TESIS DOCTORAL
SUBSAMPLING RECEIVERS
WITH APPLICATIONS TO
SOFTWARE DEFINED RADIO
SYSTEMS
José Ramón Ga cía Oya
Se illa, No iemb e de 2012
II
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
TESIS DOCTORAL
SUBSAMPLING RECEIVERS
WITH APPLICATIONS TO
SOFTWARE DEFINED RADIO
SYSTEMS
po
José Ramón Ga cía Oya
Ingenie o de Telecomunicaciones po la E.T.S. de Ingenie os
de la
Uni e sidad de Se illa
P esen ada en la
Escuela Técnica Supe io de Ingenie os
de la
Uni e sidad de Se illa
Pa a la ob ención del g ado de
Doc o po la Uni e sidad de Se illa
Se illa, No iemb e de 2012
IV
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
TESIS DOCTORAL
SUBSAMPLING RECEIVERS
WITH APPLICATIONS TO
SOFTWARE DEFINED RADIO
SYSTEMS
Au o :
José Ramón Ga cía Oya
Di ec o :
Fe nando Muñoz Cha e o

VI
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
VII
ACKNOWLEDGMENTS
Fi s ly, I would like o hank my supe iso D . Fe nando Muñoz Cha e o,
o gi ing me he oppo uni y o esea ch hese in e es ing and challenging ields,
as well as o his supe ision, guidance, op imism and suppo o become an
independen esea che .
Also I am g ea ly g a e ul o D . An onio To alba Silgado, D . Ramón
González Ca ajal and he en i e Elec onics Enginee G oup (GIE), o gi ing me
he chance o eques ing and being admi ed o a public PhD schola ship in
Uni e si y P o esso T aining and De elopmen , FPU ( om he Spanish
Fo mación de P o eso ado Uni e si a io), and o being able o de elop my
esea ch in a iendly en i onmen wi h high expe ise. This PhD schola ship was
ounded by he Minis y o Educa ion, Cul u e and Spo s (p e iously named
Minis y o Science and Inno a ion).
I would like o hank he company AT4 Wi eless, o he esea ch wo ks
join ly ca ied ou wi hin he scopes o he Telmax P ojec (PI-0553/2007), and
he Muphy P ojec (PI-0358/2009). The Telmax P ojec was pa ially unded by
CDTI–Cen o pa a el Desa ollo Tecnológico e Indus ial–, o he Spanish
Minis y o Science and Inno a ion, unde he INGENIO 2010 P og am/CENIT
call. Mo eo e , he Muphy P ojec was pa ially unded by he Andalusian
Regional Go e nmen (unde he p og am en i led “P og ama de Incen i os pa a
el Fomen o de la Inno ación y el Desa ollo Emp esa ial de Andalucía”) and he
Andalusian Technological Co po a ion (CTA).
I wish o exp ess my g a i ude o D . Da id Hely and D . Fadhel M.
Ghannouchi o accep ing me o pe o m my in e na ional in e nships in he LCIS
labo a o ies (INP, Uni e si y o G enoble) and iRadio Labs esea ch g oup
(Depa men o Elec ical and Compu e Enginee ing, Uni e si y o Calga y),
espec i ely. Specially, I would like o hank o D . Edua do Mendes (Uni e is y
o G enoble), And ew Kwan and D . Seyed Aidin Bassam (Uni e si y o Calga y)
o hei in aluable help o comple e my hesis wo k.
Finally, I wan o hank all my iends and amily, especially he people
whom his wo k is dedica ed, my pa en s and S ephanie. I ha e no enough wo ds
o hank you o e e y hing, and much less in English.
VIII Acknowledgmen s
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
IX
RESUMEN DE LA TESIS
La p esen e Tesis Doc o al p opone la u ilización sis emas basados en
submues eo como una al e na i a pa a la implemen ación de la e apa de down-
con e sion de los ecep o es de adio ecuencia empleados pa a aplicaciones
mul i-es ánda y So wa e De ined Radio. Uno de los obje i os p incipales se á el
de op imiza el diseño en cuan o a lexibilidad y simplicidad, las cuales son
p opiedades inhe en es en los sis emas basados en submues eo. Po an o, como
educi el núme o de componen es al mínimo es cla e cuando un mismo ecep o
p ocesa di e en es es ánda es de comunicación, se han seleccionado es as
a qui ec u as basadas en submues eo, pa a las que es posible alcanza un al o
g ado de eusabilidad de los componen es. De es e modo, se educi á el cos e o al
del ecep o de comunicación, así como de los equipos de es y ce i icación que
emplean es e ipo de a qui ec u as.
Un mo i o adicional po el que los sis emas basados en submues eo han
sido seleccionados es el conce nien e a la opología del ecep o . Como el obje i o
del So wa e De ined Radio es implemen a odas las uncionalidades del ecep o
( il ado, ampli icación) en el dominio digi al, el con e ido analógico-digi al
(ADC) debe á es a localizado en la cadena de ecepción lo más ce ca posible a la
an ena, siendo el obje i o inal el con e i la señal di ec amen e de RF a digi al.
Sin emba go, con los ac uales ADCs no es posible implemen a es a idea debido
al al o ancho de banda que equie en, sin pe de esolución, pa a cub i las
especi icaciones de los es ánda es de comunicaciones inalámb icas. Po an o, los
sis emas basados en submues eo se p esen an como la opción más adecuada pa a
implemen a es e ipo de ecep o es, debido a que pueden mues ea la señal de
en ada po debajo de la asa de Nyquis , si se cumplen cie as es icciones en
cuan o a la elección de la ecuencia de mues eo. De es e modo, los
eque imien os del ADC se án elajados ya que, usando las a qui ec u as
p opues as, dicho componen e p ocesa á la señal a ecuencias in e medias con
más al as p es aciones de esolución.
Una ez se han in oducido los concep os p incipales del submues eo, es a
esis doc o al p esen a el diseño de una a je a de adquisición de da os basada en
es e ipo de écnicas con la inalidad de se implemen ada como ecep o de es y
ce i icación mul i-es ánda de banda ancha. El sis ema p opues o p opo ciona una
al a esolución pa a un ele ado ancho de banda, a pa i del uso de un sample &
hold (S&H) de bajo ji e y de un ADC que abaja a ecuencias in e medias. El
p o o ipo es implemen ado usando disposi i os come ciales en una placa de
ci cui o imp eso, cuya ca ac e ización expe imen al mues a una esolución de
XVI Con en s
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS

XVII
LIST OF TABLES
Table 3.1 Valid sampling anges and op imal sampling equency o an inpu signal a
1070 MHz and signal bandwid h equal o 20 MHz .......................................................... 66
Table 4.1 Sys em pe o mance a COTS le el ................................................................ 101
Table 4.2 Sys em pe o mance a PCB le el .................................................................. 103
Table 4.3 Sys em pe o mance using mul iple clocking a COTS le el ......................... 108
Table 4.4 S anda ds speci ica ions and esul s a COTS le el ........................................ 110
Table 4.5 Sys em pe o mances using mul iple clocking a PCB le el .......................... 111
Table 4.6 S anda d speci ica ions and esul s ................................................................. 115
Table 5.1 The bounda y cons ain s o he dual band case ............................................ 124
Table 5.2 Valid sampling equencies below 2 GHz ...................................................... 126
Table 5.3 Dual band signal cons uc ion able ................................................................ 138
Table 5.4 Subsampling ecei e ’s a chi ec u es.............................................................. 140
Table 5.5 Compa a i e be ween expec ed and expe imen al SNR ................................. 143
Table 8.1 Designs ules ................................................................................................... 178
Table 8.2 Impedance calcula ion .................................................................................... 179
XVIII Lis o ables
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
XIX
LIST OF FIGURES
Figu e 1.1 Classic SDR a chi ec u e ................................................................................. 28
Figu e 2.1 SDR E olu ion ................................................................................................. 38
Figu e 2.2 Block diag am o P ima y Signal P ocessing Tasks in a ypical ansmi e and
ecei e .............................................................................................................................. 39
Figu e 2.3 Concep ual diag am o he wo-s ages down-con e sion supe he e odyne
ecei e .............................................................................................................................. 43
Figu e 2.4 Concep ual diag am o he wo-s ages down-con e sion supe he e odyne
ecei e wi h he second IF equal o DC ........................................................................... 43
Figu e 2.5 Concep ual diag am o he wo-s ages down-con e sion supe he e odyne
ecei e wi h he second IF ansla ed o DC digi ally ...................................................... 44
Figu e 2.6 Ha ley’s ecei e a chi ec u e ......................................................................... 45
Figu e 2.7 Wea e ’s ecei e a chi ec u e ........................................................................ 45
Figu e 2.8 Concep ual diag am o he ze o IF ecei e a chi ec u e ................................. 47
Figu e 2.9 Concep ual diag am o he low IF ecei e a chi ec u e.................................. 48
Figu e 2.10 Concep ual diag am o he low IF ecei e a chi ec u e wi h polyphase
il e ing .............................................................................................................................. 49
Figu e 2.11 Concep ual diag am o he double low IF ecei e ....................................... 49
Figu e 2.12 Concep ual diag am o he wideband IF ecei e a chi ec u e wi h double
con e sion ......................................................................................................................... 50
Figu e 2.13 Concep ual diag am o he subsampling ecei e a chi ec u e ...................... 51
Figu e 2.14 Concep ual diag am o he subsampling a chi ec u e wi h an in e media e
down-con e sion ............................................................................................................... 52
Figu e 2.15 Concep ual diag am o he ecei e based on in e lea ing a chi ec u e ........ 53
Figu e 2.16 Mul i-s anda d equency spec um ............................................................... 54
Figu e 2.17 Mul i-s anda d ecei e a chi ec u e by ze o IF ............................................ 55
Figu e 2.18 Mul i-s anda d ecei e a chi ec u e by low IF ............................................. 56
Figu e 3.1 A ypical wi eless link ..................................................................................... 61
Figu e 3.2 Time domain ep esen a ion o (a) 200 Hz con inuous sine wa e (b) sampled a
10 kHz and (c) sampled a 2 kHz ...................................................................................... 62
Figu e 3.3 Sampling o a signal using (a) s >> BW (b) s = BW and (c) s < BW .............. 64
Figu e 3.4 Illus a ion o he concep o subsampling: (a) F equency domain
ep esen a ion o he RF passband inpu signal along wi h he subsampling equency and
S&H ha monics and (b) signal eplicas ollowing subsampling p ocess when selec ing
s=( c- i )/k and s>BW ........................................................................................................ 65
Figu e 3.5 Ou pu spec um o he subsample when he 1070 MHz RF signal is
subsampled a a s o (a) 475.56 MHz (modd=9 and i =118.89 MHz) and (b) 480 MHz
(modd=9 and i =110 MHz) ................................................................................................. 67
XX Lis o igu es
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
Figu e 3.6 Con inuous signal spec um whe e subsampling is no posible because c-
BW<BW ............................................................................................................................ 67
Figu e 3.7 Subsampling ecei e scheme .......................................................................... 68
Figu e 3.8 Concep o ji e ............................................................................................... 68
Figu e 3.9 SNR equi emen s as a unc ion o he ji e o di e en inpu equencies ... 71
Figu e 3.10 Phase noise o a clock equency equal o 1.9 GHz ..................................... 72
Figu e 3.11 Phase noise o an oscilla o ............................................................................ 74
Figu e 3.12 Block diag am o a PLL ................................................................................. 74
Figu e 3.13 Phase noise o a PLL ..................................................................................... 75
Figu e 3.14 (a) Model o he S&H, (b) he mal noise olded in he band o in e es and (c)
e ec i e noise bandwid h.................................................................................................. 76
Figu e 3.15 The mal noise e ec depending on he sampling equency ......................... 77
Figu e 3.16 Bandpass an i-aliasing il e ing equi emen s in subsampling ...................... 78
Figu e 3.17 IF subsampling ecei e ................................................................................. 78
Figu e 3.18 Block diag am o a con inuous- ime bandpass Σ∆ modula o ....................... 79
Figu e 3.19 RF subsampling mul i-s anda d Σ∆ ecei e ................................................. 80
Figu e 3.20 Subsampling based ecei e o spec um sensing ........................................ 81
Figu e 3.21 Digi al p e-dis o ion idea .............................................................................. 82
Figu e 3.22 Dual-band digi al p edis o ion wi h subsampled eedback loop ................... 83
Figu e 4.1 Used coaxial componen s ................................................................................ 90
Figu e 4.2 S&H THD o he inpu ange 1.1-3.2 GHz ..................................................... 91
Figu e 4.3 THD measu ed and p o ided by he manu ac u e .......................................... 92
Figu e 4.4 Block diag am o he implemen ed sys em ...................................................... 94
Figu e 4.5 Implemen ed signal pa h a COTS le el .......................................................... 94
Figu e 4.6 ENOB s. inpu ampli ude o a inpu signal equency o 1001 MHz and a
sampling equency o 445.3 MHz .................................................................................... 95
Figu e 4.7 Measu ed SFDR s. inpu signal ampli ude .................................................... 96
Figu e 4.8 Measu ed IM s. inpu ampli ude .................................................................... 97
Figu e 4.9 Measu emen o he o e lapping he mal noise: ENOB ob ained o di e en
op imal sampling equencies ............................................................................................ 98
Figu e 4.10 Ou pu spec um o a 2001 MHz inpu signal subsampled a 470.8 MHz ..... 98
Figu e 4.11 Ou pu spec um o a 2001 MHz inpu signal subsampled a 216.3 MHz ..... 99
Figu e 4.12 Measu emen o he ji e noise: ENOB ob ained o di e en inpu
equencies ........................................................................................................................ 99
Figu e 4.13 ENOB s. inpu equency ........................................................................... 100
Figu e 4.14 Ou pu spec um o a 3 GHz inpu equency ............................................. 101
Figu e 4.15 Block diag am and designed PCB p o o ype ............................................... 102
Figu e 4.16 Implemen ed s ack-up .................................................................................. 103
Figu e 4.17 ENOB s. inpu equency (20 MHz signal band, up o 20 GHz inpu ca ie
equency) ....................................................................................................................... 104
Figu e 4.18 Clocking schemes o : (a) a unique clock and (b) wo di e en clocks....... 105
Figu e 4.19 Implemen ed mul iple clock sys em ............................................................ 107
Figu e 4.20 ENOB ob ained a COTS le el o a mul iple clocking a chi ec u e ........... 109
Figu e 4.21 Measu emen o he olded noise e ec (ENOB s. Op imal sampling
equencies) ..................................................................................................................... 109
Figu e 4.22 Ob ained ENOB in unc ion o he inpu equency .................................... 111
Lis o igu es XXI
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
Figu e 4.23 Mul i-s anda d ecei e a chi ec u es p oposed in [4.38] (a) and [4.39] (b) 114
Figu e 5.1 Subsampling ecei e in mul i-band nonlinea en i onmen ......................... 122
Figu e 5.2 (a) F equency loca ions in he sampled ou pu spec um and (b) Spec um o
he dual band RF signal a he inpu o he S&H wi h a io R1= 2/ 1 ............................... 123
Figu e 5.3 Powe spec um a he inpu ( op) and he ou pu (bo om) o a nonlinea
sys em ............................................................................................................................. 125
Figu e 5.4 Subsampled spec um o 1.82 and 2.4 GHz inpu equency ....................... 126
Figu e 5.5 Subsampling applica ions o mul i-band and nonlinea sys ems in he
ansmi e and ecei e sides .......................................................................................... 127
Figu e 5.6 (a) Subsampling based ecei e o spec um sensing in cogni i e adio
sys ems and (b) measu emen se up o alida ing spec um sensing concep using
subsampling ecei e ...................................................................................................... 128
Figu e 5.7 Spec a o (a) he inpu RF signal o he ecei e and (b) he subsampled RF
signal o bands (698-752 MHz, 902-928 MHz) using a subsampling equency o 255
MHz ................................................................................................................................ 129
Figu e 5.8 Spec a o he inpu and il e ed ou pu baseband signals o he 698-752 MHz
band (a) and 902-928 MHz band (b) ............................................................................... 129
Figu e 5.9 The (a) p edic ed RF undamen al and ha monics up o 4 GHz and (b)
subsampled esul using a sampling equency o 619.8 MHz ....................................... 131
Figu e 5.10 (a) RF spec a a he ou pu o he PA and (b) no malized spec a o he
cap u ed subsampled signal using an ADC ope a ing a 619.8 MHz ............................. 132
Figu e 5.11 Folded noise e ec s using single clock (a) and mul iple clock (b) ............. 133
Figu e 5.12 Folded e ec s o ha monics and in e modula ion p oduc s using a single
clock (a) and mul iple clock (b) echniques .................................................................... 134
Figu e 5.13 Op imized a chi ec u e based on mul iple clocking and BP il e s ............. 134
Figu e 5.14 Algo i hm low diag am o compu ing op imal subsampling equencies in
he p oposed a chi ec u e ................................................................................................ 136
Figu e 5.15 Expec ed SNR (a) o single and mul iple clock a chi ec u es and (b) o
di e en a chi ec u es based on BP il e s ...................................................................... 139
Figu e 5.16 Expe imen al se up o dual band subsampling ecei e ............................. 140
Figu e 5.17 Simula ed spec a a e wo-s age subsampling p ocess, using a S&H
subsampling equency o 1900 MHz, ADC subsampling equency o 400 MHz, and
signal bands a 2.12 GHz and 2.4 GHz ........................................................................... 141
Figu e 5.18 Expe imen al spec a a e wo-s age subsampling p ocess, using a S&H
subsampling equency o 1900 MHz, ADC subsampling equency o 400 MHz, and
signal bands a 2.12 GHz and 2.4 GHz ........................................................................... 141
Figu e 5.19 Theo e ical SNR o he p oposed a chi ec u es .......................................... 142
Figu e 5.20 Expe imen al SNR o he p oposed a chi ec u es ( o design 2 and 4, no sub-
sampling equency could be ound o scena ios 1, 2 and 3) ........................................ 142
Figu e 7.1 Ou pu spec um o 2-in e lea ed ADCs ( c=6 GHz, s=2.8 GHz) ................. 156
Figu e 7.2 Ou pu spec um o 3-in e lea ed ADCs ( c=6 GHz, s=2.3 GHz) ................. 156
Figu e 7.3 Ou pu spec um o 4-in e lea ed ADCs ( c=4 GHz, s=1.6 GHz) ................. 157
Figu e 7.4 Ou pu spec um wi hou implemen ing calib a ion ...................................... 158
Figu e 7.5 Ou pu spec um a e calib a ing wo couples o ADCs ............................... 159
Figu e 7.6 Measu ed esolu ion o ou in e lea ed ADCs ........................................... 159

XXII Lis o igu es
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
Figu e 7.7 (a) Th ee ADCs (ABC) o h ee imes sampling a e, (b) Th ee ADCs (ABC)
o a double sampling a e ............................................................................................... 161
Figu e 7.8 Example o a s uc u e wi h ex a ADCs ....................................................... 162
Figu e 7.9 Random clock o 5 ADCs ............................................................................. 162
Figu e 7.10 Func ional diag am using digi al il e s blocks ............................................ 164
Figu e 7.11 A chi ec u es based on: (a) one S&H and (b) se e al sub-S&H .................. 166
Figu e 7.12 A chi ec u e based on double sampling ....................................................... 166
Figu e 8.1 Block diag am o he p oposed sys em .......................................................... 172
Figu e 8.2 BGA dimensions o he Inphi 1821TH ......................................................... 172
Figu e 8.3 Pinou o E2V AT84AS001 ........................................................................... 173
Figu e 8.4 P oposed schema ic: page 1 ........................................................................... 174
Figu e 8.5 Th u-line loss calib a ion ............................................................................... 174
Figu e 8.6 P oposed schema ic: page 2 ........................................................................... 175
Figu e 8.7 P oposed schema ic: page 3 ........................................................................... 176
Figu e 8.8 P oposed schema ic: page 4 ........................................................................... 176
Figu e 8.9 P oposed schema ic: page 5 ........................................................................... 177
Figu e 8.10 P oposed schema ic: page 6 ......................................................................... 177
Figu e 8.11 P oposed s ack-up ........................................................................................ 181
Figu e 8.12 D ill cha ..................................................................................................... 182
Figu e 8.13 VIA wi h minimum size .............................................................................. 182
Figu e 8.14 ADC anou .................................................................................................. 183
Figu e 8.15 VIAs a ay s cuc u e ................................................................................... 183
Figu e 8.16 VIA employed in SMA connec ions ............................................................ 184
Figu e 8.17 VIA employed in powe supplies connec ions ............................................ 184
Figu e 8.18 Top unc ionali y .......................................................................................... 185
Figu e 8.19 Top unc ionali y (II) ................................................................................... 185
Figu e 8.20 Bo om unc ionali y .................................................................................... 186
Figu e 8.21 Powe supplies in laye VCC1 ..................................................................... 186
Figu e 8.22 Powe supplies in laye VCC2 ..................................................................... 187
Figu e 8.23 P o o ype dimensions ................................................................................... 187
XXIII
LIST OF ACRONYMS
ADC Analog o Digi al Con e e
AGC Au oma ic Gain Con ol
ASIC Applica ion Speci ic In eg a ed Ci cui
AWGN Addi i e Whi e Gaussian Noise
BER Bi E o Ra e
BGA Ball G id A ay
BoM Bill o Ma e ials
BP Band Pass
BPS Band Pass Sampling
BW Bandwid h
CMOS Complemen a y Me al-Oxide Semiconduc o
COTS Comme cial O The Shel
CPLD Complex P og ammable Logic De ice
CR Cogni i e Radio
DAC Digi al- o-Analog Con e e
DDC Digi al Down Con e e
DDS Di ec Digi al Syn hesize
DPD Digi al P e-Dis o e
DSP Digi al Signal P ocessing
EMI Elec omagne ic In e e ence
ENOB E ec i e Numbe O Bi s
FFT Fas Fou ie T ans o m
FPGA Field P og ammable Ga e A ay
XXIV Lis o ac onyms
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
FR4 Flame Re a dan 4
GP Gene al Pu pose
IC In eg a ed Ci cui
IF In e media e F equency
IMD3 3 d o de In e modula ion Dis o ion
I-Q In-phase and Quad a u e
IRF Image Rejec ion Fil e
LNA Low Noise Ampli ie
LO Local Oscilla o
LP Low Pass
LPS Low Pass Sampling
LSB Leas Signi ican Bi
LVDS Low-Vol age Di e en ial Signaling
MOS Me al-Oxide Semiconduc o
NF Noise Figu e
PA Powe Ampli ie
PCB P in ed Ci cui Boa d
PLL Phase-Locked Loop
PSD Powe Signal Densi y
RD Recei e Design
RF Radio F equency
SAW Su ace Acous ic Wa e
SDR So wa e De ined Radio
SFDR Signal F ee Dynamic Range
SMA Subminia u e e sion A
SMD Su ace Moun De ice
SNDR Signal- o-Noise and Dis o ion Radio
Lis o ac onyms XXV
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
SNR Signal- o-Noise Ra io
S&H Sample & Hold
THD To al Ha monic Dis o ion
UWB Ul a Wideband
VCO Vol age Con ol Oscilla o
VCXO Vol age Con ol C ys al Oscilla o
VIA Ve ical In e connec Access
VLSI Ve y-La ge Scale In eg a ion
WPAN Wi eless Pe sonal A ea Ne wo k
32 Chap e 1: In oduc ion
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
1.4 Re e ences
[1.1] R. Baghe i e al., “An 800 MHz-6 GHz So wa e-De ined Wi eless
Recei e in 90 nm CMOS,” IEEE Jou nal o Solid-S a e Ci cui s, j.41,
no.12, pp. 2860-2876, Dec. 2006.
[1.2] F. Agnelli e al., “Wi eless Mul i-S anda d Te minals: Sys em
Analysis and Design o a Recon igu able RF F on -End,” IEEE
Ci cui s and Sys ems Magazine, ol. 6, no.1, pp. 38-59, Jan. 2006.
[1.3] F. S el o, M. B. Vahid a , M. B andolini, “Recon igu able Si RF
Recei e F on -Ends o Mul is anda d Radios”, 1s Eu opean
Con e ence on Wi eless Technology, (EuWiT 2008), pp.33-36, 2008.
[1.4] M. B andolini, P. Rossi, D. Mans e a, F. S el o, “Towa d
Mul is anda d Mobile Te minals-Fully In eg a ed Recei e s
Reque imen s and A chi ec u es”, IEEE T ansac ions on Mic owa e
Theo y and Techniques, ol. 53, no. 3, Ma ch 2005.
[1.5] M. Vidojko ic, M. A. T. Sanduleanu, V. Vidojko ic, J. an de Tang,
P. Bal us, A. H. M. an Roe mund, “A 1.2V Recei e F on -End o
Mul i-S anda d Wi eless applica ions in 65nm CMOS LP”, 34 h
Eu opean Solid-S a e Ci cui Con e ence (ESSCIRC 2008), pp. 414-
417, 2008.
[1.6] R. Ba ak, A. Ghazel, F. Ghannouchi, “Op imized Mul is anda d RF
Subsampling Recei e A chi ec u e,” IEEE T ansac ions on Wi eless
Communica ions, ol. 8, no. 6, pp. 2901-2909, Jun. 2009.
[1.7] J. R. G. Oya, F. Muñoz, A. To alba, A. Ju ado, A. J. Ga ido, J.
Baños, “Da a Acquisi ion Sys em Based on Subsampling o Tes ing
Wideband Mul is anda d Recei e s," IEEE T ansac ions on
Ins umen a ion and Measu emen s, ol. 60, no. 9, pp. 3234-3237, Sep.
2011.
[1.8] J. R. G. Oya, F. Muñoz, A. To alba, A. Ju ado, F. Má quez, E.
López-Mo illo, “Da a Acquisi ion Sys em Base on Subsampling using
Mul iple Clocking Techniques,” IEEE Ins umen a ion and
Measu emen s, ol. 61, no. 8, pp. 2333-2335, Aug. 2012.
[1.9] J. Mi ola, “The So wa e Radio A chi ec u e” IEEE Communica ions
Magazine, ol. 33, no. 5, pp. 26-38, May 1995.
[1.10] F. Ha is, R.W. Lowde milk, “So wa e De ined Radio: Pa 22 in a
Se ies o Tu o ials on Ins umen a ion and Measu emen ,” IEEE
Ins umen a ion & Measu emen , ol. 13, no. 1, pp. 23-32, Feb. 2010.

Chap e 1: In oduc ion 33
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
[1.11] S. Haykin, “Cogni i e Radio: B ain Empowe ed Wi eless
Communica ions”, IEEE Jou nal on Selec ed A eas in
Communica ion, j. 48, no. 2, pp. 201-220, 2005.
[1.12] F. K. Jond al, “So wa e-de ined adio: basics and e olu ion o
cogni i e adio,” IEEE EURASIP, ol.3, pp. 275–283, Aug. 2005.
[1.13] N. Vun, A. B. P emkuma , “ADC Sys ems o SDR Digi al F on -
End,” P oceedings o he Nin h In e na ional Symposium on
Consume Elec onics (ISCE 2005), pp. 359-363, 2005.
[1.14] C. R. Ande son, S. Venka esh, J. E. Ib ahim, R. M. Bueh e , J. H.
Reed, “Analysis and Implemen a ion o a Time-In e lea ed ADC
A ay o a So wa e-De ined UWB Recei e ,” IEEE T ansac ions on
Vehicula Technology, ol. 58, no. 8, pp. 4046-4063, Oc . 2009.
[1.15] M. B. Romdhane, P. Loumeau, “Analog o Digi al Con e sion
speci ica ions o Ul a Wide Band ecep ion,” P oceedings o he
Fou h IEEE In e na ional Symposium on signal P ocessing and
In o ma ion Technology, pp. 157-160, 2004.
[1.16] A.A. Abidi, “Di ec -con e sion adio anscei e s o digi al
communica ions," IEEE Jou nal o Solid-S a e Ci cui s, ol. 30, no.
12, Dec. 1995.
[1.17] J. C ols, M. S eyae , “Low-IF opologies o high-pe o mance
analog on -ends o ully in eg a ed ecei e s," IEEE Jou nal o
Solid-S a e Ci cui s, ol. 45, no. 3, Ma . 1998.
[1.18] D. G ace, S. P. Pi , “Quad a u e sampling o high equency
wa e o ms," Jou nal o he Acous ical Socie y o Ame ica, ol. 44, pp.
1432-1436, 1968.
[1.19] R. Vaughan, N. Sco , D. Whi e, “The Theo y o Bandpass Sampling,”
IEEE T ansac ions on Signal P ocessing, ol. 39, no. 9, pp. 1973-
1984, Sep. 1991.
[1.20] A. Kwan, S. A. Bassam, F. M. Ghannouchi, “Sub-sampling Technique
o Spec um Sensing in Cogni i e Radio,” IEEE Radio and Wi eless
Symposium (RWS’2012), pp. 347-350, 2012.
[1.21] S. A. Bassam, A. Kwan, W. Chen, M. Helaoui, F. Ghannouchi,
“Subsampling Feedback Loop Applicable o Concu en Dual-Band
Linea iza ion A chi ec u e,” IEEE T ansac ions on Mic owa e Theo y
and Techniques, ol. 60, no.6, pa 2, pp. 1990-1999, 2012.
34 Chap e 1: In oduc ion
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
35
CHAPTER
2
OVERVIEW OF SOFTWARE
DEFINED RADIO AND MULTI-
STANDARD RECEIVER
ARCHITECTURES
CHAPTER CONTENTS
2.1 So wa e de ined adio sys ems ...................................................................... 37
2.1.1 So wa e de ined adio idea and e olu ion ............................................................. 37
2.1.2 So wa e de ined adio a chi ec u e ........................................................................ 38
2.1.3 Bene i s and incon eniences o he so wa e de ined adio .................................... 40
2.1.4 In oduc ion o cogni i e adio ............................................................................... 41
2.2 Recei e a chi ec u es .................................................................................... 42
2.2.1 Supe he e odyne ecei e ....................................................................................... 42
2.2.2 Ze o-IF ecei e ...................................................................................................... 46
2.2.3 IF ecei e s ............................................................................................................. 47
2.2.3.1 Low IF ecei e s ................................................................................................ 47
2.2.3.2 Doble low IF ecei e s ...................................................................................... 49
2.2.3.3 Wideband IF ecei e s wi h double con e sion ................................................ 50
2.2.4 Subsampling ecei e .............................................................................................. 50
2.2.5 Recei e s based on in e lea ing ............................................................................. 53
36 Chap e 2: O e iew o so wa e de ined adio and ecei e a chi ec u es
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
2.2.6 Mul i-s anda d ecei e s ......................................................................................... 54
2.2.6.1 Mul i-s anda d ze o IF ecei e ......................................................................... 54
2.2.6.2 Mul i-s anda d low IF ecei e .......................................................................... 55
2.3 Re e ences ....................................................................................................... 56
This chap e is dedica ed o desc ibing he con ex and he applicabili y o
his hesis. A i s sec ion e iews he main concep s abou SDR, desc ibing i s
e olu ion and de ailing i s bene i s and he cu en p oblems ha a oid ge ing
his pa adigm. This sec ion p esen s an in oduc ion o cogni i e adio as well. A
second sec ion e iews he main ecei e a chi ec u es, ocusing on hei main
ad an ages and disad an ages when implemen ed as mul i-s anda d ecei e s.
This sec ion concludes wi h he con enience o using subsampling a chi ec u es.
The chap e inalizes desc ibing se e al published mul i-s anda d ecei e
a chi ec u es.
Chap e 2: O e iew o so wa e de ined adio and ecei e a chi ec u es 37
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
2.1 So wa e de ined adio sys ems
2.1.1 So wa e de ined adio idea and e olu ion
A SDR adio is a communica ion sys em ha pe o ms mos o i s signal
p ocessing asks in a p og ammable digi al signal p ocessing (DSP), being ully
adap able by eal ime downloadable so wa e, and enabling he adjus men o
a ious communica ions scena ios au oma ically, adap ing o di e en egional
in e aces as well [2.1]. In o he wo ds, SDR means a adio whe e unc ionali y
and signal p ocessing a e de ined in so wa e, and i suppo s mul i-band mul i-
use adio communica ion.
These sys ems will ha e a key ole in u u e adio con igu a ions because
he eme gence o new wi eless echnologies, and he necessi y o in eg a ion o a
la ge numbe o communica ion s anda ds in mul i-s anda d and mul iband adios
in o de o implemen a uni e sal handse ha p o ides wo ldwide access.
Some bene i s a op le el a e, o subsc ibe s, an easie in e na ional
oaming, imp o ed and mo e lexible se ices, and inc eased pe sonaliza ion. Fo
mobile ne wo k ope a o s, some op-le el bene i s a e he po en ial o apidly
de elop and in oduce new, pe sonalize, and cus omized se ices [2.2].
A SDR sys em uses a single ha dwa e on end bu can be ep og ammed
by so wa e i s equency o ope a ion, occupied bandwid h, and adhe ence o
se e al wi eless s anda ds by calling a ious so wa e algo i hms, allowing
inexpensi e, e icien in e ope abili y, and inc easing lexibili y ia inc eased
p og ammabili y. A he same ime, SDR a chi ec u es simpli y ha dwa e
componen adeo s and p o ide new ways o managing he complexi y o he
eme ging s anda ds.
The SDR so wa e ep og ams he DSP segmen in o de o econ igu e
he sys em and, hus, implemen mul iple adios. This segmen pe o ms he signal
p ocessing and condi ions he signal o be modula ed and demodula ed, being he
p ocessing engine a combina ion o gene al pu pose (GP) mic op ocesso s,
applica ion speci ic in eg a ed ci cui s (ASICs), ield p og ammable ga e a ays
(FPGAs) and specialized co-p ocesso s [2.3]. The DSP sys em is coupled o he
ai in e ace and an enna by analog- o-digi al and digi al- o-analog con e e s, as
shown in Figu e 1.1.
No e ha , al hough he SDR pa adigm allows a single e minal o adap o
mul iple adio in e ace s anda ds by so wa e con e ing di ec ly o digi al
domain a e he an enna, an e olu ion p ocess is s ill ac i e, which has been
desc ibed and p edic ed. This e olu ion is illus a ed in he ecei e o Figu e 2.1
and is e e ed o he pa o he adio a chi ec u e ha is co e ed by he so wa e

38 Chap e 2: O e iew o so wa e de ined adio and ecei e a chi ec u es
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
p ocessing ools [2.2]. The e o e, new da a con e sion echniques a e cu en ly
esea ched in o de o ge his pa adigm in he u u e.
RF
F on -
end
Analog
IF A/D
D/A
Baseband
Modem
P ocessing Bi s eam
P ocessing Da a
In e ace
RF
F on -
end
Analog
IF A/D
D/A
Baseband
Modem
P ocessing Bi s eam
P ocessing Da a
In e ace
RF
F on -
end
Digi al
IF
P ocessing
A/D
D/A
Baseband
Modem
P ocessing Bi s eam
P ocessing Da a
In e ace
So awa e p ocessing
So awa e p ocessing
So awa e p ocessing
Figu e 2.1 SDR E olu ion
On he o he hand, he e a e di e en download mechanisms [2.1]: s a ic,
pseudo-s a ic and dynamic so wa e download. S a ic download is he si ua ion
whe e SDR can suppo a a ie y o s anda ds and is p og ammed in a s a ic
manne o add ess one o hese possibili ies. This means a i s s ep abou
econ igu abili y capabili ies. Pseudo-s a ic download e e s o using he ai
in e ace o download and p e-con igu e a e minal o accommoda e a de ined se
o applica ions, and p o ocols. This op ion inc eases he lexibili y o he adio
o e he s a ic op ion and, mo eo e , his upg ade can appea anspa en o he
use . Finally, he dynamic op ion o e s a highe lexibili y, due o allowing he
econ igu a ion du ing, o ins ance, a call, p o iding a conc e e con igu a ion on
demand.
2.1.2 So wa e de ined adio a chi ec u e
In his sec ion a gene al scheme o a SDR anscei e will be desc ibed,
p esen ing he main unc ionali ies implemen ed in bo h sides, i.e., ansmi e and
ecei e . Since his wo k is o ien a ed o he ecei e implemen a ion, in ollowing
Chap e 2: O e iew o so wa e de ined adio and ecei e a chi ec u es 39
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
sec ions he mos con enien a chi ec u es o his pa o he sys em will be
desc ibed mo e conc e ely.
A block diag am o he signal p ocessing unc ionali ies in a SDR
anscei e is illus a ed in Figu e 2.2. This diag am anks o asks known as
P ima y Signal P ocessing Tasks [2.3], which includes equency ansla ion,
il e ing and da a con e sion. Two mo e signal p ocessing g oups a e classi ied
o a SDR sys em, one dedica ed o synch oniza ion and ano he one dedica ed o
imp o e he esponse o he di y RF [2.4] (Seconda y and Te ia y Signal
P ocessing Tasks, espec i ely). Bo h o hem a e desc ibed b ie ly in he end o
his sec ion.
S-PI-Q
Table
Digi al
Low Pass
Digi al
Low Pass
Digi al
Low Pass
Digi al
Low Pass
DAC
IF
S age RF
S age
PA
Gain
Con ol
IF
S age
ADC
Digi al
Low Pass
Digi al
Low Pass
Digi al
Low Pass
Digi al
Low Pass
DDS Ca ie
PLL
De ec S-P
Timing
PLL
DDS
LNA VGA
Bi s
Modula o
Demodula o
Channel
Shape &
Upsample
In e polla e
IF
Analog Analog
RF Ca ie
RF Ca ie
Analog
Ca ie
Wa e o m
Decima e
Ma ched
Fil e
Bi s
Figu e 2.2 Block diag am o P ima y Signal P ocessing Tasks in a ypical
ansmi e and ecei e
Cen e ing he desc ip ion in he ecei e side (Figu e 2.2), he ecei ed
signal will be ampli ied and ansla ed o sui able in e media e equencies (IF)
p io o being con e ed o digi al domain. Howe e , his a chi ec u e is only a
i s app oxima ion because he SDR pa adigm leads o implemen his down-
con e sion digi ally, i.e., placing he ADC jus a e he an enna. The con e ed
signal is sampled by he demodula o , which down-con e s he IF-cen e ed signal
wi h a digi al down-con e e (DDC). The cu en baseband-cen e ed signal has a
sample a e abo e he equi ed o sa is y he Nyquis c i e ia so i will be educed
by he decima ing il e . The e o e, hese wo p ocesses mi o and cancel he up-
con e sion and he in e pola ion1, espec i ely, implemen ed in he ansmi e
side. Finally, he educed sample- a e signal is used as inpu o he de ec o and is
p ocessed o maximize i s ou pu SNR.
1 Usually his ou pu sample a e is ixed a some su icien ly high alue ha can be used o a
la ge IF ange.
40 Chap e 2: O e iew o so wa e de ined adio and ecei e a chi ec u es
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
Al hough i is possible o obse e om Figu e 2.2 ha he ansmi e and
he ecei e side a e almos complemen a y, he ecei e also pe o ms a numbe
o asks no p esen in he ansmi e , such as o es ima e unknown pa ame e s o
he ecei ed signal as ampli ude, equency o se , o iming o se . To implemen
hese unc ionali ies some elemen s as a channel equalize , a digi al au oma ic
gain con ol (AGC), a DC cancelle , and a SNR es ima o a e necessa y in he
ecei e pa , al hough hey a e omi ed om he block diag am by simplici y.
On he o he hand, some o he Seconda y Signal P ocessing Tasks in he
ecei e a e he modula ion ca ie alignmen , symbol clock alignmen , and
scaling he ecei ed signal wi h i s SNR. In ac , when he alignmen p ocess is
pe o med p ope ly, he ecei e is able o collec all he ene gy in he ecei e
signal, es ima ing he ansmi ed wa e o m’s ampli ude wi h maximum SNR.
Consequen ly, SDR adios wo k o e a la ge ange o s essed signal le els.
Finally, he Te ia y Signal P ocessing Tasks help o minimize he
p oblems caused by he ole ance, and gain and phase imbalances om he analog
wo ld. Also, o he non ideali ies con ibu e o he signal deg ada ion. These
e ec s include he non linea i y in he powe ampli ie , ADC and mixe s, DC
o se s and coupled spec al lines, oscilla o phase noise, sampling ape u e ji e
and clock ji e . The e o e, he ecei e includes some compensa ing p ocessing
DSP blocks o educe he e ec s om he analog componen s. These blocks a e
DC cancel, Phase and Gain Balance, and Channel Equaliza ion.
2.1.3 Bene i s and incon eniences o he so wa e de ined adio
An SDR has ad an ages in cos and pe o mance. Since almos all he
unc ionali ies a e pe o med in he digi al domain, hese designs a e less
expensi e o manu ac u e, due o he Moo e’s Law, and o e s a be e gene al
pe o mance as well as educed sensi i i y o age, empe a u e and en i onmen al
in luences. Analog componen s such as esis o s o capaci o s a e manu ac u ed
wi h speci ic ole ances. Also ci cui s designed om analog componen s su e
om misma ch and imbalance e ec s, which limi he pe o mance o he sys em.
Mo eo e , no e ha , al hough an SDR design uses digi al echniques,
di e s om a DSP adio (also called so wa e-con olled digi al adios [2.1]) in
ha i s adio pa ame e s a e no ixed, i.e., hey a e econ igu able. The main
ad an age o his econ igu abili y is ha an SDR sys em suppo s
communica ions be ween a wide ange o communica ion sys ems. This
p og ammabili y includes p og ammable RF bands, channel access modes, and
channel modula ion. Mo eo e , as new wa e o ms, ea u es and s anda ds a e
de eloped and inco po a ed, he SDR can be ep og ammed, h ough so wa e
upg ades, o inc ease i s capabili ies and be a new adio. These bene i s will
educe he cos s delaying he obsolescence o he communica ion sys ems.
The e o e, in applica ions whe e access o mul iple bands wi h mul iple adio
Chap e 2: O e iew o so wa e de ined adio and ecei e a chi ec u es 41
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
access modes is a necessi y, he SDR can educe ha dwa e size, complexi y and
powe h ough ewe adio uni s.
Howe e , SDR is s ill a adio and, he e o e, he ules o design a good
adio mus be applied as well. As his design equi es a la ge bandwid h and
dynamic ange, he DSP sec ion will need educed le els o di y RF [2.4] in he
analog ai in e ace. The phase noise, he ji e , he le el o he hi d o de
in e cep will ha e o imp o e because DSP by i sel canno epai all hese
sou ces o signal deg ada ion in oduced by he analog componen s.
Ano he RF p oblem in SDR is he necessi y o a oid he in oduc ion o
p ocesso clock ha monics in o he analog RF and IF ci cui s. Also, when a
mul iple ansmi e is implemen ed elec omagne ic in e e ence (EMI) p oblems
will occu . Ne e heless, some o hese incon eniences a e p esen in mul iple
ha dwa e adios as well.
Mo eo e , i is di icul o enginee wideband, low loss an ennas, and RF
and da a con e e s, so new echniques o implemen he anscei e s mus be
de eloped in o de o co e mos wi eless communica ion s anda ds placing he
da a con e e s (ADCs and DACs) as close o he an enna as possible. The p esen
wo k will be ocused in hese objec i es, abou he op imal con e sion echniques
in he ecei e side.
An addi ional p oblem o an SDR sys em is ha , due o i s lexibili y, i
can be used o pe o m unc ions ha a e p ohibi ed by legal es ic ions, like o
ansmi in unlicensed equency bands. This p oblem mus be add essed as well.
Finally, ano he known d awback is he di icul y o place he ADC igh
a e he an enna, due o he cu en ADC speci ica ions. This p oblem will be
add ess h oughou his hesis wo k.
2.1.4 In oduc ion o cogni i e adio
A his poin i is necessa y o in oduce he cogni i e adio idea. Due o
ano he unc ionali y o he SDR is i s capabili y o adap i sel he ansmission
scena io in o de o minimize he in e e ence wi h o he signals in he ai
in e ace, he sys em will equi e he abili y o scan he spec um om low o high
equency using so wa e. Wi h his objec i e in mind, he idea o CR [2.5] is o
build on an SDR, whe e he adio adap s i sel o he en i onmen by op imizing
he ca ie equency, modula ion, and choice he adio s anda d o minimize
in e e ence and main ain communica ion in a gi en scena io. One o he mos
p omising objec i es o CR is o inc ease he spec um occupancy, he adio
u ilizing spec um ha is no used by o he adio a his momen . Ano he main
objec i e will be o be able o implemen a highly eliable communica ion
whene e and whe e e needed. The h ee undamen al cogni i e asks can be
ound desc ibed in [2.6]:
48 Chap e 2: O e iew o so wa e de ined adio and ecei e a chi ec u es
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
The nex down con e sion om low IF o baseband is implemen ed in he
digi al domain, a oiding he p oblems caused by he I-Q misma ches in he analog
domain.
BPF LNA LO1
Ampli ie
RF
90º LPF
LPF
Ampli ie
ADC
ADC
sin(ωLO2 ) DSP
cos(ωLO2 )
cos(ωLO2 )
Figu e 2.9 Concep ual diag am o he low IF ecei e a chi ec u e
This s uc u e is also simple and s ill allows a high le el o in eg a ion and
does no su e om he ze o-IF a chi ec u e p oblems (DC o se s o licke
noise) because he desi ed signal is no olded o DC. Howe e , in his case, he
image equency disad an age is ein oduced, due o being di icul o ejec i ,
hus being he majo d awback o his a chi ec u e. Bo h image and desi ed signal
will be digi ized by he ADC. Thus a digi al il e ing o channel-selec ion will be
implemen ed by he DSP as well. An addi ional p oblem is abou he ADC powe
consump ion, which is inc eased because now a highe con e sion a e is equi ed.
Ano he op ion o elimina e he image in e e e is o use an IF polyphase
a e he down-con e sion, as shown in Figu e 2.10 [2.7]. This makes his
a chi ec u e e y sui able o mul i-s anda d ecei e s, since his il e can be
sha ed by di e en s anda ds. Any changes in he RF equencies can be sol ed by
using co esponding LO equencies and down-con e ing he RF signal o he
same IF, il e ing he co esponding image by he polyphase il e .

Chap e 2: O e iew o so wa e de ined adio and ecei e a chi ec u es 49
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
BPF LNA
Ampli ie
RF
LPF
LPF
Ampli ie
ADC
ADC
sin(ωLO2 ) DSP
cos(ωLO2 )
cos(ωLO2 )
cos(ωLO1 )
sin(ωLO1 )
Polyphase
Fil e
Figu e 2.10 Concep ual diag am o he low IF ecei e a chi ec u e wi h
polyphase il e ing
2.2.3.2 Doble low IF ecei e s
The basic idea o hese ecei e s, whose block diag am is illus a ed in
Figu e 2.11, is o up con e he I-Q channels, gene a ed p e iously a IF, o a
high equency using a ixed equency syn hesize [2.15]. The signal hen eeds
o an IF il e , usually o chip, being i s in eg a ion le el lowe han in he low IF
case.
The double low IF a chi ec u e is mo e immune o DC p oblems, like in
he low IF case. Also simila o a low IF ecei e , he close p oximi y o he
image signal means i s supp ession is no possible by only he RF il e . Al hough,
hese ecei e s would be limi ed o s anda ds wi h mode a e adjacen channels o
s ingen equi emen s o he IRF, unlike he low IF ecei e , he signal is up
con e ed o a high IF, whe e a e y high-Q disc e e il e is employed o emo e
he image. This il e will con ibu e o he supp ession o he image in
conjunc ion wi h low pass il e a he i s IF s age, p o iding a highe selec i i y
pe o mance and a comp omise be ween selec i i y and in eg a ion le el.
BPF LNA LO1
RF
90º LPF
LPF
Ampli ie
ADC
LO2
90º BPF
Figu e 2.11 Concep ual diag am o he double low IF ecei e
50 Chap e 2: O e iew o so wa e de ined adio and ecei e a chi ec u es
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
2.2.3.3 Wideband IF ecei e s wi h double con e sion
A ecei e which uses a RF channel-selec equency syn hesize and an IF
o baseband channel-selec il e is a na owband ecei e . An al e na i e
a chi ec u e based on IF ecei e s uc u es o use o mul i-s anda d applica ions
is he wideband IF ecei e (Figu e 2.12), whe e he en i e RF band con aining he
in o ma ion is ansla ed o IF by mul iplying he LO ou pu wi h a ixed
equency [2.15]. All he channels a IF a e hen ansla ed o DC using a unable,
channel-selec LO and, inally, he selec ed low pass channel eeds he ADC.
P e iously a a iable gain has been p o ided. As in he case o ze o IF ecei e s,
channel il e ing can be pe o med a baseband, whe e digi al p og ammable
il e s can enable mo e mul i-s anda d ecei e ea u es.
This app oach is simila o supe he e odyne ecei e s in ha he equency
ansla ion is accomplished in mul iple s eps. Howe e , unlike a con en ional
supe he e odyne a chi ec u e, he i s LO equency ansla es he whole RF band,
main aining a la ge bandwid h a IF. Mo eo e , he wideband IF ecei e has an
addi ional ad an age abou a highe capabili y o acili a e he syn hesize
in eg a ion han he es o ecei e s desc ibed p e iously. The e o e, he
in eg a ion o his a chi ec u e is also easible, al hough he I-Q misma ches
p oblems in he analog domain a e ein oduced. Simila o he ze o IF ecei e s,
o he incon enience o wideband IF a chi ec u es is i s suscep ibili y o licke
noise, DC o se and dis o ion due o second o de in e modula ion.
BPF LNA LO1
RF
90º LPF
LPF
ADC
sin(ωLO2 )
cos(ωLO2 )
cos(ωLO2 )
Ampli ie
LPF
LPF
ADC
Ampli ie
IF
Figu e 2.12 Concep ual diag am o he wideband IF ecei e a chi ec u e wi h
double con e sion
2.2.4 Subsampling ecei e
A easible al e na i e o he p e ious solu ions is he ecei e based on
subsampling, which is illus a ed in Figu e 2.13. The ecei ed signal is il e ed by
a RF band pass il e ha can be a unable il e o a bank o il e s. The incoming
band pass signal is sampled unde Nyquis c i e ia [2.16,2.17], using some
Chap e 2: O e iew o so wa e de ined adio and ecei e a chi ec u es 51
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
sampling p ope ies and a oiding aliasing, as is desc ibed in Chap e 2. This
sampled signal is con e ed o digi al using an ADC a in e media e sampling a e.
The main ad an age o his scheme is i s simplici y, he numbe o componen s
being educed, and being possible o place he da a con e sion close o he
an enna. The e o e, lo s o unc ions like il e ing, equency ansla ion and
demodula ion can be implemen ed in digi al domain, aking ad an age o low cos
digi al VLSI solu ions, leading o a high in eg a ion and elimina ing p oblems
such as DC o se , 1/ noise.
In addi ion o sys em cos educ ion, pushing hese unc ions in digi al
domain elimina es he many o he sensi i i ies o analog solu ions, such de ice
ma ching, en i onmen al sensi i i y, and pe o mance a ia ion o e ime. The
lexibili y and econ igu abili y equi ed by SDR applica ions is also inc eased by
mo ing he ADC in o IF s age and, mo eo e , i is possible o use his a chi ec u e
o wideband and mul i-s anda d applica ions because o i s la ge analog
bandwid h. In his a chi ec u e, a single ADC can sample mul iple signal channels,
which a e hen sepa a ed and demodula ed in pa allel in digi al domain.
Howe e , some c i ical equi emen s exis when his s uc u e is employed,
as he needed analog inpu bandwid h o he S&H. Since his bandwid h mus
include he RF ca ie equency, he bandwid h o he S&H inside he ADC
canno be la ge enough o he equi ed dynamic ange, esolu ion and sample a e
using he cu en echnologies. A common solu ion is o place a p e ious ex e nal
S&H.
BPF LNA S&H ADC
BPF
RF
s/2
Figu e 2.13 Concep ual diag am o he subsampling ecei e a chi ec u e
An in e media e al e na i e, abou simplici y and S&H equi emen s, is o
do a p e ious ansla ion o IF, as shown in Figu e 2.14. This concep is called IF-
sampling [2.18], being his digi al-IF a chi ec u e a s ep owa d SDR idea whe e
he las down-con e sion s age in he e odyne ecei e s is eplaced by an ADC
s age.
I he Nyquis heo em is me he sampling unc ion is called Low Pass
Sampling (LPS) and he IF signal is di ec ly sampled and con e ed by he ADC,
52 Chap e 2: O e iew o so wa e de ined adio and ecei e a chi ec u es
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
a oiding he I-Q misma ches p oblems. Howe e , since he IF signal ha e o be
la ge, in o de o a oid he il e ing ejec ion p oblems, he ADC equi emen s
will be e y demanding. The e o e, o his a chi ec u e, o sample wi hou
mee ing Nyquis heo em (only wi h a sampling equency highe han wice he
in o ma ion banwid h) allows o ob ain mo e elaxed ADC. This sampling
unc ion is called Band Pass Sampling (BPS) and i is equi alen o ealize a down
con e sion mixing by spec al olding, as shown in Chap e 2. This scheme will
a oid he I-Q misma ches p oblems as well.
BPF LNA S&H ADC
BPF
RF
s/2
LO
BPF
IF
Figu e 2.14 Concep ual diag am o he subsampling a chi ec u e wi h an
in e media e down-con e sion
Finally, subsampling ecei e s ha e addi ional p oblems as some noise
sou ces, as he ji e and he mal noise olded in he in e es band, inally hese
e ec s being minimized in his hesis wo k.
Mo eo e , RF band pass il e ing is equi ed when a oiding o e lap
be ween olded signals is necessa y. These BP il e s, especially on-chip il e s,
a e di icul o implemen a high equencies. Al hough ex e nal il e s, such as
SAW il e s, can be used, hey a e only a ailable a limi ed numbe o equencies,
so i is no a p ac ical solu ion o design mul i-s anda d ecei e s.
Al e na i ely, highe sampling equency is o en used o educe he
equi ed selec i i y. Howe e , his solu ion has some d awbacks as he high
echnology and high cos equi ed by he ADC, whose esolu ion and dynamic
ange will be deg aded as compa ed o lowe sample a e ADC al e na i es. Also
powe consump ion is inc eased wi h sample a e. The e o e, he cos ,
pe o mance, and powe consump ion o o he de ices (such as ADC clock
sou ces, digi al ci cui s a e he ADC) also will be impac ed by he ADC sample
a e. In his hesis some no el echniques, abou he sampling equency plan, a e
add essed in o de o a oid his o e lapping be ween signals, educing he
complexi y o he RF il e ing. On he o he hand, addi ional adjacen in e e e s
no o e lapped wi h he desi ed signal can be supp essed by addi ional channel
il e ing in digi al domain.
Chap e 2: O e iew o so wa e de ined adio and ecei e a chi ec u es 53
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
2.2.5 Recei e s based on in e lea ing
An al e na i e scheme o he subsampling echniques, in o de o place he
analog- o-digi al con e sion close o he an enna, is a ime-in e lea ed ADC
a chi ec u e. This scheme is an e ec i e app oach o achie ing e y high
sampling a es [2.19,2.20]. A ime-in e lea ed ADC ope a es M pa allel ADCs a
di e en sampling imes, c ea ing he image o a single ADC ope a ing a a much
highe sampling a e. The concep is illus a ed in Figu e 2.15 [2.20]. Ideally, he
i h ADC, i = 0, ..., M – 1, samples pe iodically he inpu signal a ime ins an s
i, i+M, i+2M, wi h sample a e s/M, whe e m=mTs and Ts=1/ s is he sampling
pe iod o he ime-in e lea ed ADC. The inal ou pu is c ea ed by mul iplexing
all o he indi idual ADC ou pu s in he p ope o de (e.g. ADC0, ADC1, …,
ADCM – 1, ADC0, ADC1, e c.).
The eby, he inal e ec is as i he inpu signal we e sampled once e e y
Ts seconds, i.e., wi h sample a e s. This app oach has been widely adop ed in he
indus y, since he con e e s can be wo king a lowe speeds wi hou sac i icing
he o e all sys em pe o mance.
Howe e , i should be no ed ha each indi idual ADC deals wi h he
en i e analog inpu signal, and, he e o e, i s S&H ci cui mus be able o p ese e
he ull inpu signal bandwid h. This is he main incon enien o use hese
a chi ec u es o mul i-s anda d ecei e s, besides he misma ch be ween ADCs,
being his hesis ocused mainly in subsampling echniques. Ne e heless, an
app oach o ecei e s based on in e lea ing is add essed in Appendix A.
Figu e 2.15 Concep ual diag am o he ecei e based on in e lea ing a chi ec u e

54 Chap e 2: O e iew o so wa e de ined adio and ecei e a chi ec u es
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
2.2.6 Mul i-s anda d ecei e s
S acking se e al ecei e s o di e en s anda ds in o a single ecei e ,
ope a ing in pa allel, is no a easible op ion o implemen a mul i-s anda d
ecei e because he a ea and he powe consump ion would be ex emely high.
The e o e, a mul i-s anda d ecei e should sha e he a ailable ha dwa e esou ces
as possible and make use o he unable and p og ammable de ices, inc easing he
le el o in eg a ion. F om he iew o high le el in eg a ion, he ze o IF ecei e ,
low IF ecei e , wideband IF ecei e and subsampling ecei e a e mos sui able,
while he double IF ecei e p esen s a comp omise be ween selec i i y and
in eg a ion le el.
F om he iew o placemen o he ADC, bo h he ze o IF ecei e and he
subsampling ecei e a e candida es o SDR implemen a ion because he ADC
di ec ly has an in e ace o RF o highe IF signals. Howe e , due o some
incon eniences o ze o IF ecei e s, as DC o se o LO leakage, his hesis will
be add essed o subsampling ecei e s.
Rega ding he adio sec ion, no e ha he di e en s anda ds equi e
di e en on -end pe o mance. The mos s aigh o wa d solu ion would be
sa is ying he mos c i ical speci ica ions o each one. Some o he mo e ypical
s anda ds ha need o be co e ed by mul i-s anda d ecei e s a e illus a ed in
Figu e 2.16.
GHz
GSM
1 65432
GPS
GSM
UMTS/W-CDMA
Blue oo h
IEEE 802.11b/g
WiMAX
IEEE802.11a
Figu e 2.16 Mul i-s anda d equency spec um
In his sec ion wo examples o con en ional mul i-s anda ds ecei e s a e
desc ibed. The mul i-s anda d ecei e based on subsampling, designed in his
hesis wo k, is desc ibed in he ollowing chap e s and compa ed wi h o he
published subsampling ecei e s o mul i-s anda d applica ions.
2.2.6.1 Mul i-s anda d ze o IF ecei e
Figu e 2.17 [2.21] illus a es a single-chip mul i-mode ecei e o ou
s anda ds (GSM900, DCS1800, PCS1900 and W-CDMA), which was designed in
a ze o IF scheme. An ex e nal digi al con olle selec s he di e en s anda ds and
he ha dwa e is sha ed as much as possible by di e en s anda ds. These s anda ds
Chap e 2: O e iew o so wa e de ined adio and ecei e a chi ec u es 55
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
use wo di e en channel selec ion il e s, and he quad a u e LO signals a e
ob ained by using a di ide-by- wo-ci cui o he mixe s. Since he LO signal is
gene a ed on-chip, he LO leakage on he PCB o he RF inpu , which is a ypical
p oblem o homodyne ecei e s, is elimina ed mo e e icien ly. Howe e , he
le el in eg a ion will be lowe in his case.
Mul i-Band
LNA
1ADC
DCS/PCS/GSM
WCDMA
1ADC
DCS/PCS/GSM
WCDMA
/2
WCDMA
GSM900
DCS1800
PCS1900
LO
Figu e 2.17 Mul i-s anda d ecei e a chi ec u e by ze o IF
2.2.6.2 Mul i-s anda d low IF ecei e
Figu e 2.18 [2.22] illus a es ano he ully in eg a ed mul i-s anda d
ecei e designed in low IF, wi h se e al SAW BP il e s and a mul i-band LNA
p eceding his a chi ec u e. This design suppo s i e wi eless communica ion
s anda ds, Blue oo h, GSM (DCS1800 o Eu ope, o PCS1900 o USA), UMTS,
802.11b/g and 802.11a. I is possible o obse e in he diag am block how he
Blue oo h channel is ac i e all he ime, while he o he ou s anda ds, which
co e i e di e en equency bands, a e ac i a ed by an RF swi ch be o e eeding
he es o he low IF ecei e due o how hey do no need o be co e ed a he
same ime, i.e., when an applica ion is ac i e, he o he s can be swi ched o o in
idle mode, in o de o sa e powe and euse ha dwa e esou ces. O he wise,
Blue oo h needs o ope a e concu en ly o o he s anda ds, allowing o ha e
ac i a ed he wi eless link du ing a phone call o da a communica ion.
56 Chap e 2: O e iew o so wa e de ined adio and ecei e a chi ec u es
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
0º90ºLO
ADC
ADC
Mul i-Band
LNA
Blue oo h SAW
Blue oo h Recei e
Pa h
GSM SAW
UMTS SAW
IEEE 802.11b/g SAW
IEEE 802.11a SAW
Figu e 2.18 Mul i-s anda d ecei e a chi ec u e by low IF
In his sec ion a compa a i e o mul i-s anda d ecei e s abou
digi aliza ion echniques (i.e., mixing o subsampling based sys ems) ha e been
in oduced. The e is a second esea ch ield abou he band s a egy (i.e.,
wideband o na ow band s a egy). While he examples desc ibed in his sec ion
[2.21,2.22] a e na owband mul i-s anda d ecei e s because o hey employ
dedica ed channels, o he published wo ks [2.23-2.26] can be conside ed uni e sal
ecei e s, co e ing a la ge inpu equency ange. Al hough hese wideband
solu ions a e mo e lexible, hei main incon enience is he RF on -end mus
mee he equi emen s o each s anda d and hey a e no op imum o any
s anda d. These al e na i e ecei e s will be s udied mo e p ecisely when hey a e
compa ed wi h he wo ks desc ibed in Chap e 3.
2.3 Re e ences
[2.1] J. Mi ola, “The So wa e Radio A chi ec u e” IEEE Communica ions
Magazine, ol. 33, no. 5, pp. 26-38, May 1995.
[2.2] W. H. W. Tu lebee, “So wa e-De ined Radio: Face s o a
De eloping Technology,” IEEE Pe sonal Communica ions, ol. 6, no.
2, pp. 38-44, 1999.
[2.3] F. Ha is, R.W. Lowde milk, “So wa e De ined Radio: Pa 22 in a
Se ies o Tu o ials on Ins umen a ion and Measu emen ,” IEEE
Ins umen a ion & Measu emen , ol. 13, no. 1, pp. 23-32, Feb. 2010.
[2.4] G. Fe weis, M. Lohning, D. Pe o ic, M. Windisch, P. Zillmann, and
W. Ra e, “Di y RF: A new Pa adigm,” In e na ional. Jou nal o
Wi eless In o ma ion Ne wo ks, ol. 14, no. 2, pp. 133-148, Jun. 2007.
Chap e 2: O e iew o so wa e de ined adio and ecei e a chi ec u es 57
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
[2.5] J. Mi ola, G. Q. Magui e, “Cogni i e adio: Making so wa e adios
mo e pe sonal,” IEEE Pe sonal Communica ions, ol. 6, no. 4, pp.
13–18, Aug. 1999.
[2.6] S. Haykin, “Cogni i e Radio: B ain Empowe ed Wi eless
Communica ions”, IEEE Jou nal on Selec ed A eas in
Communica ion, j. 48, no. 2, pp. 201-220, 2005.
[2.7] L. Zhang, “Sys em and Ci cui Design Techniques o WLAN-
Enabled Mul i-S anda d Recei e ,” Doc o al Disse a ion, Ohio S a e
Uni e si y, USA, 2005.
[2.8] Y. R. Sun, “Gene alized Bandpass Sampling Recei e s o So wa e
de ined Radio,” Doc o al Disse a ion, School o In o ma ion and
Communica ion Technology (ICT), S ockholm, Sweden, 2006.
[2.9] J. R. Macleod, M. A. Beach, P. A. Wa , T. Nesimoglu, “A So wa e
De ined Radio Recei e Tes -bed,” IEEE Vehicula Technology
Con e ence (VTC 2001), ol. 3, no. 2, pp. 1565-1569, Fall 2001.
[2.10] R. Ha ley, “Modula ion Sys em,” U.S. Pa en 1,666,206, Ap . 1928.
[2.11] D. Wea e , “A Thi d Me hod o Gene a ion and De ec ion o Single-
Sideband Signals,” P oceedings o he IRE, pp 1703-1705, Dec. 1956.
[2.12] A.A. Abidi, “Di ec -con e sion adio anscei e s o digi al
communica ions," IEEE Jou nal o Solid-S a e Ci cui s, ol. 30, no.
12, Dec. 1995.
[2.13] J. C ols, M. S eyae , “Low-IF opologies o high-pe o mance
analog on -ends o ully in eg a ed ecei e s," IEEE Jou nal o
Solid-S a e Ci cui s, ol. 45, no. 3, Ma . 1998.
[2.14] P. C uz, N. B. Ca alho, K. Remley, “Designing and Tes ing
So wa e-De ined Radios,” IEEE Mic owa e Magazine, ol. 11, no. 4,
pp. 83-94, 2010.
[2.15] J. C. Rudell, “F equency T ansla ion Techniques o High-In eg a ion
High-Selec i i y Mul i-S anda d Wi eless Communica ion Sys ems,”
Doc o al Disse a ion, Uni e si y o Cali o nia, Be keley, USA, Fall
2000.
[2.16] D. G ace, S. P. Pi , “Quad a u e sampling o high equency
wa e o ms," Jou nal o he Acous ical Socie y o Ame ica, ol. 44, pp.
1432-1436, 1968.
64 Chap e 3: Subsampling ecei e s
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
Choosing a sampling equency o a band limi ed signal2 a ec s he
econs uc ion p ocess. A band limi ed signal wi h o al bandwid h (BW) is
deno ed in Figu e 3.3. Once he con inuous signal is ep esen ed by a sequence o
disc e e sample alues, i s spec um akes a eplica ed o m, wi h hese eplicas
sepa a ed s. In o he wo ds, a con inuous signal canno be ep esen ed in a digi al
machine in i s cu en band limi ed o m.
Fo Figu e 3.3a, he sampling equency s=1/Ts=ωs/2π is much la ge han
he bandwid h, and pe ec econs uc ion is possible because he in ini e
eplica ions a e no aliased. Simila ly, o he case when s=BW as in Figu e 3.3b,
he spec ums do no o e lap, o alias, o e each o he and he signal can s ill be
decoded p ope ly. Howe e , in Figu e 3.3c, s is less han BW, and aliasing occu s
o e he signal. This aliasing co up s he in o ma ion in he signal and is
un eco e able.
The minimum sampling a e, o he Nyquis sampling a e, should be s >=
BW in o de o co ec ly decode he signal. In p ac ice, an an i-aliasing il e (LP
il e ) will be necessa y be o e he analog- o-digi al con e sion in o de o
elimina e any ene gy signal loca ed abo e BW/2 o below –BW/2 ha would be
olded o e he band o in e es .
X( ) X( ) X( )
(a) (b)
s
- s s2 s
- s
-2 s s2 s
- s
-2 s(c)
Figu e 3.3 Sampling o a signal using (a) s >> BW (b) s = BW and (c) s < BW
3.3 Subsampling heo y
3.3.1 Concep o subsampling
As men ioned be o e, mo ing he ADC close o he an enna inc eases he
lexibili y o he ecei e . Howe e , his con e sion jus a e he an enna would
p ohibi i ely inc ease he bandwid h and sampling equency equi emen s o he
ADC. Ne e heless, he bandwid h o a bandpass signal is usually a ac ion o i s
cen e equency, so ha i is possible o subsample he signal (i.e., iola ing he
Nyquis condi ion) a oiding aliasing be ween eplicas.
Subsampling is he p ocess o sampling a signal wi h a equency lowe
han wice he highes signal equency, and highe han he signal bandwid h BW.
Using an ideal S&H de ice wi h sampling equency s will gene a e ha monics a
2 F om a p ac ical s andpoin , he e m band-limi ed signal me ely implies ha any signal ene gy
ou side he ange [-BW/2,BW/2] is below he sensi i i y o he sys em.

Chap e 3: Subsampling ecei e s 65
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
s, 2 s...m s, whe e m is an in ege . In he case o Figu e 3.4a, a bandpass RF
signal is cen e ed a c, while he m h closes ha monic gene a ed by he S&H and
lowe han c is k, whe e k = loo ( c/ s). The eplicas o he signal ha a e
gene a ed by he S&H exis a -m s + c, while he mi o ed e sions eplicas exis
a (m + 1) s – c. Figu e 3.4b shows hese eplicas, and he signal eplica wi hin he
[0- s/2] ange (cen e ed a i = c-k s) can be used o ex ac he o iginal RF signal.
(a)
(b)
sk s(k+1) s
BW
- s
-k s
-(k+1) s
s/2 3 s/2
c
c
- s/2
-3 s/2
- c
- c
(k+1/2) s
-(k+1/2) s
Figu e 3.4 Illus a ion o he concep o subsampling: (a) F equency domain
ep esen a ion o he RF passband inpu signal along wi h he subsampling
equency and S&H ha monics and (b) signal eplicas ollowing subsampling
p ocess when selec ing s=( c- i )/k and s>BW
3.3.2 Selec ing he sampling equency
This sec ion p o ides he me hod o selec he op imal sampling equency
( s) o a gi en signal bandwid h (BW) and ca ie equency ( c). Usually, he
minimal sampling equency is de e mined by he Nyquis Theo em:
s>2( c+BW/2). Howe e , o a bandpass signal a sampling equency lowe han
he Nyquis equency can be selec ed i equa ion (3.16) s ill holds [3.3]:
mBW mBW csc /)2/(2)1/()2/(2
(3.16)
whe e m is he numbe o eplicas o he signal spec um in he ange [0, c-
BW/2], and lies be ween 1 and loo (( c+BW/2)/BW). An app op ia e alue is
s=4 c/modd which p oduces a eplica a s/4 and some imes i is called “op imal”
sampling equency. Using an odd in ege modd ensu es ha he signal is a s/4,
while me en gene a es he low equency alias o he signal a 3 s/4.
As an example, o a gi en inpu signal a 1070 MHz, wi h a signal
bandwid h equal o 20 MHz, Table 3.1 shows i s i s en alid anges and i s i s
en op imal sampling equencies.
On he o he hand, an example ha illus a es he con enience o sampling
a 4 c/modd can be obse ed in Figu e 3.5, which shows he ou pu spec um when
an inpu signal a 1070 MHz is sampled a s o 475.56MHz (Figu e 3.5a) and a
nea by equency o 480 MHz (Figu e 3.5b). I can be seen how he second o de
66 Chap e 3: Subsampling ecei e s
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
ha monic a 237.78 MHz is placed u he om away he desi ed signal in he case
o Figu e 24a compa ed wi h Figu e 3.5b (a 220 MHz). The e o e, sampling a
4 c/modd esul s in a la ge subsampling equency bandwid h and elaxes he
il e ing equi emen s a e he S&H. As s, c and i a e all di ec ly ela ed, he e
a e bandwid h and equency adeo s when selec ing he subsampling equency.
Table 3.1 Valid sampling anges and op imal sampling equency o an inpu
signal a 1070 MHz and signal bandwid h equal o 20 MHz
Uppe Limi (MHz)
Lowe limi (MHz)
Op imal equency (MHz)
2040
1120
1462.7
1020
746.7
856
680
560
611.4
510
448
475.6
408
373.3
389.1
340
320
329.2
291.4
280
285.3
255
248.9
251.8
226.7
224
225.3
212
196.4
203.8
Chap e 3: Subsampling ecei e s 67
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
(a) (b)
237.78 MHz
118.89 MHz 110 MHz
220 MHz
Figu e 3.5 Ou pu spec um o he subsample when he 1070 MHz RF signal is
subsampled a a s o (a) 475.56 MHz (modd=9 and i =118.89 MHz) and (b) 480
MHz (modd=9 and i =110 MHz)
Ano he op ion in o de o elax he ADC equi emen s could be o place
he eplica a a lowe equency. Howe e , in p ac ice, high pe o mance ADCs
a e ine- uned du ing design and manu ac u e o ensu e maximum linea i y a
equency, hei use a lowe equencies no being ecommended. Mo eo e , o
sample close o he limi s o he alid ange could no be p uden because he
analog band pass il e s ha e non ideali ies and sample a es a he clock gene a o
can p esen ins abili ies. The e o e, in hese cases i would be con enien o
conside a band gua d be added o he signal bandwid h [3.4].
Finally, he e is a case when subsampling is no possible. I a con inuous
bandpass signal’s lowes equency is less han he bandwid h, we ha e a no
pe missible si ua ion. This condi ion is shown in Figu e 3.6 whe e c-BW<BW.
The e is no way o old any spec al eplicas be ween his lowes equency
componen and DC. In his case, i is only possible o sample he signal mee ing
Nyquis c i e ia, i.e., using a sample a e o a leas wice he highes equency
componen , i.e., s >2( c+BW).
c
- c c+B/2
c-B/2
BB
Figu e 3.6 Con inuous signal spec um whe e subsampling is no posible because
c-BW<BW
68 Chap e 3: Subsampling ecei e s
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
3.4 Non ideali ies in subsampling
A gene al scheme o a subsampling ecei e is shown in Figu e 3.7. I
dese es o be men ioned ha his ecei e is e y simple, especially i i is
compa ed o he con en ional he e odyne a chi ec u e. Howe e , as he S&H
p ocesses high equency signals, i s equi emen s a e much mo e es ic i e han
hose expec ed om he signal bandwid h. The main non-ideali ies o be
conside ed in he S&H a e ji e and olded he mal noise and will be desc ibed in
he sub-sec ions.
S&H ADC Digi al
P ocessing
Figu e 3.7 Subsampling ecei e scheme
3.4.1 Ji e and phase noise
3.4.1.1 Phase noise
Clock ji e is an impo an limi a ion in he da a acquisi ion sys ems a
high signals equencies because leads o sampling ime unce ainly. Ji e is he
de ia ion o he e e ence edges o he clock signal wi h espec o hei ideal
posi ion in ime. In his chap e we will conside his de ia ion as a andom noise.
As shown in Figu e 3.8, a andom e o τn om he nominal sampling ime ins an
n causes a andom e o ετ(n) in he ampli ude o he sampled signal [3.5]. This
e ec can be seen as an addi ion o noise o he ou pu signal, esul ing in a
deg ada ion o he ou pu Signal- o-Noise Ra io (SNR).
y( )
y( n)
y( n+τn)
n n+τn
τn
ετ(n)
Figu e 3.8 Concep o ji e
The ampli ude e o ( e o ) is p opo ional o he de i a i e o he inpu
signal [3.5,3.6]:
Chap e 3: Subsampling ecei e s 69
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
d
d
in
e o
(3.17)
Wi h a ji e alue o ∆ . Fo a sine wa e o equency in and ampli ude Ain,
he maximum e o is [3.5,3.6]:
inine o A 2
max_
(3.18)
The e a e wo main sou ces o ji e noise, he phase noise associa ed o he
clock e e ence and he ape u e ji e o he S&H. The ape u e ji e o an S&H
implemen ed wi h a MOS ansis o is signal-dependen , as he ansis o
h eshold ol age depends on he inpu signal. Conce ning he sys em clock, he e
a e wo p ima y mechanisms ha cause ji e : he he mal noise and he coupling
noise. The la e can be caused by c oss alk and/o g ound loops wi hin, o
adjacen o, he a ea o he ci cui . Special ca e has o be aken designing he
powe lines in he da a acquisi ion boa d ha will be desc ibed in Chap e 3.
In a i s o de app oach, hese wo sou ces o ji e noise can be
conside ed as unco ela ed Gaussian s ochas ic p ocesses, each one wi h a
pa icula s anda d de ia ion. Being ∆ ms he s anda d de ia ion o ji e (o oo
mean squa e), which usually de ined as a pe cen age o he sampling pe iod, he
sampling e o in equa ion (3.17) can be e-w i en as [3.5]:
2
2)/()( in
in msin mse o A
d d
(3.19)
whe e σ() is he s anda d de ia ion.
The e o e, he esul ing SNR on he sampled signal is hen [3.5]:
)2log(20
2
2
2/
log20 in ms
in
in ms
in
ji e
A
A
SNR
(3.20)
This app oxima ion will be ue i 2π in∆ ms<<1, o he wise he gene al
exp ession o he SNR due o he unco ela ed andom ji e noise o a sinusoidal
inpu signal can be exp essed as ollows [3.7,3.8]:
o he wisee
SNR msin
msin msin
ji e :)1(2/1
12:4/1
log20 222
2
222
(3.21)
The exp ession o SNR o 2π in∆ ms<<1 is alid o all ji e dis ibu ions
while he o he SNR exp ession only applies o a andom ji e wi h Gaussian
dis ibu ion N(0, ∆ ms) [3.7]. Mo eo e , small ji e noise can be app oxima ely

70 Chap e 3: Subsampling ecei e s
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
ega ded as sampled Addi i e Whi e Gaussian Noise (AWGN) while o la ge
ji e , his assump ion is no alid anymo e.
No e ha he SNR is deg aded when he inpu equency inc eases. This
SNR will be added o he SNR in he analog- o-digi al con e sion s age (SNRADC
in equa ion (3.22) [3.5]), i.e., he deg ada ion caused by he mal noise and
quan iza ion noise.
1010 1010log20
ji e
ADC SNR
SNR
To al
SNR
(2.22)
In he pa icula case o subsampling, ji e noise is an impo an limi a ion
due o high inpu equencies a e p ocessed. Thus, in o de o alida e his
heo e ical s udy, ji e noise is simula ed using ypical alues o ecei e s
subsampling based, i.e., he inpu equencies in he GHz ange, he sampling
equencies a ound 500 MHz (which is a ypical limi o high esolu ion
comme cial ADCs, as is desc ibed in Chap e 3), and a 20 MHz signal bandwid h,
due o how i is a ypical alue o many communica ions s anda ds and is used o
cha ac e ize expe imen ally he da a acquisi ion boa d p oposed in Chap e 3.
The e o e, using hese alues, ji e noise has been simula ed (using
MATLAB) as a s ochas ic p ocess wi h a e age equal o ze o and s anda d
de ia ion equal o ∆ ms. Figu e 3.9 [3.9] illus a es he maximum admi ed ji e
(Axis X) o ob ain a conc e e SNR (Axis Y) o h ee di e en inpu equencies
(1, 2 and 4 GHz) sampling a he op imum equency ( om equa ion s=4 c/modd)
immedia ely lowe han 500 MHz. In his example, he ji e noise is in eg a ed in
a signal bandwid h equal o 20 MHz and i can be obse ed how he SNR will
dec ease a ound 6 dB each ime he inpu equency is doubled, as can be
p edic ed by equa ion (3.21).
Chap e 3: Subsampling ecei e s 71
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
x 10-12
40
45
50
55
60
65
70
75
80
Des iación ípica de ji e (seg)
SNDR (dB)
c=2GHz, s=470.59MHz, BW=20MHz
c=1GHz, s=444.44MHz, BW=20MHz
c=4GHz, s=484MHz, BW=20MHz
Ape u e ji e (s)
SNR (dB)
Figu e 3.9 SNR equi emen s as a unc ion o he ji e o di e en inpu
equencies
3.4.1.2 Phase noise
This adi ional me hod, based on equa ion (3.22) o ob ain he SNR as a
unc ion o clock ji e and signal equency, has some limi a ions, as he
assump ion o a ull scale scena io. Al hough his si ua ion may happen in some
applica ions, mos commonly he inpu signal ene gy is sp ead o e some
bandwid h. In hese cases i is mo e ealis ic o s udy he ji e e ec om he
spec um domain.
Since he spec um o ji e is e y di icul o measu e di ec ly, he mos
common me hod o s udy i s e ec is by measu ing he phase noise, which is he
mos widely employed pa ame e o compa e be ween di e en clock sou ces and
oscilla o s.
The phase noise is de ined as he equency domain ep esen a ion o he
phase modula ion o he clock signal due o he ji e . The clock signal being a
sine wa e o equency s [3.5]:
))(2sin()))((2sin( A A ssclock
(3.23)
whe e υ( )is he phase noise in he ime domain. Assuming υ( ) has small
a ia ion a ound ze o, equa ion (3.23) can be w i en as [3.5]:
)()2cos()2sin( A A ssclock
(3.24)
The second e m o his exp ession is he addi i e noise due o he phase
modula ion. Since he phase noise appea s mul iplied by a cosine in he abo e
ime domain exp ession, in he equency domain he spec um o he phase noise,
Φ( ), is con ol ed wi h he noise- ee clock and appea s as sidebands a ound i s
cen e equency. This noise is usually ep esen ed as L( ) (single-sideband phase
72 Chap e 3: Subsampling ecei e s
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
noise powe spec um) and is equal o he noise powe spec al densi y pe He z
a he equency s+ no malized by he clock o oscilla o signal powe A2/2. I is
called single-sideband because only one side o he noise powe is aken in o
accoun ; hence i includes only hal he noise ene gy. Thus [3.5]:
10
)(
2
102)(
)(
2
1
log10)(
L
L
(3.25)
L( ) is ep esen ed in dBc/Hz. Figu e 3.10 [3.1] shows an example o phase
noise o a clock equency equal o 1.9 GHz, which is a ypical equency ange
o he S&H clock sou ce in he implemen ed sys ems, as i will be de ailed in he
ollowing sec ions. These expe imen al measu emen s show a phase noise o
a ound -95 dBc/Hz, -110 dBc/Hz and -125 dBc/Hz a 100 Hz, 1 KHz and 10 KHz
espec i ely.
Figu e 3.10 Phase noise o a clock equency equal o 1.9 GHz
F om equa ion (3.23), he ela ionship be ween υ( ) and ji e is [3.5]:
)(2)( sss kT kT
(3.26)
Tha is equi alen o e e encing ji e o he clock pe iod. In he equency
domain, whe e he clock phase noise is mos commonly ep esen ed, i is hen
equal o he clock ji e scaled by 2π s [3.5]:
)(2)( T s
(3.27)
The e o e, we ha e he ollowing exp ession o ob ain he o al ji e om
phase noise [3.5]:
Chap e 3: Subsampling ecei e s 73
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
0
10/)(
0
2102
2
1
)(
2
1d
d
L
ss
ms
(3.28)
Finally, he SNR deg ada ion can be calcula ed om he phase noise
measu emen s as well. Assuming a sine wa e as inpu , [3.5,3.10] calcula es he
exp ession o he ol age sampling e o due o he ji e in he equency domain:
)(
2
2)( in
in
ine o T
A
V
(3.29)
I we subs i u e (3.27) in (3.29) he ollowing exp ession is ob ained:
)(
2
)( in
in
s
in
e o
A
V
(2.30)
In o de o ob ain he in-band noise ha will a ec he SNR, equa ion 30
o e he sys em pass band ( min, max), employing (3.25), esul s3:
max
min
max
min
10/)(
2
2
10log20
)(
2
log10
L
s
in
e o
in
ji e d
d V
A
SNR in
(3.31)
On he o he hand, when an oscilla o is used as clock gene a o i is
necessa y ake accoun ha i s phase noise is composed o wo main egions as
illus a ed in Figu e 3.11 [3.5,3.11,3.12]. The la ge egion is due o he he mal
noise whose e ec is simila o a equency modula ion, gene a ing sidebands ha
all in e sely p opo ional wi h equency o se , in a slope o -20 dBc/dec. A low
equency o se s, he e is a egion wi h a slope o -30 dB/dec due o up-
con e sion o 1/ noise. The co ne equency be ween wo egions, x, is
dependen on he oscilla o implemen a ion. Finally, he la cu e is usually
called “whi e” phase noise [3.11] and he dashed cu e illus a es he combina ion
o he desc ibed cu es.
3 No ice ha he e is no ac o 2 because he in eg a ion is o e only one sideband.
80 Chap e 3: Subsampling ecei e s
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
ad an age o using a con inuous Σ∆ modula o in on end is o educe he noise
o subsampling p ocess wi hou inc easing he powe consump ion o he ADC.
Mo eo e , using he con inuous- ime il e o he modula o he
equi emen s o he an i-aliasing il e be o e he modula o will be less es ic ed.
O he wise, wi h he pu pose o ansla ing he di e en RF bands o same IF, his
implemen a ion selec s a sampling equency such as he equency dis ance
be ween he op imal IF ( s/4) and he IF o each RF band is minimal.
The Σ∆ modula ion (whose block diag am is simila as shown in Figu e
3.18) is ollowed by a digi al p og ammable decima ion il e in o de o emo e
he unwan ed componen s equency and educe he o e sampled a e o he
Nyquis a e o he channel. This decima ion il e ing is implemen ed by using
mul iple s ages, each s age being designed o he desi ed band o each s anda d.
Finally, each s anda d is down-con e ed o baseband h ough he I-Q
pa hs and con olled by a nume ical oscilla o . A e a LP il e ing s age he
baseband signal is digi ally p ocessed.
RF mul iband
il e LNA
Con inuous- ime
BP Σ∆ modula o Decima ion
il e
Con inuous- ime subsampling BP Σ∆ ADC
LO
90º LPF
LPF
DSP
Figu e 3.19 RF subsampling mul i-s anda d Σ∆ ecei e
Ano he in e es ing applica ion based on subsampling is he impulse adio
a chi ec u es used in ul a wideband (UWB) adio [3.20,3.21]. UWB p esen s a
high a ie y o applica ions, including imaging, su eillance, high-speed da a
communica ion and high- esolu ion loca ion [3.22,3.23]. A common applica ion
ha wo ks in he 3.1-10.6 GHz band (wi h a minimum bandwid h o 500 MHz) is
indoo communica ion, which has implemen a ions in high speed sho dis ances
(less han 10 m) sys ems o wi eless pe sonal a ea ne wo k (WPAN), o in low
da a a e communica ions, such as senso ne wo ks. [3.20] p oposes o implemen
a low complexi y 3.1-10.6 GHz UWB sys em which ansmi s passband pulses
using a pulse and an an enna, and down-con e s he ecei ed signal ia
subsampling.
A challenge o UWB is o ully exploi he ea u es o he wideband adios
o low powe and low cos designs in o de o inc ease he e iciency o
na owband sys ems. The e o e, a easible al e na i e o implemen hese sys ems
will be he subsampling a chi ec u es. In his case, he ecei e chain will p ocess
non sinusoidal ca ie s, so called impulse adios.

Chap e 3: Subsampling ecei e s 81
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
Mo eo e , UWB has a ela i ely low ecei ed SNR, due o he signal
ansmission powe is limi ed by egula ions, a la ge in-band noise due o he
wideband ci cui noise and possible in e e ences. As a esul , a UWB ecei e
equi es only a mode a e ADC esolu ion, i.e., 4-6 bi s [3.19]. Thus, he
quan iza ion noise is mo e dominan han he olded he mal noise, being he
subsampling ecei e a e y p omising a chi ec u e o UWB. Fo he same eason,
he ji e cons ain will be much less s ingen .
On he o he hand, [3.21] p esen s a lexible subsampling ecei e based
on line spec um es ima ion echniques, which a e applied in he equency
domain in o de o eco e he posi ion and he ampli ude o he ecei ed pulses
[3.24]. These pulses will be dis o ed by he anscei e an ennas and he channel,
being his pulse dis o ion e y a iable among he mul ipa hs employed in hese
sys ems. The e o e, delay lines will be implemen ed in he analog domain in o de
o equalize he di e en pa hs. On he o he hand, digi al based ecei e s p o ide
mo e lexibili y and accu acy, bu equi e ADCs sampling a Nyquis a e which
a e ha dly ealizable and highly powe consuming. An a chi ec u e based on
in e lea ed ADCs could be a easible op ion bu he a ea is inc eased and i
equi es app op ia e echniques o compensa e ci cui misma ches be ween he
pa allel b anches. As a esul , a subsampling a chi ec u e is an al e na i e o
implemen his applica ion because i p o ides he lexibili y o a digi al design
wi hou inc easing he powe consump ion o a ea.
Mo eo e , some applica ions speci ically o ien ed o mul iband and non
linea sys ems employ subsampling echniques. This sec ion in oduces a couple
o examples which wo k in hese scena ios. Howe e , hese applica ions will be
s udied in mo e de ail in Chap e 4, due o he necessi y o addi ional es ic ions
in o de o a oid he o e lapping be ween signals and ha monics.
Fi s ly, [3.25] p esen s a subsampling ecei e o cogni i e adio
applica ions. The main u ili y o his mul iband sys em, whose block diag am is
illus a ed in Figu e 3.20, is o scan he wi eless spec um o know wha slo s in
he whole spec um a e unde u ilized in o de o be managed mo e e icien ly and
o able o be eused dynamically. This ecei e p esen s he ad an ages p ope o
he subsampling based sys ems, such as, low complexi y and lexibili y, bu he e
will be an addi ional challenge in o de o a oid he o e lapping be ween di e en
slo s when his spec um is subsampled.
Bank o BP
il e s
S&H ADC Spec um sensing uni
(baseband
p ocessing)
Figu e 3.20 Subsampling based ecei e o spec um sensing
82 Chap e 3: Subsampling ecei e s
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
Secondly, an example o a subsampling applica ion in a non linea
scena io is p esen ed in [3.26]. The p oposed sys em consis s o he
implemen a ion o a subsampling ecei e in he eedback loop o a ansmi e .
This eedback loop is u ilized o accomplish he digi al p e-dis o ion (DPD)
linea iza ion p ocess o he powe ampli ie [3.27,3.28], whose idea is illus a ed
in Figu e 3.21.
This concep is based on ex ac ing he beha io al model o he nonlinea
ansmi e , es ima ing i s in e se beha io model, in o de o p e-dis o he
digi al baseband signal (p e iously o be ansmi ed) o compensa e he ex ac ed
e ec s in he powe ampli ie . The e o e, since g()= -1() in Figu e 3.21, i is
possible o ob ain he inpu signal as y= (g(x))= ( -1(x))=x.
(u)g(x)
DPD
x u y
x
u
u x
yy
Figu e 3.21 Digi al p e-dis o ion idea
The concep behind he DPD echnique is based on ex ac ing he
beha io al model o he nonlinea ansmi e , es ima ing i s in e se beha io
model, in o de o p e-dis o he digi al baseband signal (p e iously o be
ansmi ed) o compensa e he ex ac ed e ec s in he powe ampli ie .
Since he powe ampli ie is he main sou ce o nonlinea i y, because i s highes
e iciency s a e is ope a ing close o he maximum ou pu powe , a scheme as
showed in Figu e 3.22 will be necessa y. This a chi ec u e implemen s a
ansmi e o dual-band applica ions. I is possible o obse e how he baseband
signals x1 and x2 a e p e-dis o ed o ob ain xpd1 and xpd2 by he digi al block ha
implemen s he in e se unc ion o he powe ampli ie . These signals a e
con e ed o analog domain and up-con e ed o RF and combined in o de o
eed he powe ampli ie and be ansmi ed. This RF signal will be coupled o a
eedback loop ha is composed by a subsampling ecei e , i.e., a BP il e , an
S&H and an ADC.
The subsampling ecei e educes he powe consump ion and he
complexi y in compa ison wi h he ypical eedback loops o linea iza ion and i
is employed o ex ac he IF signal which be digi ally down-con e ed in o de o
ob ain he baseband nonlinea signals y1 and y2. These baseband signals, besides
xpd1 and xpd2, will eed he digi al analyze block in o de o ob ain he necessa y
coe icien s ha will be employed as inpu s by he p e-dis o ed blocks, besides
he inpu signals x1 and x2, in o de o adjus he p e-dis o e s dynamically.
Chap e 3: Subsampling ecei e s 83
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
P e-dis o e
P e-dis o e
DAC
DAC
RF up-con e e
RF up-con e e
S&H
ADC
Digi al
down-con e sion
Analyzing
S age
Dual-band
powe ampli ie
Couple
x1
x2
xpd1
xpd2
y1
y2
Figu e 3.22 Dual-band digi al p edis o ion wi h subsampled eedback loop
3.6 Re e ences
[3.1] J. R. G. Oya, A. Kwan, F. Muñoz, F. M. Ghannouchi, M. Healoui, F.
Má quez, E. López-Mo illo, A. To alba, “Subsampling Recei e s
wi h Applica ions o So wa e De ined Radio,” Da a Acquisi ion,
InTech, Chap e 7, pp. 165-194, 2012.
[3.2] C. L. Phillips, E. Riskin, “Signals, Sys ems and T ans o ms,” 4 h
Edi ion, P en ice Hall, Uppe Saddle Ri e , NJ, 2008.
[3.3] R. Vaughan, N. Sco , D. Whi e, “The Theo y o Bandpass Sampling,”
IEEE T ansac ions on Signal P ocessing, ol. 39, no. 9, pp. 1973-
1984, Sep. 1991.
[3.4] R. G. Lyons, “Unde s anding Digi al Signal P ocessing,” P en ice
Hall, Uppe Saddle Ri e , NJ, 2001.
[3.5] C. Aze edo-Leme, “Clock Ji e E ec s on Sampling: A Tu o ial,”
IEEE Ci cui s and Sys ems, ol.11, no. 3, pp. 26-37, 2011.
[3.6] B. B annon, A. Ba low, “Ape u e unce ain y and ADC sys em
pe o mance,” Analog De ices, Inc., Applica ion No e AN-501, 2006.
[3.7] Y. R. Sun, “Gene alized Bandpass Sampling Recei e s o So wa e
de ined Radio,” Doc o al Disse a ion, School o In o ma ion and
Communica ion Technology (ICT), S ockholm, Sweden, 2006.
[3.8] S. Ka onen, “Cha ge-Domain Sampling o High F equency Signals
wi h Embedded Fil e ing,” Doc o al Disse a ion, Facul y o
Technology, Depa men o Elec ical and In o ma ion Enginee ing,
Uni e si y o Oulu, Finland, Jan. 2006.
84 Chap e 3: Subsampling ecei e s
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
[3.9] J. R. G. Oya, A. Ju ado, F. Muñoz, A. To alba, “High F equency
Analog- o-Digi al Con e sion Based on Subsampling”, XXIV
Con e ence o Design o Ci cui s and In eg a ed Sys ems (DCIS’2009),
Za agoza, Spain, No . 2009.
[3.10] V. A kes eijn, E. Klumpe ink, B. Nau a, “Ji e equi emen s o he
sampling clock in so wa e adio ecei e s,” IEEE T ans. Ci cui s
Sys . II, ol. 53, no. 2, pp. 90–94, Feb. 2006.
[3.11] W. Kes e , “Con e ing oscilla o phase noise o ime ji e ,” Analog
De ices, Inc., Tu o ial MT-008, 2009.
[3.12] T. Lee, A. Hajimi i, “Oscilla o phase noise: A u o ial,” IEEE J.
Solid-S a e Ci cui s, ol. 35, no. 3, pp. 326–335, Ma . 2000.
[3.13] D. Lee, “Analysis o ji e in phase-locked loops,” IEEE T ans.
Ci cui s Sys . II, ol. 49, no. 11, pp. 704–711, May 2002.
[3.14] M. R. Yuce, W. Liu, “Al e na i e Wideband F on -End A chi ec u es
o Mul i-S anda d So wa e Radios,” IEEE 60 h Vehicula
Technology Con e ence (VTC’2004), ol. 3, pp. 1968-1972, Fall 2004.
[3.15] J. R. G. Oya, F. Muñoz, A. To alba, A. Ju ado, A. J. Ga ido, J.
Baños, “Da a Acquisi ion Sys em Based on Subsampling o Tes ing
Wideband Mul is anda d Recei e s," IEEE T ansac ions on
Ins umen a ion and Measu emen s, ol. 60, no. 9, pp. 3234-3237, Sep.
2011.
[3.16] H. Pekau, J. W. Hasle , “A compa ison o analog on end
a chi ec u es o digi al ecei e s,” Canadian Con e ence on
Elec ical and Compu e Enginee ing (CCECE’2005), pp. 1073-1077,
2005.
[3.17] M. B. Dadi, R. Bouallegue, “On he RF Subsampling Con inuous-
Time ΣΔ Downcon e sion S age o Mul is anda d Recei e s,”
In e na ional Con e ence on Compu e Enginee ing and Technology
(ICCET), ol. 6, pp. 167-171, June 2010.
[3.18] A. I. Hussein, W. B. Kuhn, “Bandpass Σ∆ modula o employing
unde sampling o RF signals o wi eless communica ion,” IEEE
T ansac ions on Ci cui s and Sys ems II: Analog and Digi al Signal
P ocessing, ol. 47, no. 7, pp. 614–620, July 2000.
[3.19] M. B. Dadi, R. Bouallegue, “Subsampling Con inuous-Time Bandpass
ΣΔ Modula o o Radio F equency A/D Con e sion,” 10 h
In e na ional Con e ence on In o ma ion Sciences Signal P ocessing
and hei Applica ions (ISSPA’2010), pp. 181-184, 2010.
Chap e 3: Subsampling ecei e s 85
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
[3.20] S-W. M. Chen, R. W. B ode sen, “A Subsampling Radio A chi ec u e
o Ul awideband Communica ions,” IEEE T ansac ions on Signal
P ocessing, ol. 55, no. 10, pp. 5018-5031, Oc . 2007.
[3.21] Y. Vande pe en, W. Dehaene, G. Leus, “A Flexible Low Powe
Subsampling UWB Recei e Based on Line Spec um Es ima ion
Me hods,” IEEE In e na ional Con e ence o Communica ions
(ICC’2006), ol. 10, pp. 4694-4699, 2006.
[3.22] S. Roy, J. R. Foe s e , V. S. Somayazulu, D. G. Leepe ,
“Ul awideband adio design: The p omise o high-speed, sho - ange
wi eless connec i i y,” P oceedings o he IEEE, ol. 4, no. 2, pp.
295–311, Feb. 2004.
[3.23] G. R. Aiello, G. D. Roge son, “Ul a-wideband wi eless sys ems,”
IEEE Mic owa e Mag., ol. 4, no. 2, pp. 36–47, June 2003.
[3.24] J. Zhang, T. Abhayapala, R. Kennedy, “P incipal Componen s
T acking Algo i hms o Synch oniza ion and Channel Iden i ica ion
in UWB Sys ems,” IEEE Eigh h In e na ional Symposium on Sp ead
Spec um Techniques and Applica ions, pp. 369-373, Sep . 2004.
[3.25] A. Kwan, S. A. Bassam, F. M. Ghannouchi, “Sub-sampling Technique
o Spec um Sensing in Cogni i e Radio,” IEEE Radio and Wi eless
Symposium (RWS’2012), pp. 347-350, 2012.
[3.26] S. A. Bassam, A. Kwan, W. Chen, M. Helaoui, F. Ghannouchi,
“Subsampling Feedback Loop Applicable o Concu en Dual-Band
Linea iza ion A chi ec u e,” IEEE T ansac ions on Mic owa e Theo y
and Techniques, ol. 60, no.6, pa 2, pp. 1990-1999, 2012.
[3.27] F. M. Ghannouchi, O. Hammi, “Beha io al modeling and
p edis o ion,” IEEE Mic owa e Magazine, ol. 10, no. 7, pp. 52–64,
Dec. 2009.
[3.28] S. A. Bassam, M. Helaoui, F. M. Ghannouchi, “C osso e Digi al
P edis o e o he Compensa ion o C oss alk and Nonlinea i y in
MIMO T ansmi e s,” IEEE T ansac ions on Mic owa e Theo y and
Techniques, ol. 57, no. 5, pp. 1119-1128, May 2009.

86 Chap e 3: Subsampling ecei e s
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
CHAPTER
4
DATA ACQUISITION
SYSTEMS BASED ON
SUBSAMPLING FOR TESTING
WIDEBAND MULTI-
STANDARD RECEIVERS
CHAPTER CONTENTS
4.1 Da a acquisi ion sys ems based on COTS .............................................................. 89
4.1.1 Choice o componen s ....................................................................................... 89
4.1.2 Expe imen al esul s .......................................................................................... 90
4.1.2.1 S&H cha ac e iza ion ............................................................................ 90
4.1.2.2 Subsampling ecei e cha ac e iza ion ................................................. 92
4.2 Da a acquisi ion sys ems based on PCB ............................................................... 101
4.2.1 PCB design ...................................................................................................... 101
4.2.2 Expe imen al esul s ........................................................................................ 103
4.3 Noise pe o mance op imiza ion based on mul iple clocking echniques ............ 104
4.3.1 Theo e ical s udy ............................................................................................. 105
4.3.2 Expe imen al esul s ........................................................................................ 106
4.3.2.1 COTS le el .......................................................................................... 107
4.3.2.2 PCB le el ............................................................................................ 110
88 Chap e 4: Da a acquisi ion sys ems based on subsampling o es ing wideband mul i-
s anda d ecei e s
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
4.4 Compa ison wi h o he implemen ed mul i-s anda ds ecei e s ........................... 112
4.5 Re e ences ............................................................................................................. 116
As desc ibed in Chap e 2, i a mul i-s anda d ecei e is implemen ed by
s acking di e en ecei e s o di e en s anda ds in o a single ecei e , he a ea
and powe consump ion will be ex emely high. The e o e, a p ope ly designed
mul i-s anda d ecei e mus sha e he ha dwa e esou ces and use unable and
p og ammable de ices, educing he a ea and powe consump ion, which is a
e y impo an app oach o ba e y powe de ices. O he wise, o mul i-s anda d
applica ions such as ins umen a ion o alida ion, he main cons ain is he
capabili y o co e ing he maximum numbe o s anda ds as possible.
A da a acquisi ion boa d based on subsampling o high pe o mance low-
cos mul i-s anda d es equipmen is p esen ed in his chap e . Due o he
necessi y o lexible and low cos ecei e s in he es indus y, he selec ed
a chi ec u e is based on subsampling echniques. P e iously, a s a e o he a
s udy and an expe imen al e alua ion a COTS le el we e pe o med in o de o
alida e he design and de e mine he eal speci ica ions o he da a acquisi ion
sys em. Wi h a signal bandwid h o 20 MHz i achie es 8.5 bi esolu ion o a
p og ammable ca ie equency anging om 0 up o 3.3 GHz, and mo e han 8
bi esolu ion up o 4 GHz. By a p ope selec ion o he cen e equency and
signal bandwid h, he p oposed boa d can be used o digi ize he signal in mos o
p esen wi eless s anda ds. This design is in ended o be pa o a es sys em; ha
is, he inpu signal o he subsampling ecei e is assumed o be il e ed and ee
o in e e ences.
A las sec ion desc ibes he achie ed imp o emen o he noise
pe o mance by using wo clocking s ages a chi ec u e. This app oach allows he
sampling equency o he i s s age o inc ease, esul ing in a lowe con ibu ion
o he i s S&H o he o al olded he mal noise. This sec ion is s uc u ed in a
i s pa dedica ed o he heo e ical s udy, whe e he exp essions o he expec ed
imp o emen a e deduced, and a second pa dedica ed o he expe imen al
alida ion o hese echniques. Conside ing a signal bandwid h o 20 MHz, he
imp o ed da a acquisi ion sys em achie es an ENOB o mo e han 9 bi s o a
p og ammable ca ie equency up o 2.9 GHz and 8 bi s up o 6.5 GHz,
p esen ing an imp o emen in he esolu ion o 0.5-1 bi . The chap e inalizes
p esen ing a compa ison wi h o he implemen ed mul i-s anda d ecei e s.
Chap e 4: Da a acquisi ion sys ems based on subsampling o es ing wideband mul i-s anda d
ecei e s 89
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
4.1 Da a acquisi ion sys ems based on COTS
4.1.1 Choice o componen s
Since he da a acquisi ion sys em is based on comme cial de ices, hey
will ha e o be chosen in o de o minimize he e ec s o he main p oblems
encoun e ed using subsampling, i.e., ji e and olded he mal noise, as de ailed in
Chap e 2. These non ideali ies ix he speci ica ions o he main building blocks
o he sys em, i.e., he S&H and he ADC.
A e a s udy on comme cial componen s, we decided o use an ex e nal
S&H be o e he ADC since an in e nal S&H bandwid h is limi ed o 3 GHz
app oxima ely wi h a esolu ion a ound 8 bi s. Howe e , when using an ex e nal
S&H, i is possible o ob ain a highe esolu ion o a wide bandwid h. This is a
easible al e na i e because he S&H can subsample he RF inpu s co e ing a high
analog bandwid h and he ob ained IF eplica can be con e ed o digi al by a high
esolu ion, in e media e equency ADC.
A e ealizing a s udy o he s a e o he a , he chosen S&H is he Inphi
1821 TH [4.1], wi h he ollowing ea u es:
Wide bandwid h (18 GHz), in o de o co e mos o wi eless
communica ion s anda ds.
Minimum ape u e ji e (50 s).
In eg a ed noise o e he i s Nyquis band (clocking a 1 GHz) equal o
0.64 mV. This alue is e y simila o he o he s udied S&H om Hi i e
[4.2] o Teledyne [4.3].
Wide equency ange (0-6 GHz) o 10-bi linea i y.
Al hough he maximum sampling equency is equal o 2 GS/s, his S&H
has he capabili y o sample a he in e es ed equency (a ound 500 MS/s)
o his applica ion.
This equi emen (500 MS/s) is gi en by he maximum sampling equency
o comme cial 10-bi A/D con e e s, since using a maximum sampling
equency is con enien in o de o educe he o e lapping noise e ec s.
Conc e ely, he A/D con e e chosen is E2V AT84AS001 [4.4].
The es o componen s will be desc ibed as hey a e u ilized in he
expe imen al cha ac e iza ion ha is de ailed in he ollowing sec ion.
96 Chap e 4: Da a acquisi ion sys ems based on subsampling o es ing wideband mul i-
s anda d ecei e s
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
Figu e 4.7 Measu ed SFDR s. inpu signal ampli ude
4.1.2.2.2.3 3 d o de in e modula ion dis o ion (IMD3) measu emen
Fo hese expe imen s a single one modula ed in ampli ude (AM
modula ion) wi hou ca ie supp ession is employed as inpu and, he e o e, he
signal inpu is composed by h ee ones equally spaced 1 MHz.
Figu e 4.8 illus a es he IM measu ed o he same inpu equencies (1, 2
and 3 GHz) clocking a he op imal sampling equency. In his case he inpu
ampli ude mus be educed in o de o op imize he pe o mance, dis ibu ing he
o al ene gy in he h ee di e en ones. Al hough he cha ac e iza ion o SNDR
o he all equency ange has been implemen ed o a single one case, as
desc ibed in he ollowing sec ions, i is possible o obse e how he e is an
ampli ude ange (a ound -3 dBm) whe e he ob ained IMD3 is no a p oblem and
i s SNDR ha e accep able alues, i.e., e y close o he op imal pe o mance.
The e o e, using se e al ones as inpu in his ampli ude ange would lead o e y
simila SNDR esul s as desc ibed in he ollowing sec ions.
-3 -2 -1 012345
50
52
54
56
58
60
62
64
66
68
Ain (dBm)
SFDR (dB)
Fin=1001 MHz y Fs=445,3 MHz
Fin=2001 MHz y Fs=471,2 MHz
Fin=3001 MHz y Fs=480,4 MHz

Chap e 4: Da a acquisi ion sys ems based on subsampling o es ing wideband mul i-s anda d
ecei e s 97
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
Figu e 4.8 Measu ed IM s. inpu ampli ude
4.1.2.2.2.4 Subsampling p ope ies in he implemen ed sys em
As desc ibed abou he S&H cha ac e iza ion, his implemen ed ecei e a
COTS le el can be used o illus a e some subsampling e ec s s udied in he
heo e ical sec ions. Fi s ly, he dependency o he o e lapping he mal noise
wi hin band signal on he subsampling equency was expe imen ally
cha ac e ized, wi h i s e ec illus a ed in Figu e 4.9 [4.21]. This igu e ep esen s
ENOB ob ained o di e en op imal subsampling equencies o an inpu signal
a 2001 MHz wi h 20 MHz o inpu bandwid h. As expec ed he o al esolu ion is
dec eased when lowe sampling equencies a e used.
The same e ec can be app ecia ed om Figu e 4.10 and Figu e 4.11,
whe e i is possible o obse e how he noise loo inc eases in he case o using
he lowes sampling equency (Figu e 4.11).
-9 -8 -7 -6 -5 -4 -3 -2 -1 0 1
20
25
30
35
40
45
50
55
60
65
70
Ain (dBm)
IM (dB)
Fc=1001 MHz y Fs=444,9 MHz
Fc=2001 MHz y Fs=470,8 MHz
Fc=3001 MHz y Fs=480,2 MHz
98 Chap e 4: Da a acquisi ion sys ems based on subsampling o es ing wideband mul i-
s anda d ecei e s
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
Figu e 4.9 Measu emen o he o e lapping he mal noise: ENOB ob ained o
di e en op imal sampling equencies
Figu e 4.10 Ou pu spec um o a 2001 MHz inpu signal subsampled a 470.8
MHz
200 250 300 350 400 450 500
0
1
2
3
4
5
6
7
8
9
s (MHz)
enob
10
3
10
4
10
5
10
6
10
7
10
8
10
9
0
20
40
60
80
100
120
Hz
dBm
Chap e 4: Da a acquisi ion sys ems based on subsampling o es ing wideband mul i-s anda d
ecei e s 99
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
Figu e 4.11 Ou pu spec um o a 2001 MHz inpu signal subsampled a 216.3
MHz
Ano he in e es ing measu emen is on he in luence o he ji e , which is
mo e c i ical a highe inpu equencies. This e ec is illus a ed in Figu e 4.12
[4.21], showing he ENOB ob ained o di e en inpu equencies using he
op imal subsampling equency immedia ely less han 500 MHz and an inpu
bandwid h o 20 MHz.
Figu e 4.12 Measu emen o he ji e noise: ENOB ob ained o di e en inpu
equencies
1000 1500 2000 2500 3000 3500
8.2
8.3
8.4
8.5
8.6
8.7
8.8
8.9
c (MHz)
enob
10
3
10
4
10
5
10
6
10
7
10
8
10
9
0
20
40
60
80
100
120
Hz
dBm
100 Chap e 4: Da a acquisi ion sys ems based on subsampling o es ing wideband mul i-
s anda d ecei e s
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
4.1.2.2.2.5 Measu ed esolu ion o all he inpu ange
Finally, he o al esolu ion o he p oposed sys em is illus a ed in Figu e
4.13 [4.21]. These esul s a e ob ained o an inpu equency up o 3.3 GHz
in eg a ing in an inpu bandwid h o 20 MHz and clocking a a equency close o
he op imal sampling equency immedia ely lowe han 500 MHz. The e o e,
hese inal esul s show a da a acquisi ion sys em a COTS le el ha con e s o
digi al signals wi h he ollowing esolu ion:
A ound 9 bi s up o 2 GHz inpu equency.
Mo e han 8 bi s up o 3.1 GHz inpu equency.
Figu e 4.13 ENOB s. inpu equency
The sys em pe o mance ob ained om measu emen s is summa ized in
Table 4.1.
As an example, Figu e 4.14 [4.21] illus a es he ou pu spec um o a 3
GHz inpu equency and a 480.2 MHz sampling equency.
500 1000 1500 2000 2500 3000 3500
0
1
2
3
4
5
6
7
8
9
10
Fin (MHz)
enob
Chap e 4: Da a acquisi ion sys ems based on subsampling o es ing wideband mul i-s anda d
ecei e s 101
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
Figu e 4.14 Ou pu spec um o a 3 GHz inpu equency
Table 4.1 Sys em pe o mance a COTS le el
Maximum Inpu F equency
3.1 GHz
Signal Bandwid h
20 MHz
Sampling F equency
<500 MHz
ENOB (SNDR)
>8.1 bi s
SFDR
>61.8 dBc
Vol age Supply S&H
-5,2 V
Vol age Supplies ADC
5 V (Analog), 3.3 V (Digi al, Ou pu )
Powe Consump ion
3.7 W
4.2 Da a acquisi ion sys ems based on PCB
4.2.1 PCB design
A da a acquisi ion module a PCB le el o high pe o mance low-cos
mul i-s anda d es equipmen was p esen ed in [4.22]. This wo k p o ides high
esolu ion o e a la ge bandwid h wi h only a low-ji e wideband S&H and an
in e media e equency ADC, by means o subsampling. Using comme cial
de ices on a mul ilaye p in ed ci cui boa d, expe imen al esul s showed mo e
103104105106107108109
-20
0
20
40
60
80
100
120
Hz
dBm
BW=20MHz

102 Chap e 4: Da a acquisi ion sys ems based on subsampling o es ing wideband mul i-
s anda d ecei e s
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
han 8 bi s esolu ion o a 20 MHz signal bandwid h wi h up o 3.8 GHz cen e
equency, enough o co e he equi emen s o es sys ems o mos o p esen
wi eless communica ion s anda ds. These comme cial de ices a e he same ha
used a COTS le el, i.e., he S&H Inphi 1821TH and he ADC E2V AT84AS001.
These de ices a e he main componen s o he p oposed da a acquisi ion
sys em o which a mul i laye PCB p o o ype was designed and ab ica ed
(Figu e 4.15 [4.22]). Mo eo e , his p o o ype includes o he componen s, such as
baluns [4.23], bias ees [4.24] and LP passi e il e s (Minici cui s LFCN-160
[4.25]). Thanks o a highe a ailabili y o su ace moun de ices (SMD), he
chosen il e s le o implemen a be e adjus men o he il e ing s age because i s
cu o equency is lowe and, he e o e, he 2nd o de ha monics, which a e in he
200-250 MHz9, a e mo e a enua ed.
This design uses a Class 7 boa d wi h DE104i FR4 dielec ic, six me al
laye s and mic os ip lines adap ed o 50 . The ea u es o his class and his
dielec ic a e ob ained om [4.26] and a e de ailed in Appendix B. On he o he
hand, he ules and exp essions employed o adap he componen s (designing he
dimensions o aces and laye s) we e ob ained om [4.27] and also a e desc ibed
in Appendix B.
S&H
IN+
IN-
CLK+CLK-
ADC
IN+
IN-
OUT+
OUT-
OUT
D11-D012
Fil e
LP
Fil e
LP
Bias
ee
Bias
ee
Balun
CLK+CLK-
Balun
Figu e 4.15 Block diag am and designed PCB p o o ype
Employing he ex e nal me al laye s (1 and 6) o signals, he adjacen
me al laye s o g ound (2 and 5) and he mos in e nal laye s (3 and 4) o powe
supplies, he esul an s ack-up is as showed in Figu e 4.16, which ollows a
9 Since he chosen op imal sampling equency is in he ange o 400-500 MHz in o de o old he desi ed IF eplica o
s/4, he 2nd o de ha monic will be loca ed in he ange o 200-250 MHz.
Chap e 4: Da a acquisi ion sys ems based on subsampling o es ing wideband mul i-s anda d
ecei e s 103
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
ypical s uc u e p oposed by he manu ac u e Labci cui s [4.26] and has a
hickness equal o 1.22 mm.
Coppe (Signal)
Coppe (Signal)
Coppe (G ound)
Coppe (G ound)
Coppe (Powe )
Coppe (Powe )
P ep eg
P ep eg
P ep eg
Co e
Co e
0.669 mils
0.669 mils
0.669 mils
0.669 mils
0.669 mils
0.669 mils
3.397 mils
3.397 mils
6.693 mils
6.693 mils
6.693 mils
Figu e 4.16 Implemen ed s ack-up
The ci cui was ca e ully laid ou in o de o minimize he ji e e ec . The
dis ance be ween signal acks, pads and me al laye s was ca e ully chosen in
o de o educe c oss alk and in e -symbol in e e ence, which cause ji e . O he
ules ollowed o educe ji e we e a co ec decoupling om he powe lines and
he signal planes employing he echnique called picke ences [4.28], consis ing in
placing closely spaced VIAs ( e ical in e connec access) be ween di e en
g ound planes. A ypical dis ance be ween VIAs equal o 1/20 wa eleng h (λ) was
selec ed. Finally, he dimensions o he designed PCB a e 16.05 x 10.64 cm2.
4.2.2 Expe imen al esul s
The da a acquisi ion sys em was expe imen ally cha ac e ized, wi h he
ob ained pe o mance summa ized in Table 4.2 [4.22]:
Table 4.2 Sys em pe o mance a PCB le el
Signal Bandwid h
20 MHz
Signal Inpu F equency
3.3 GHz
Sampling F equency
<500 MHz
ENOB (SNDR)
>8.2 bi s
Linea i y (SFDR)
>9.49 bi s
IMD3
>60.4 dB
Powe Consump ion
3.7 W
104 Chap e 4: Da a acquisi ion sys ems based on subsampling o es ing wideband mul i-
s anda d ecei e s
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
The measu ed SFDR and IMD3 we e abou 60 dB ega dless o he inpu
equency, and hus he ENOB (based on he maximum SNDR) is la ge han 8.2
bi s o sinusoidal inpu signals up o 3.3 GHz.
Figu e 4.17 [4.22] shows he sys em esolu ion as a unc ion o he ca ie
equency up o 20 GHz, o a sampling equency e y close o 500 MHz. The
esul s p o ide a use ul cha ac e iza ion o he sys em esponse and clea ly show
he e ec o ji e . Up o 3.3 GHz he e is no signi ican in luence o ji e , as he
ENOB is nea ly cons an in his ange. The esolu ion alls as he inpu equency
inc eases, mos ly due o he in luence o ji e . The sys em esponse p o ides an
ENOB la ge han 7 bi s up o 6 GHz, 6 bi s up o 13 GHz and 5.5 bi s a 20 GHz.
Figu e 4.17 ENOB s. inpu equency (20 MHz signal band, up o 20 GHz inpu
ca ie equency)
The e o e, his wo k p esen s a da a acquisi ion module o es ing wi eless
ecei e s based on subsampling, which co e s mos p esen wi eless
communica ion s anda ds equi emen s, wi h an ENOB be ween highe han 8 bi s
up o 4 GHz cen e equency o a 20 MHz signal bandwid h.
4.3 Noise pe o mance op imiza ion based on mul iple
clocking echniques
As desc ibed in Sec ion 4.2, a limi a ion o he da a acquisi ion sys ems
based on subsampling is he maximum sampling equency, which will be gi en
by he ADC speci ica ion and is a ound 400-500 MHz o comme cial ADCs wi h
la ge enough esolu ion.
In o de o educe he olded noise e ec , his sec ion desc ibes a me hod
o imp o e he esolu ion, which employs wo consecu i e subsampling s ages
1091010
4
5
6
7
8
9
in (Hz)
ENOB
Chap e 4: Da a acquisi ion sys ems based on subsampling o es ing wideband mul i-s anda d
ecei e s 105
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
[4.29]. The use o wo subsampling p ocesses allows he sampling equency o
he i s s age o be inc eased, esul ing in a lowe con ibu ion o he i s S&H o
he olded he mal noise, such as desc ibed in Sec ion 3.4.1.
4.3.1 Theo e ical s udy
Figu e 4.18 shows wo di e en al e na i es o implemen a subsampling
based ecei e . Figu e 4.18a illus a es he scheme o a unique subsampling
p ocess implemen ed in Sec ion 4.2 while Figu e 4.18b illus a es he scheme wi h
wo di e en clocks. The sampling equency o he i s S&H in Figu e 4.18b is
selec ed be ween 1.2 GHz and 2 GHz (gi en by he maximum sampling equency
clock speci ica ions o he S&H), ob aining a band-limi ed signal a he ou pu . As
he i s sampling equency is e y la ge he olded he mal noise added by his
s age is educed. A e il e ing, he esul ing signal is subsampled again by a
second S&H a 400-500 MHz. Howe e , some d awbacks o he p oposed sys em
a e a highe complexi y and powe consump ion han he one s age subsampling
ecei e .
=1-20 GHz
Fil e
LP
= 400-500 MHz
=1-20 GHz
Fil e
BP
clock
= 1,2-2 GHz
(b)
(a)
c
c
s
1s
S&H1
clock
S&H1
clock
S&H2
ADC
clock
S&H2
ADC
= 400-500 MHz
2s
Figu e 4.18 Clocking schemes o : (a) a unique clock and (b) wo di e en clocks
The S&H in Figu e 4.18 can be modeled as shown in Figu e 3.14a, whe e
he swi ch in oduces he mal noise o powe spec al densi y Sin( )=4kTRon ha
will be il e ed by he RC ci cui , esul ing in an ou pu noise powe o Pn,ou =kT/C,
as was ob ained in Sec ion 3.4.2.
In his sec ion, he ou pu noise was conside ed o be a Gaussian he mal
noise il e ed by a b ick-wall il e o bandwid h equal o Be (noise bandwid h),
as shown in Figu e 3.14c:
dB
on
e
CR
B3
24
1
(4.1)
Whe e 3dB is he 3-dB bandwid h o he RC il e .
On he o he hand, he SNR in [-Be ,Be ] is de ined as [19]:
oi
sNmN
P
SNR )1(
(4.2)
112 Chap e 4: Da a acquisi ion sys ems based on subsampling o es ing wideband mul i-
s anda d ecei e s
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
4.4 Compa ison wi h o he implemen ed mul i-
s anda ds ecei e s
The e a e wo main classi ica ions o mul i-s anda ds ecei e s. In sec ion
2.2.6 a compa a i e o mul i-s anda d ecei e s abou digi aliza ion echniques
(i.e., mixing o subsampling based sys ems) was in oduced. Al hough
subsampling echniques ha e some p oblems as he olded he mal noise e ec o
he aliasing in mul i-band scena ios (as de ailed in Chap e 4), o SDR
applica ions a e e y con enien in o de o place, using a ew building blocks, he
analog- o-digi al con e sion s age as close he an enna as possible.
Mo eo e , mul i-s anda ds ecei e s migh be classi ied abou hei band
s a egy, i.e., i hey implemen a na ow-band s a egy o a wide-band s a egy. A
na ow-band s a egy is implemen ed by he ecei e s ha a e designed o some
speci ics s anda ds, employing dedica ed channels, while a wide-band s a egy is
implemen ed by he ones ha co e a highe numbe o wi eless communica ion
s anda ds. The e o e, na ow-band ecei e s migh p o ide a ine op imiza ion
o speci ic s anda ds while wide-band ecei e s migh be conside ed uni e sal
ecei e s and a e used o mo e gene al applica ions. Al hough hese wideband
solu ions a e mo e lexible, hei main incon enience is he RF on -end mus
mee he equi emen s o each s anda d and hey a e no op imum o any
s anda d.
Since i has been desc ibed in his chap e , he wo ks p esen ed in
[4.22,4.29] migh be an app oach o he idea o uni e sal ecei e o SDR
applica ions. O he wo ks ha e been published, which can be conside ed mul i-
s anda d ecei e s. Howe e , some o hese wo ks a e based on mixing echniques,
losing pa o he lexibili y and simplici y p o ided by he subsampling based
sys ems. On he o he hand, he e a e also mul i-s anda d ecei e s, which
al hough hey a e based on subsampling, a e op imized o a gi en numbe o
wi eless s anda ds, wi hou co e ing all he applica ions.
Examples in bo h di ec ions a e p esen in he li e a u e. [4.36,4.37] a e
examples o wo ks which employ wideband s a egies. [4.37] p esen s a ecei e
on -end o mul i-s anda d wi eless applica ions, i s analog bandwid h being up
o 3.5 GHz. [4.36] p esen s a wideband-mul i-s anda d sys em ha can be
conside ed as an uni e sal ecei e , co e ing inpu equencies be ween 0.8 and 6
GHz and being based in mixing echniques. Al hough [4.36] is a mo e complex
solu ion han [4.29], his wo k has a la ge uning ange and o he bene i s as a
high linea i y and low powe consump ion, because o i s implemen a ion in IC
(In eg a ed Ci cui ). As a main incon enience, since [4.36,4.37,4.22,4.29] a e
wideband solu ions and, he e o e, hey a e no op imum o any s anda d.

Chap e 4: Da a acquisi ion sys ems based on subsampling o es ing wideband mul i-s anda d
ecei e s 113
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
Abou he na ow-band s a egy, [4.34] p oposes an al e na i e mul i-
s anda d ecei e solu ion sepa a ing in o wo di e en RF channels, one o he
2.4 GHz and 5 GHz WLANs and he o he o he GSMs. The di e en channels
sha e a common p og ammable baseband, his solu ion being highly e icien ,
because e e y pa h is op imized o a speci ic s anda d. On he hand, he main
d awback o his wo k is, besides he limi a ion o he numbe o s anda ds, he
a ea consump ion, due o he high selec i i y is achie ed by means o many
induc o s.
An example o a solu ion based on na ow-band s a egies is showed in
Figu e 4.23a [4.38]. In his ecei e , dedica ed Blue oo h and GPS links allow
connec i i y while making a phone call o /and sending o ecei ing da a h ough a
WLAN. The WLAN pa h connec s o IEEE802.11a/b/g/n ou e s, while he
cellula -dedica ed channel can swi ch om one o he GSM bands o he
UMTS/WCDMA. Mo eo e , he selec ion is p o ided by o chip SAW il e s,
which elax he linea i y equi emen s.
O he wo ks p o ide a high le el o ha dwa e sha ing, as [4.39], whe e he
di e en speci ic s anda ds employ a common acquisi ion and digi aliza ion s ages.
This wo k p oposes a solu ion o Blue oo h, GSM, UMTS and WLAN, whe e he
las h ee s anda ds sha e he same ci cui y a e he il e bank (Figu e 4.23b),
allowing euse some building blocks in he ecei e a chi ec u e.
This ha dwa e sha ing maximiza ion means a minimum a ea consump ion,
making i possible hank o all he conside ed s anda ds, excep Blue oo h, do no
need o be co e ed a he same ime, i.e., when an applica ion is ac i e, he o he s
can be swi ched o in o de o sa e powe .
On he o he hand, da a acquisi ion sys ems o di e en communica ions
s anda ds use subsampling echniques in o de o p ocess high equency signals
wi h only a ew componen s. [4.40] p oposes a subsampling ecei e o h ee
di e en s anda ds (GSM, UMTS and IEEE 802.11g), which alida es hese
opologies a a simula ion le el in o de o be applied o mul i-s anda d adio
design. An addi ional goal o his wo k is he design o he RF and IF il e s o
he di e en s anda ds, in o de o a oid he aliasing caused by he subsampling
p ocess.
In o he published wo ks [4.41,4.42], he ecei e s based on subsampling
a e implemen ed expe imen ally only o ixed bands. [4.41] p oposes a low noise
subsampling implemen a ion o he 2.1 GHz band, and [4.42] o 2.4 GHz (IEEE
802.11a/g WLAN s anda ds). In [4.41] an IC ecei e designed in 0.18 µm CMOS,
whose main goal is a unable LC il e implemen a ion, is p oposed. [4.42] shows
a 0.18 µm CMOS ecei e which ep esen s he mos comple e subsampling
ecei e e e ence, hanks o he op imiza ion pe o med o pa ame e s as he mal
noise le el, ji e -induced noise and nonlinea i y. Finally, he e a e also ecei e s
114 Chap e 4: Da a acquisi ion sys ems based on subsampling o es ing wideband mul i-
s anda d ecei e s
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
based on subsampling o UWB applica ions, like he one in [4.43], which
ope a es in he 3.1-10.6 GHz band wi h low powe consump ion.
0º90º
VGA1VGA2
VGA1VGA2
VCO
ADC
ADC
n
n
Tunable
SAW
Mul i-Band
LNA
Digi al Gain and Band Selec ion
Cellula
Radios
0º90º
VGA1VGA2
VGA1VGA2
VCO
ADC
ADC
n
n
Tunable
SAW
Mul i-Band
LNA
Digi al Gain and Band Selec ion
Wi eless
LANs
GPS SAW
GPS Recei e Pa h
Blue oo h
SAW
Blue oo h Recei e
Pa h
(a)
0º90º
VGA1VGA2
VGA1VGA2
VCO
ADC
ADC
n
n
Mul i-Band
LNA
Digi al Gain and Band Selec ion
Blue oo h
SAW
Blue oo h Recei e
Pa h
(b)
Figu e 4.23 Mul i-s anda d ecei e a chi ec u es p oposed in [4.38] (a) and
[4.39] (b)
Chap e 4: Da a acquisi ion sys ems based on subsampling o es ing wideband mul i-s anda d
ecei e s 115
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
Finally, Table 4.6 [4.29] shows he speci ica ions o mos common
wi eless communica ion s anda ds [4.34] and he esul s ob ained in some o hese
p e iously published wo ks abou he noise pe o mance. These esul s a e
compa ed wi h he ob ained in [4.22] in o de o obse e he bene i s o
implemen a mul iple clock echnique o mul i-s anda d ecei e s based on
subsampling. I can be seen ha , employing he da a acquisi ion sys em designed
in [4.29], only he ENOB speci ica ions o IEEE 802.11a a e no achie ed,
al hough hey a e e y close. Mo eo e , no e ha some speci ica ions, like Noise
Figu e (NF) o UMTS (I) and 802.11b/g, o esolu ion o Blue oo h, we e no
achie ed wi hou he imp o emen p oposed in [4.29], i.e., when a unique clock
sou ce is used [4.22]. Compa ing wi h he o he published wo k, simila esul s
abou NF and Noise PSD can be obse ed wi h espec o [4.29], showing a la ge
in luence o he ji e (i.e., educing he esolu ion wi h he inpu equency) in he
wo k p esen ed in [4.29].
Table 4.6 S anda d speci ica ions and esul s
S anda d
GSM
1800
UMTS
(I)
Blue-
oo h
802.11b/g
802.11a
S anda d
equi emen s
Ca ie F eq.
(MHz)
1805.2-
1879.8
2110-
2170
2400
2400
5000
Signal
Bandwid h
(MHz)
0.2
5
1
20
20
ENOB (bi s)
9
6
11
8
9
NF (dB)
9.3
4.6
10.7
6.5
18.2
Expe imen al
esul s o
p e iously
published
acquisi ion
sys ems
ENOB [4.22]
(bi s)
11.76
9.56
10.36
8.2
7.41
NF [4.22] (dB)
6.1
6.5
8.3
8.3
13.1
NF [4.34] (dB)
5.2
5.6
5.8
NF [4.37] (dB)
5.8
6
6.5
6.5
NF [4.40] (dB)
7.5
7.2
Noise PSD
[4.42]
(dBm/Hz)
-131
Expe imen al
esul s o [4.29]
ENOB (bi s)
12.47
9.97
10.86
8.7
8.34
NF (dB)
3.6
4.4
6.2
6.2
9.3
Noise PSD
(dBm/Hz)
-129.7
-128.8
-126.9
-126.9
-123.8
116 Chap e 4: Da a acquisi ion sys ems based on subsampling o es ing wideband mul i-
s anda d ecei e s
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
4.5 Re e ences
[4.1] 1821TH, 18 GHz Bandwid h 2 GS/s THA, Inphi Co ps. Da ashee .
[4.2] HMC660LC4B, 0.02-4.5 GHz Wideband 3 GS/s T ack-and-Hold
Ampli ie , Hi i e Mic owa e Co po a ion. Da ashee .
[4.3] RTH010-060 Se ies, 8-16 GHz Bandwid h 1-4 GS/s Dual T ack-and-
Hold, Teledyne Scien i ic Company. Da ashee .
[4.4] AT84AS001, 12-bi 500 Msps ADC, E2V Technologies. Da ashee .
[4.5] 1821TH, 18 GHz T ack and Hold, Inphi Co ps. Applica ion No e.
[4.6] On line: www.minici cui s.com
[4.7] E8257D PSG, Mic owa e Analog Signal Gene a o , Agilen
Technologies. Da ashee .
[4.8] SMIQ, Vec o Signal Gene a o , Rohde & Schwa z. Da ashee .
[4.9] 16720-A, Measu emen s Modules o he 16900 Se ies, Agilen
Technologies. Da ashee .
[4.10] ZAPDJ-2, Powe Spli e /Combine 2 Way-180º 50Ω, Minici cui s.
Da ashee .
[4.11] ZFSCJ-2-4, Powe Spli e /Combine 2 Way-180º 50Ω, Minici cui s.
Da ashee .
[4.12] ZFSCJ-1-2, Powe Spli e /Combine 2 Way-180º 50Ω, Minici cui s.
Da ashee .
[4.13] BLK-89, DC-Block 50 Ω, Minici cui s. Da ashee .
[4.14] ZFTB-4R2GW-FT, Bias-Tee 50 Ω Wideband, Minici cui s. Da ashee .
[4.15] ANNE-50, Te mina ion SMA 50 Ω, Minici cui s. Da ashee .
[4.16] SLP-250, Low Pass Fil e 50 Ω, Minici cui s. Da ashee .
[4.17] Cable, 2-FT SMA M-N M, Minici cui s. Da ashee .
[4.18] Cable, 1.5-FT SMA M-SMA M, Minici cui s. Da ashee .
[4.19] Adap e , SMA F-SMA F, Minici cui s. Da ashee .
[4.20] AT84AS001-EB, E alua ion Boa d, E2V Technologies. Use Guide.
[4.21] J. R. G. Oya, A. Ju ado, F. Muñoz, A. To alba, “High F equency
Analog- o-Digi al Con e sion Based on Subsampling”, XXIV
Chap e 4: Da a acquisi ion sys ems based on subsampling o es ing wideband mul i-s anda d
ecei e s 117
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
Con e ence o Design o Ci cui s and In eg a ed Sys ems (DCIS’2009),
Za agoza, Spain, No . 2009.
[4.22] J. R. G. Oya, F. Muñoz, A. To alba, A. Ju ado, A. J. Ga ido, J.
Baños, “Da a Acquisi ion Sys em Based on Subsampling o Tes ing
Wideband Mul is anda d Recei e s," IEEE T ansac ions on
Ins umen a ion and Measu emen s, ol. 60, no. 9, pp. 3234-3237, Sep.
2011.
[4.23] SBTCJ-1W, Powe Spli e /Combine 2 Way-180º 50Ω, Minici cui s.
Da ashee .
[4.24] TCBT-6G, Bias-Tee 50 Ω Wideband, Minici cui s. Da ashee .
[4.25] LFCN-160, Low Pass Fil e 50 Ω, Minici cui s. Da ashee .
[4.26] On line: www.labci cui s.com
[4.27] K. Mi zne , “Comple e PCB Design Using O CAD Cap u e and
Layou ,” Newnes & Else ie , Bu ling on, MA, 2007.
[4.28] D. B ooks, “Signal In eg i y and P in ed Ci cui Boa d Design,”
P en ice Hall, Uppe Saddle Ri e , NJ, 2003.
[4.29] J. R. G. Oya, F. Muñoz, A. To alba, A. Ju ado, F. Má quez, E.
López-Mo illo, “Da a Acquisi ion Sys em Base on Subsampling using
Mul iple Clocking Techniques,” IEEE Ins umen a ion and
Measu emen s, ol. 61, no. 8, pp. 2333-2335, Aug. 2012.
[4.30] R. Vaughan, N. Sco , D. Whi e, “The Theo y o Bandpass Sampling,”
IEEE T ansac ions on Signal P ocessing, ol. 39, no. 9, pp. 1973-
1984, Sep. 1991.
[4.31] J. R. G. Oya, A. Ju ado, F. Muñoz, A. To alba, F. J. Má quez, E.
López-Mo illo, “Mul iple Clocking High Analog- o-Digi al
Con e sion Based on Subsampling”, XXVI Con e ence o Design o
Ci cui s and In eg a ed Sys ems (DCIS’2011), Albu ei a, Po ugal,
No . 2011.
[4.32] SLP-550, Low Pass Fil e 50 Ω, Minici cui s. Da ashee .
[4.33] SHP-300, High Pass Fil e 50 Ω, Minici cui s. Da ashee .
[4.34] F. Agnelli e al., “Wi eless Mul i-S anda d Te minals: Sys em
Analysis and Design o a Recon igu able RF F on -End,” IEEE
Ci cui s and Sys ems Magazine, ol. 6, no.1, pp. 38-59, Jan. 2006.
[4.35] RBP-400, Band Pass Fil e 50 Ω, Minici cui s. Da ashee .

118 Chap e 4: Da a acquisi ion sys ems based on subsampling o es ing wideband mul i-
s anda d ecei e s
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
[4.36] R. Baghe i e al., “An 800 MHz-6 GHz So wa e-De ined Wi eless
Recei e in 90 nm CMOS,” IEEE Jou nal o Solid-S a e Ci cui s, j.41,
no.12, pp. 2860-2876, Dec. 2006.
[4.37] M. Vidojko ic, M. A. T. Sanduleanu, V. Vidojko ic, J. an de Tang,
P. Bal us, A. H. M. an Roe mund, “A 1.2V Recei e F on -End o
Mul i-S anda d Wi eless applica ions in 65nm CMOS LP”, 34 h
Eu opean Solid-S a e Ci cui Con e ence (ESSCIRC 2008), pp. 414-
417, 2008.
[4.38] F. S el o, M. B. Vahid a , M. B andolini, “Recon igu able Si RF
Recei e F on -Ends o Mul is anda d Radios”, 1s Eu opean
Con e ence on Wi eless Technology, (EuWiT 2008), pp.33-36, 2008.
[4.39] M. B andolini, P. Rossi, D. Mans e a, F. S el o, “Towa d
Mul is anda d Mobile Te minals-Fully In eg a ed Recei e s
Reque imen s and A chi ec u es”, IEEE T ansac ions on Mic owa e
Theo y and Techniques, ol. 53, no. 3, Ma ch 2005.
[4.40] R. Ba ak, A. Ghazel, F. Ghannouchi, “Op imized Mul is anda d RF
Subsampling Recei e A chi ec u e,” IEEE T ansac ions on Wi eless
Communica ions, ol. 8, no. 6, pp. 2901-2909, Jun. 2009.
[4.41] H. Pekau, J. W. Hasle , “A 0.18µm CMOS 2.1GHz Sub-sampling
Recei e F on end wi h Fully In eg a ed Second- and Fou h-O de
Q-Enhanced Fil e s,” IEEE In e na ional Symposium on Ci cui s and
Sys ems, pp. 3103-3106, 2007.
[4.42] D. Jakonis, K. Folkesson, J. Dab owski, P. E iksson, C. S ensson, “A
2.4-GHz RF Sampling Recei e F on -End in 0.18-μm CMOS,” IEEE
Jou nal o Solid-S a e Ci cui s, ol 40, pp. 1265-1277, Jun. 2005.
[4.43] Y. Vande pe en, W. Dehaene, G. Leus, “A Flexible Low Powe
Subsampling UWB Recei e Based on Line Spec um Es ima ion
Me hods,” IEEE In e na ional Con e ence o Communica ions
(ICC’2006), ol. 10, pp. 4694-4699, 2006.
CHAPTER
5
SUBSAMPLING TECHNIQUES
FOR NONLINEAR AND
MULTI-BAND APPLICATIONS
CHAPTER CONTENTS
5.1 S udied scena ios .................................................................................................. 121
5.2 Subsampling in nonlinea en i onmen s............................................................... 123
5.3 Subsampling o mul i-band sys ems .................................................................... 123
5.4 Subsampling o mul i-band sys ems in non linea en i onmen s ........................ 124
5.4.1 Implemen ed algo i hm ................................................................................... 125
5.4.2 Subsampling applica ions o mul i-band and nonlinea sys ems ................... 126
5.4.2.1 Spec um sensing in cogni i e adio ................................................... 127
5.4.2.2 Subsampling eedback loop o concu en dual band powe ampli ie
linea iza ion ......................................................................................................... 129
5.4.3 Op imiza ion o dual band ecei e s in nonlinea en i onmen s ..................... 132
5.4.3.1 Subsampling o concu en dual band and nonlinea sys ems using
mul iple clocking echniques .............................................................................. 132
5.4.3.2 Op imiza ion o he ecei e a chi ec u e ............................................ 137
5.4.3.3 Expe imen al alida ion ...................................................................... 139
5.5 Re e ences ............................................................................................................ 143
120 Chap e 5: Subsampling echniques o nonlinea and mul i-band applica ions
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
Chap e 5 desc ibes he addi ional challenges o implemen ing
subsampling echniques o mul i-band and nonlinea applica ions. A i s sec ion
o his chap e is dedica ed o in oduce he con ex whe e hese applica ions a e
use ul, while he nex wo sec ions in oduce he p e iously published exp essions
o implemen , sepa a ely, subsampling in mul i-band and non linea scena ios. A
ou h sec ion in eg a es bo h e ec s, de ailing he designed algo i hm o ind he
alid sampling equency anges and desc ibing an op imiza ion o a pa icula
case o dual band ecei e s in a nonlinea en i onmen . This op imiza ion is
based on he mul iple clocking echniques desc ibed in Chap e 4 and on a mul i-
il e s uc u e implemen a ion be ween he S&H and he ADC. Finally his
op imiza ion is expe imen ally alida ed.
Chap e 5: Subsampling echniques o nonlinea and mul i-band applica ions 121
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
5.1 S udied scena ios
Due o he eme gence o se e al co-exis ing wi eless echnologies in he
cellula indus y, he e is a end o design mul i-s anda d ecei e s a ge ing he
op imiza ion o hei lexibili y, simplici y and powe consump ion.
The wo main d awbacks o he sys ems based on subsampling a e he ji e
and he he mal olded noise, which make sys em implemen a ion e en mo e
di icul o mul i-band o non-linea applica ions. The e is a challenge when using
subsampling concu en ly in a mul i-signal en i onmen and/o nonlinea
condi ions, because he eplicas o he gene a ed ha monics a e olded back in he
band o in e es and may o e lap wi h he desi ed signals. This issue was
add essed in [5.1] whe e a uni e sal o mula o subsampling in nonlinea sys em
was de eloped o single band applica ions. In dual band ecei e applica ions, he
main p oblem o subsampling is he possible o e lapping be ween he eplicas o
he wo desi ed signals in he IF equency band. This p oblem was s udied in
[5.2] o mul iband linea and non in e e ing en i onmen .
As one example o applica ion, a dual band subsampling ecei e has been
p oposed o use in a eedback loop o a dual band ansmi e o linea iza ion
pu poses [5.3] using digi al p edis o ion. In [5.4] a subsampling ecei e o dual
band applica ions was p oposed, due i s simplici y, o allow he cogni i e adio
sense di e en bands and check and see i hey a e in use. In [5.4], he designed
subsampling ecei e does no conside any in e e e s, ha monics o
in e modula ion e ec s.
This chap e ex ends he abo e s udy [5.3,5.4] by op imizing he SNR o
concu en dual-band signals a he ecei e in a mul i-signal o nonlinea
en i onmen . The equi emen o inc easing he analog bandwid h and educing
he e ec o he olded noise leads o p opose new ecei e opologies wi h he
objec i e o imp o ing hese ea u es o a la ge numbe o communica ion
s anda ds. In e e ences and spu ious signals in he ecei ed spec um can be
ea ed as in e modula ion p oduc s using he same op imiza ion echnique ha
will be desc ibed la e , so when hese signals a e subsampled he esul ing aliasing
componen s wi h hese unwan ed and spu ious signals mus no o e lap wi h he
desi ed signal.
As an addi ional bene i , hese ex a condi ions used in he sampling
equency selec ion can lead o mo e elaxed RF il e equi emen s, due o he
known unwan ed signals in he spec um will no a ec he desi ed signal
bandwid h a IF and, he e o e, hey will be il e ed mo e easily a e being
subsampled.
128 Chap e 5: Subsampling echniques o nonlinea and mul i-band applica ions
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
(a) (b)
Figu e 5.6 (a) Subsampling based ecei e o spec um sensing in cogni i e adio
sys ems and (b) measu emen se up o alida ing spec um sensing concep using
subsampling ecei e
As an example o spec um sensing using a subsampling ecei e , wo RF
bands a e selec ed: he o icial digi al ideo b oadcas ing band a 698-752 MHz,
and an unlicensed band a 902-928 MHz. Wi h hese wo bands and using he
algo i hm desc ibed in he p e ious sec ion, a subsampling equency o 255 MHz
is selec ed.
Figu e 5.6b [5.4] shows he measu emen se up used o spec um sensing.
Two signal gene a o s a e used o ob ain he wo RF bands, hen he bands a e
combined using a powe combine and passed in o he ecei e . A SP De ices
de elopmen boa d using wo TI ADS5474 ADCs [5.9,5.10] ope a ing in a ime
in e lea ed manne is used as he subsampling de ice. A logic analyze is used o
cap u e he digi al da a s eaming om he ADC boa d, while ano he signal
gene a o p o ides a clock sou ce o he ADC.
A h ee channel signal is sen in he DVB band, while a 2 channel signal is
sen in he unlicensed band. Di e en powe le els a e con igu ed o each
channel o simula e di e en ecei ed signals. Figu e 5.7a shows he spec a o
hese wo bands in he RF domain. Figu e 5.7b [5.4] shows he wo bands
subsampled using a equency o 255 MHz. Since he ADCs a e ope a ing in a
ime in e lea ing ashion, he di e ences be ween each ADC may cause gain
misma ches and iming skew [5.11].

Chap e 5: Subsampling echniques o nonlinea and mul i-band applica ions 129
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
(a) (b)
Figu e 5.7 Spec a o (a) he inpu RF signal o he ecei e and (b) he
subsampled RF signal o bands (698-752 MHz, 902-928 MHz) using a
subsampling equency o 255 MHz
Figu e 5.8 [5.4] shows he inpu signals o e laid wi h hei subsampled
ou pu a e digi al demodula ion. Wi h he subsampling ecei e , he cap u ed
signal has app oxima ely a 50 dB signal o noise loo .
(a) (b)
Figu e 5.8 Spec a o he inpu and il e ed ou pu baseband signals o he 698-
752 MHz band (a) and 902-928 MHz band (b)
The echnique may be ex ended o mul iple bands, whe e changing he
subsampling clock may allow di e en RF bands o be demodula ed concu en ly.
In [5.4], he cogni i e adio senses up o 14 bands, sensing wo bands a any gi en
ime.
5.4.2.2 Subsampling eedback loop o concu en dual band powe
ampli ie linea iza ion
The powe ampli ica ion (PA) uni is ypically he mos ine icien
componen in wi eless ansmi e s. This is caused by an in e se ela ionship ha
exis s be ween e iciency and signal quali y based on he signal powe being
ansmi ed [5.12]. A low inpu powe , e iciency is low and he ampli ie
ope a es in a linea beha io , which esul s in good signal quali y a he ampli ie
ou pu . Howe e , ope a ing he ampli ie a i s highes e iciency s a e close o he
maximum ou pu powe causes he gain cha ac e is ics o become comp essed,
and he inpu -ou pu ela ionship becomes nonlinea . The nonlinea beha io
130 Chap e 5: Subsampling echniques o nonlinea and mul i-band applica ions
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
educes he in-band signal quali y and causes ou -o -band spec al eg ow h.
Nonlinea i y is u he complica ed in a dual band ope a ion, whe e he de ice
p oduces many in e modula ion and c oss modula ion signals. This in e se
ela ionship causes di icul ies o he wi eless ope a o , and ypically a linea
ope a ion mode is used such ha signal quali y is good, and spec al eg ow h is
minimal and does no cause in e e ence in o he channels.
Digi al p edis o ion allows o he ope a ion o signal in he high
e iciency egion while educing spec al eg ow h and imp o ing signal quali y
[5.13]. This is pe o med by analyzing he inpu and ou pu signals o he powe
ampli ie , and gene a ing an in e se beha io al model (p edis o e ) o he
ampli ie . The cascade o bo h he digi al p edis o e and he powe ampli ie
esul s in a linea gain a he ou pu o he ull powe ange. The p oposed
a chi ec u e was illus a ed in Figu e 3.22 (sec ion 3.5.2).
A dual band PA ope a ing a 880 MHz and 1978 MHz is used o es he
subsampling eedback loop o concu en dual band linea iza ion [5.3]. Two
communica ion signals wi h 5 MHz bandwid hs a e sen a he cen e o hese
bands. The PA is p edic ed o ha e a 5 h o de nonlinea i y, and all he ha monics,
in e modula ion, and c oss-modula ion p oduc s up o 4 GHz a e accoun ed. In
addi ion, a 25 MHz gua d band is placed a ound each band equency o accoun
o he spec al eg ow h ha will happen du ing he ini ial analysis s age.
The ha monics, c oss-modula ion, and in e modula ion signals may be
igno ed; hei only es ic ion is o no lie inside he gua d band o he signals, in
o de o compu e he eg ow h e ec a he DPD block when he signal is
downcon e ed. The subsampling algo i hm ou lined in sec ion 5.4.1 calcula es
he minimal subsampling equency o be ween 619.7 MHz and 620.1 MHz.
Figu e 5.9a shows a simula ion o he RF spec a o all he componen s om DC
o 4 GHz, while Figu e 5.9b shows he subsampled componen s using a equency
o 619.8 MHz [5.3].
Chap e 5: Subsampling echniques o nonlinea and mul i-band applica ions 131
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
(a) (b)
Figu e 5.9 The (a) p edic ed RF undamen al and ha monics up o 4 GHz and (b)
subsampled esul using a sampling equency o 619.8 MHz
The same se up desc ibed in Figu e 5.6b is used o gene a e he dual band
signal, and cap u e he RF ou pu . Figu e 5.10a shows he RF spec a a he ou pu
o he PA [5.3]. Compa ed o Figu e 5.9a, he e is an ex a e m p, which is a 7 h
o de in e modula ion p oduc a 436 MHz. The ejec ion o he i and j e ms a e
due o he design o he PA ou pu ma ching ne wo k. Figu e 5.10b [5.3] shows
he no malized spec a o he subsampled RF PA ou pu . The e is a enua ion om
he uppe band signal caused by he limi a ion o he ADC’s bandwid h o 1.4
GHz. The cap u ed ime domain signal may be digi al il e ed and demodula ed
and o e ie e he ampli ie ou pu o he wo bands, and u he pos -p ocessing
can de e mine he digi al p edis o ion model.
132 Chap e 5: Subsampling echniques o nonlinea and mul i-band applica ions
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
(a) (b)
Figu e 5.10 (a) RF spec a a he ou pu o he PA and (b) no malized spec a o
he cap u ed subsampled signal using an ADC ope a ing a 619.8 MHz
5.4.3 Op imiza ion o dual band ecei e s in nonlinea
en i onmen s
An analysis wi h he objec i e o noise pe o mance op imiza ion is
p esen ed in his sec ion. An independen clock solu ion o he S&H, and ADC is
p oposed o limi he noise e ec s; and a bank o bandpass il e s is used o il e
ou mos o he aliased nonlinea and in e e ing componen s. Se e al di e en
subsampling a chi ec u es and il e con igu a ions a e analysed in heo e ical and
measu emen en i onmen s.
5.4.3.1 Subsampling o concu en dual band and nonlinea sys ems
using mul iple clocking echniques
The ecei e pe o mance is de ined in e ms o noise, linea i y, sensi i i y,
dynamic ange and bandwid h [5.14]. These pa ame e s de ine he applicabili y o
a ecei e o be employed o a gi en s anda d. In he case o implemen a mul i-
s anda d ecei e , i is necessa y o maximize he analog inpu bandwid h and a
he same ime, he es o pa ame e s mus be op imized in o de o co e as many
wi eless communica ion s anda d equi emen s as possible.
This pa o he hesis is ocused on an op imized concu en dual-band
subsampling ecei e o noise pe o mance and e sa ili y, in o de o co e mos
wi eless communica ion s anda ds. The op imiza ion akes ad an age o he
lexibili y p ope o subsampling, i is possible o s udy di e en alid al e na i es
o clock he ecei e in o de o maximize i s noise pe o mance.
In sec ion 3.4 he main sou ces o noise in a subsampling ecei e we e
desc ibed. Ji e noise mainly depends on he inpu signal equency (which is a
Chap e 5: Subsampling echniques o nonlinea and mul i-band applica ions 133
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
cha ac e is ic o he s anda d), while inc easing he sampling equency can
educe he olded he mal noise. As desc ibed in sec ion 4.3, clocking he S&H
and he ADC wi h he same clock limi s he maximum sampling equency o he
sys em o ha o he ADC, which is usually, signi ican ly lowe han he
maximum sampling equency o he S&H. O he wise, employing an addi ional
highe equency clocking he S&H i is possible o inc ease he SNR o he
ecei e . Mo eo e , o his dual-band applica ion, addi ional deg ees o eedom
can be achie ed using a mul iple clock scheme, in o de o co e a highe numbe
o dual-band combina ions o wi eless communica ion s anda ds.
Figu e 5.11 illus a es he olded noise (g een line) o he single clock
(Figu e 5.11a) and he mul iple clock (Figu e 5.11b) cases, conside ing o bo h
cases he same he mal noise le el a he inpu o he S&H ( ed line) and om he
ADC (blue line). Since he e ec i e noise bandwid h o he S&H is ypically
much la ge han ha o he ADC, he imp o emen achie ed a he S&H in
Figu e 5.11b is usually dominan . In Figu e 5.11b a BP il e will be necessa y in
o de o dec ease he ou -o -band noise olded by he second subsampling p ocess,
while a LP il e wi h a cu o equency equal o s/2 is enough in Figu e 5.11a.
Howe e , his BP il e migh educe he lexibili y o he ecei e when i is used
in mul i-band applica ions and in a nonlinea en i onmen . This wo k ies o ind
he op imal il e bandwid h ha educes he olded noise, while a oiding a
signi ican educ ion in he numbe o alid sampling equencies in o de o ind
a high alue wi hin his ange.
S&H
S&H
ADC
ADC
Be 1
Be 1
Be 2
Be 2
s
s1 s2
s/2 s/2
s1/2 s2/2
(a)
(b)
Figu e 5.11 Folded noise e ec s using single clock (a) and mul iple clock (b)
Figu e 5.12 shows he e ec s o a hi d o de nonlinea i y when a dual
band signal passes h ough a nonlinea subsampling ecei e . In he i s scena io
he S&H and he ADC a e clocked a he same a e. In he second scena io he
clock a e o S&H and ADC has been chosen di e en ly. Figu e 5.12a p esen s
he i s scena io whe e bo h S&H and ADC a e using he same clock a e. The
wo ca ie equencies a 880 MHz and 1.82 GHz a e sampled a 400 MHz, his
equency being calcula ed by he algo i hm desc ibed in sec ion 5.4.1. Using his
sampling equency, Figu e 5.12a also shows he di e en Nyquis bands a he
S&H inpu , along which he inpu signals and hei ha monics and

134 Chap e 5: Subsampling echniques o nonlinea and mul i-band applica ions
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
in e modula ion p oduc s a e dis ibu ed. The subsampled signals and hei
ha monics and in e modula ion p oduc s a e all olded o IF. A e subsampling,
he signal is il e ed and con e ed o digi al domain, whe e he Nyquis Theo em
is me . Simila ly, Figu e 5.12b p esen s he scena io whe e he S&H and ADC use
di e en clock a es. In o de o educe he olded noise e ec , he S&H sampling
equency is inc eased o 2 GHz, while he inpu RF signal a he S&H is he same
as in he case o Figu e 5.12a. Since he e is a second subsampling p ocess,
aliasing be ween he a ge signals and hei ha monics and in e modula ion
p oduc s ha e o be a oided and, he e o e, he second sampling equency has o
be ca e ully selec ed using he same me hod de ailed in sec ion 5.4.1. In his
pa icula case, 400 MHz has been chosen o be he second sampling equency.
s=400MHz 2=1.82GHz
1=880MHz
-2 1+ 2 - 1+ 2 2 1 3 1
1+ 2
- 1+2 2 2 1+ 2
2 2
S&H+LPF
s/2=200MHz
2 1+ 2 - 1+2 2
2 2
-2 1+ 2
1+ 2
- 1+ 2
2 1
3 1
ADC
2 1+ 2 s/2=200MHz
- 1+2 2
2 2
-2 1+ 2
1+ 2
- 1+ 2
2 1
3 1
s=2GHz
S&H+LPF
s/2=1GHz
-2 1+ 2 2 1
ADC
2 2
2 1+ 2 3 1
1+ 2
- 1+2 2 - 1+ 2
s/2=200MHz
(a)
(b)
Figu e 5.12 Folded e ec s o ha monics and in e modula ion p oduc s using a
single clock (a) and mul iple clock (b) echniques
Besides op imizing he ecei e on i s noise pe o mance, he a chi ec u e
based on wo independen clock sou ces is employed o ind a alid sampling
equency o mos o combina ions in case o dual band RF inpu o a hi d o de
nonlinea ecei e . F om he wo k p esen ed in [5.6] o some s udied
combina ions i was no possible o ind a sampling equency wi hou a oiding
o e lapping be ween he desi ed signals and hei ha monics. Wi h he objec i e
o co e all he s udied cases, a bank o il e s be ween he S&H and he ADC was
p oposed o emo e some ha monics in o de o ha e mo e alid equency anges
o he second sampling p ocess. Only one il e in he bank o bandpass il e s
may be ac i e a any gi en ime. The gene al idea o his ecei e is illus a ed in
Figu e 5.13.
BP1
BPn
s1 s2
LNA S&H ADC DSP
Figu e 5.13 Op imized a chi ec u e based on mul iple clocking and BP il e s
Chap e 5: Subsampling echniques o nonlinea and mul i-band applica ions 135
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
These il e s will emo e some ha monics and in e modula ion p oduc s
elaxing he equi emen s o ind a alid second sampling equency. The
d awback o his a chi ec u e is a mo e es ic i e speci ica ion abou whe e he
desi ed signal mus be olded by he S&H wi hou being il e ed. The e o e, in
some cases, his es ic ion can lead o a low alid i s sampling equency (a he
S&H), which leads o mo e olded noise han when an a chi ec u e wi hou using
a bank o il e s is employed.
Figu e 5.14 demons a es he subsampling equency plan selec ion used
o he op imized a chi ec u e. Fi s , he dual band signal’s cen e equencies a
( 1, 2), bandwid hs (BW1, BW2), and an es ima e o he nonlinea i y o de o he
sys em a e used o p edic he numbe o in e modula ion, ha monics, and c oss
modula ion p oduc s. Then, he algo i hm ou lined in sec ion 5.4.1 gene a es he
ange o alid subsampling equencies o he S&H, F1. A loop is en e ed o ind
he maximum subsampling equency (less han o equal o he S&H maximum
equency ope a ion), whe e bo h dual band signal’s subsampled IFs all in o one
band pass il e in he il e bank, deno ed by BPselec ed. A signal’s subsampled IF
can be de e mined by he ollowing equa ion:
oddis
loo i em
e enis
loo i em
s
x
sxs
s
x
sx
i x
2/
),(
2/
),(
1
11
1
1
(5.5)
whe e x is he inpu equency be o e subsampling and x is ei he 1 o 2,
i x is he equency o he signal a e subsampling, s1 is he subsampling
equency, and em(.) is he emainde o he di ision ope a ion.
136 Chap e 5: Subsampling echniques o nonlinea and mul i-band applica ions
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
Nonlinea i y
O de
Band 1
1, BW1
Gene a e in e modula ion, ha monics,
c oss modula ion
Gene a e ange o subsampling
equencies in he i s s age, F1
i 11, i 12 bo h lie in
one o BPFs (BPselec ed)?
Selec lowe
s1 om F1
Compu e i 11, i 12 om
1, 2 and s1
Band 2
2, BW2
Se s1 o maximum a ailable
subsampling equency
s1
NO
YES
s2
Gene a e ange o subsampling
equencies in he second s age, F2
Se s2 o maximum a ailable
subsampling equency
Subsample band 1, band 2,
in e modula ion, ha monics, and c oss
modula ion using s1
Fil e ou o band signals using
BPselec ed
Compu e i 21, i 22 om
1, 2, s1 and s2
STOP
Figu e 5.14 Algo i hm low diag am o compu ing op imal subsampling
equencies in he p oposed a chi ec u e
A e de e mining he i s alid subsampling equency, s1, all he signals
gene a ed by he nonlinea i y a e subsampled. The esidual signals ou o he
bandpass il e band a e emo ed om u he analysis. The algo i hm p esen ed
in sec ion 5.4.1 is hen e-used o gene a e ano he ange o alid subsampling
Chap e 5: Subsampling echniques o nonlinea and mul i-band applica ions 137
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
equencies o he ADC, F2. The subsampling equency o he ADC, s2, will be
selec ed as he closes o he maximum sampling equency o he ADC wi hin he
ange F2.
5.4.3.2 Op imiza ion o he ecei e a chi ec u e
In o de o op imize he noise pe o mance o a dual band ecei e in a non
linea scena io a pa icula case has been esea ched in [5.6], whe e se en inpu
equencies ha e been selec ed o s udy he selec i e combina ions o di e en
dual band applica ions. These chosen s anda ds a e WCDMA (V) a 880 MHz,
GSM-DCS a 1.82 GHz, WCDMA (I) a 2.12 GHz, Blue oo h a 2.4 GHz,
WiMAX a 3.5 and 5.8 GHz, and 802.11a a 5.2 GHz.
Since he main ocus is o co e he maximum numbe o s anda ds, i is
manda o y o use an S&H be o e he ADC in o de o ha e enough analog
bandwid h. The S&H ea u es om Inphi wi h pa numbe 1821TH has been
selec ed as e e ence o his wo k, because o i s high inpu analog bandwid h (up
o 18 GHz), minimum ape u e ji e (50 s) and a maximum clock equency
equal o 2 GHz.
The i s s udied scena io is based on high esolu ion ADC wi h a high
sampling equency o educe he olded noise e ec . Wi h his ocus in mind he
selec ed ADC was a 12-bi ADS5400 om Texas Ins umen s wi h maximum
clock equency o 1 GHz [5.15]. Using a sampling equency o almos 1 GHz, i
is possible o co e all he dual band applica ions, as illus a ed in Figu e 5.15a
(Case 1) [5.6], whe e he meaning o axis x is de ailed in Table 5.3. This able
de ines each dual-band signal scena io as he combina ion o wo di e en
communica ion s anda ds. Using as e e ence he ypical esolu ion gi en by he
da ashee s, he heo e ical SNR o each dual-band scena io was es ima ed om
equa ions (3.21) and (4.3), aking in accoun he ji e and he olded noise e ec s
espec i ely.
Ano he op ion is o use a highe esolu ion ADC, like he 14-bi ADS5474
om Texas Ins umen s (Case 2 in Figu e 5.15a). This de ice was selec ed
because i s maximum sampling equency is 400 MHz. Howe e , as shown in
Figu e 5.15a, his op ion is less lexible, because i is no possible o ind any
sampling equency lowe han 400 MHz o he i s h ee scena ios.
In o de o imp o e he SNR wi hou losing lexibili y, wo s eps
subsampling app oach is p oposed, whe e he sampling equency o S&H was se
a a ound 2 GHz and he sampling equency o ADC a a ound 1 GHz (Case 3 in
Figu e 5.15a). The e o e, a heo e ical 3 dB imp o emen is achie ed om
equa ion (4.6) in espec o Case 1. Howe e , since his i s app oach is
implemen ed wi hou BP il e s, a new olded noise e ec will be added om
equa ion (4.3) because a second subsampling p ocess may be necessa y. Despi e
no using BP il e s, no e ha a LP il e wi h a cu o equency equal o s1/2
144 Chap e 5: Subsampling echniques o nonlinea and mul i-band applica ions
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
[5.4] A. Kwan, S. A. Bassam, F. M. Ghannouchi, “Sub-sampling Technique
o Spec um Sensing in Cogni i e Radio,” IEEE Radio and Wi eless
Symposium (RWS’2012), pp. 347-350, 2012.
[5.5] H. Hashemi, A. Hajimi i, “Concu en mul iband low-noise
ampli ie s- heo y, design, and applica ions,” IEEE T ansac ion on
Mic owa e Theo y and Techniques, ol.50, no.1, pp.288-301, Jan
2002.
[5.6] J. G. Oya, A. Kwan, S. A. Bassam, F. Muñoz, and F. M. Ghannouchi,
“Op imiza ion o Subsampling Dual Band Recei e s Design in a
Nonlinea Sys ems,” IEEE MTT-S In e na ional Mic owa e
Symposium Diges (IMS’2012), pp. 1-3, Mon eal, QC, Canada, June
2012.
[5.7] J. H. Kim, H. Wang, H-J. Kim H-J, J-U. Kim, “Bandpass Sampling
Digi al F on end A chi ec u e o Mul i-Band Access Cogni i e
Radio,” IEEE Global Telecommunica ions Con e ence (GLOBECOM
2009), pp. 1-6, 2009.
[5.8] S. Haykin, “Cogni i e Radio: B ain Empowe ed Wi eless
Communica ions”, IEEE Jou nal on Selec ed A eas in
Communica ion, j. 48, no. 2, pp. 201-220, 2005.
[5.9] ADS5474, 14-Bi 400-MSPS Analog- o-Digi al Con e e , Texas
Ins umen s. Da ashee .
[5.10] ADS5474, ADS5440/44/63/74 EVM, Texas Ins umen s. Use Guide.
[5.11] N. Ku osawa, H. Kobayashi, K. Ma uyama, H. Sugawa a, K.
Kobayashi K, “Explici Analysis o Channel Misma ch E ec s in
Time-In e lea ed ADC Sys ems,” IEEE T ansac ions on Ci cui s and
Sys ems I Fundamen al Theo y Applica ions, ol. 48, no. 3, pp. 261-
271, 2001.
[5.12] E. McCune, “High-e iciency, Mul i-mode, Mul i-band Te minal
Powe Ampli ie s,” IEEE Mic owa e Magazine, ol. 6, no. 1, pp. 44-
55, 2005.
[5.13] F. M. Ghannouchi, O. Hammi, “Beha io al modeling and
p edis o ion,” IEEE Mic owa e Magazine, ol. 10, no. 7, pp. 52–64,
Dec. 2009.
[5.14] F. Agnelli e al., “Wi eless Mul i-S anda d Te minals: Sys em
Analysis and Design o a Recon igu able RF F on -End,” IEEE
Ci cui s and Sys ems Magazine, ol. 6, no.1, pp. 38-59, Jan. 2006.

Chap e 5: Subsampling echniques o nonlinea and mul i-band applica ions 145
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
[5.15] ADS5400, 12-Bi 1-GSPS Analog- o-Digi al Con e e , Texas
Ins umen s. Da ashee .
[5.16] E8257D PSG, Mic owa e Analog Signal Gene a o , Agilen
Technologies. Da ashee .
[5.17] ZX60-6013E, Connec o ized Ampli ie 50 Ω, Minici cui s. Da ashee .
[5.18] E8663D PSG, RF Analog Signal Gene a o , Agilen Technologies.
Da ashee .
[5.19] SMIQ, Vec o Signal Gene a o , Rohde & Schwa z. Da ashee .
146 Chap e 5: Subsampling echniques o nonlinea and mul i-band applica ions
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
CHAPTER
6
CONCLUSIONS AND
POSSIBLE FUTURE
DIRECTIONS
CHAPTER CONTENTS
6.1 Conclusions .......................................................................................................... 148
6.2 Possible u u e di ec ions ...................................................................................... 149
This chap e p esen s he main conclusions and con ibu ions o his hesis
and desc ibes he possible u u e di ec ions.
148 Chap e 6: Conclusions and possible u u e di ec ions
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
6.1 Conclusions
This hesis exploi s he bene i s o he subsampling based sys ems o be o
he RF down-con e sion s age in ecei e s o So wa e De ined Radio (SDR).
The e a e h ee dis inguishable con ibu ions in his hesis wo k ha come om
he da a acquisi ion design o es ing wideband mul i-s anda d pu poses, he
noise pe o mance op imiza ion om mul iple clocking echniques, and ex ension
o subsampling echniques o mul i-band and non linea en i onmen s. All hese
con ibu ions ha e been heo e ically s udied and expe imen ally alida ed.
Since he SDR objec i es can only be ob ained wi h a ecei e whose
on -end accommoda es a wide ange o equency bands and channel
bandwid hs, subsampling a chi ec u es a e p esen ed as a easible al e na i e o
he di e en possible ecei e a chi ec u es e iewed in Chap e 1. Mo eo e , a
subsampling scheme has been selec ed because i leads o ex emely lexibili y
and simplici y, he educ ion o he numbe o componen s o a minimum being a
key when he same ecei e has o p ocess di e en s anda ds. In addi ion, since
he digi iza ion s age will be placed jus a e he an enna by using subsampling
echniques, mos o he signal p ocessing will be pe o med in he digi al domain,
hus a oiding he limi a ions o he cu en ADCs. The e o e, he p oposed
a chi ec u e educes he numbe o analog o building blocks, and elaxes he
speci ica ions o he ADC, which is he cu en bo leneck owa ds a ully digi al
mul i-s anda d so wa e adio.
A e desc ibing he subsampling idea in Chap e 2, de ailing he main non
ideali ies p ope o hese echniques (i.e., ji e noise and olded he mal noise),
Chap e 3 discusses he use ulness and po en ial o subsampling echnique o
design a simple and lexible uni e sal ecei e . A da a acquisi ion module o
es ing wi eless ecei e s based on subsampling has been p esen ed which co e s
mos p esen wi eless communica ion s anda ds equi emen s wi h only one single
boa d. Expe imen al esul s o he p oposed module show ( o a 20 MHz signal
bandwid h) an ENOB highe han 8 bi s up o 5 GHz cen e equency. Ano he
cha ac e is ic o he implemen ed module is i s simplici y, wi h only a ew
componen s on a p in ed-ci cui boa d. These esul s show ha , o es ing
pu poses, he subsampling based ecei e is a iable al e na i e o o he ypical
ecei e a chi ec u es, wi h enhanced econ igu abili y and p og ammabili y.
As a second main con ibu ion, he noise pe o mance o he subsampling
based ecei e has been op imized by using a no el me hod based on mul iple
clocking echniques in o de o educe he olded noise e ec . An analy ical
exp ession o he imp o emen ac o in he SNR wi h espec o he single-clock
solu ion has been ob ained. Finally, a new e sion o he da a acquisi ion module
o es ing o wi eless ecei e s has been p esen ed. Fo he selec ed equency
plan, wi h wo successi e subsampling p ocesses, he ENOB has been shown o
imp o e in app oxima ely 0.5-1 bi . Expe imen al esul s show, o a 20 MHz
Chap e 6: Conclusions and possible u u e di ec ions 149
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
signal bandwid h, an ENOB o mo e han 9 bi s up o 2.9 GHz, and mo e han 8
bi s up o 6.5 GHz cen e equency. Measu emen esul s show ha he design
co e s he mos impo an wi eless s anda ds (i.e., GPS, GSM, GPRS, UMTS,
Blue oo h, Wi-Fi, WiMAX) in e ms o uning equency and noise pe o mance.
As a hi d main con ibu ion, he subsampling concep s ha e been
ex ended o mul i-band and nonlinea sys ems, whe e he e is an addi ional
p oblem abou he ha monics and di e en channel which migh be olded in he
band o in e es . A e he challenges and issues on inding he alid subsampling
equencies in mul i-band and nonlinea sys ems ha e been discussed, an
op imized design app oach o a concu en dual band mul i-s anda d subsampling
ecei e in a nonlinea and / o in e e ing en i onmen has been p esen ed,
p oposing a mul i- il e a chi ec u e along wi h dual subsampling p ocess. In
addi ion, an e icien algo i hm has been de eloped in o de o ind he alid
sampling equencies, inc easing he lexibili y o he ecei e and co e ing he
maximum numbe o dual band applica ions o di e en communica ion
s anda ds. As an addi ional ad an age, hese condi ions used o selec he alid
sampling equency can lead o mo e elaxed RF il e equi emen s. Expe imen al
esul s p o ed he easibili y and he ad an ages o he p oposed a chi ec u e,
which can be used o di e en unc ionali ies in a wi eless communica ion
anscei e , on bo h he ansmi e and he ecei e side.
6.2 Possible u u e di ec ions
Fi s ly, u u e wo ks will be in ended o in eg a e all hese con ibu ions in
a unique da a acquisi ion boa d o mul i-s anda d and mul i-band es ing pu poses.
The e o e, implemen ing he designed bank o il e s be ween he S&H and he
ADC, i will be possible o ex end he unc ionali y o he PCB o mul i-band and
nonlinea en i onmen s.
Ano he u he objec i e o imp o e his hesis wo k will be he
implemen a ion o an au onomous p o o ype. A FPGA will be included in he
boa d o au onomously selec he op imal sampling equency and he op imal
il e bandwid h om he de eloped algo i hms, ins ead o ca ying ou hese asks
by an ex e nal compu e . Also, his FPGA will pe o m he equi ed digi al signal
p ocessing, po en ially inc easing he use ulness o including his componen .
The selec ed sampling equency will be used o p og am a VCO in o de
o inc ease he au onomy o he boa d, a oiding use o an ex e nal clock gene a o .
Ne e heless, when he second sampling equency is no mul iple o he i s
sampling equency, a second VCO will be necessa y, inc easing he complexi y
and he o al powe consump ion. The e o e, o he u u e app oach will be o
ex end he algo i hm o dual-band and nonlinea scena io in o de o ind a new
sampling equency plan based on wo mul iple equencies, which educes he
powe consump ion a oiding a signi ican noise pe o mance penal y.

150 Chap e 6: Conclusions and possible u u e di ec ions
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
Mo eo e , hese esea ches can be associa ed o o he cu en p ojec s.
Besides inc easing he au onomy by using a unique da a acquisi ion boa d, a
second main u u e challenge is o u ilize he op imized dual-band ecei e wi hin
a cu en p ojec de eloped by he iRadio Labs, om he Uni e si y o Calga y.
This p ojec includes a subsampling ecei e in a eedback loop o a dual-band
ansmi e o linea iza ion pu poses, as desc ibed in Chap e 4. The e o e, he
ocus will be o imp o e he noise pe o mance in he ecei e , a oiding he
d awbacks o he p e ious wo k (desc ibed in sec ion 5.4.2.2), such as he
p esence o addi ional spu ious due o use an in e lea ed ADC a chi ec u e, o a
low analog bandwid h ha limi s i s use ulness o mul i-s anda d applica ions.
In addi ion, he inal objec i e o his p ojec consis s o a whole
ansmi e design whose p omising alida ion will be implemen ed by COTS.
Using subsampling, he S&H and he ADC implemen he eedback loop o
linea iza ion, whe eas a DAC is used o se he baseband I-Q signals p e iously o
be modula ed. A FPGA is connec ed o bo h daugh e boa ds, implemen ing he
digi al p e-dis o ion p ocess and he signal gene a ion om he connec ion wi h
he ADC and he DAC, espec i ely. The in e ace be ween he FGPA and he
daugh e boa ds has al eady implemen ed, as well as he communica ion wi h an
en elope modula o employed o inc ease he e iciency o he powe ampli ie .
On he o he hand, an al e na i e o clock he ecei e by using VCOs is
u ilizing a clock gene a o boa d based on di ec digi al syn hesize s (DDSs),
which is being cu en ly designed wi hin a p ojec de eloped by he Elec onics
Enginee ing G oup, om he Uni e si y o Se ille, and he company AT4
Wi eless. Clocking he p oposed ecei e wi h his gene a o can be a easible
al e na i e when a equency hopping implemen a ion is equi ed, in o de o
ecei e di e en inpu signals by apidly swi ching he ca ie among many
wi eless communica ion s anda ds.
CHAPTER
7
APPENDIX A: DATA
ACQUISITION SYSTEMS
BASED ON INTERLEAVING
TECHNIQUES
CHAPTER CONTENTS
7.1 Theo y o ope a ion .............................................................................................. 153
7.1.1 In e lea ing idea .............................................................................................. 153
7.1.2 Analysis o ime-in e lea ed ADCs ................................................................. 153
7.2 Time-in e lea ed ADCs alida ion ....................................................................... 155
7.2.1 Valida ion a simula ion le el .......................................................................... 155
7.2.2 Valida ion a expe imen al le el ...................................................................... 157
7.3 Calib a ion echniques .......................................................................................... 160
7.4 Implemen ed sys ems............................................................................................ 164
7.5 Re e ences ............................................................................................................ 166
The idea o connec ing se e al ADCs in pa allel is o maximize he o al
da a acquisi ion a e. In his appendix hese a chi ec u es will be in oduced,
ocusing his desc ip ion on hei ad an ages and incon eniences, such as he
152 Appendix A: Da a acquisi ion sys ems based on in e lea ing echniques
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
misma ches e o s be ween he in e lea ed ADCs, which will need o be
compensa ed. These heo e ical concep s will be alida ed a simula ion and
expe imen al le els. Finally, special a en ion is de o ed o he published
co ec ions me hods om a bibliog aphic s udy de ailing he mos app op ia e
calib a ion echniques in e ms o powe consump ion o digi al p ocessing
capabili ies.
Appendix A: Da a acquisi ion sys ems based on in e lea ing echniques 153
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
7.1 Theo y o ope a ion
7.1.1 In e lea ing idea
A ime-in e lea ed ADC ope a es M pa allel ADCs a di e en sampling imes,
such as illus a ed in Figu e 2.15 [7.1]. Ideally, he i h ADC, i = 0, ..., M – 1,
samples pe iodically he inpu signal a ime ins an s i, i+M, i+2M, wi h sample a e
s/M, whe e m=mTs and Ts=1/ s is he sampling pe iod o he ime-in e lea ed
ADC. The inal ou pu is c ea ed by mul iplexing all o he indi idual ADC
ou pu s in he p ope o de (e.g. ADC0, ADC1, …, ADCM – 1, ADC0, ADC1, ...).
The e o e, he inal e ec is as i he inpu signal we e sampled once e e y Ts
seconds, i.e., wi h sample a e s.
Al hough he da a acquisi ion a e is inc eased wi hou penal y o e o he
ADC ea u es, hese sys ems p esen se e al disad an ages. Fi s ly, i should be
no ed ha each indi idual ADC deals wi h he en i e analog inpu signal, and,
he e o e, i s S&H ci cui mus be able o p ese e he ull inpu signal bandwid h.
Secondly, di e en spu s a e caused by he misma ches be ween ADCs, i s
loca ion in he spec um being p edic ed in he nex sec ion.
7.1.2 Analysis o ime-in e lea ed ADCs
Le x( ) be an analog signal wi h Fou ie ans o m Xa(ω). Conside ha
he ime-in e lea ed ADC ou pu s he sequence:
),...(),(),...,(),...,(),(),(1210 MMm x x x x x x
(7.1)
De ine he disc e e- ime Fou ie ans o m by10:
)exp()()( kk j xX
(7.2)
Ideally, he samples a e spaced Ts seconds apa . Then, i can be shown
ha :
)2()( sas kX X
(7.3)
which is he well-known spec um ep esen a ion o a uni o mly sampled
signal. I esul s in a pe iodic spec um wi h a pe iod equal o he sampling a e
[7.2].
10 In he li e a u e, i is a common no a ional p ac ice o eplace ω k wi h a single
a iable ω´=ω k, called no malized equency. Since ω ep esen s o dina y
equency ( adians pe second), ω´ is exp essed in uni s o adians (pe sample).
Recall also ha by sampling he disc e e- ime Fou ie ans o m, we ob ain he
disc e e Fou ie ans o m (DFT) [7.1].
B. Papa i, D. Asemani, A. Khakpou , “A Wide-Band Time-In e lea ed A/D Con e e
Fo Cogni i e Radio Applica ion Wi h Adap i e O se Co ec ion,” 2011 Wi eless
Ad anced, pp. 144-148, 2011.
C. R. Pa key, M. T. Hun e , D. B. Ches e , W. B. Mikkael, “Simulink Modeling o
Analog o Digi al Con e e s o Pos Con e sion Co ec ion De elopmen and
e alua ion,” IEEE 54 h In e na ional Midwes Symposium on Ci cui s and Sys ems
(MWCAS 2011), pp. 1-4, 2011.
H. Pekau, J. W. Hasle , “A 0.18µm CMOS 2.1GHz Sub-sampling Recei e F on end
wi h Fully In eg a ed Second- and Fou h-O de Q-Enhanced Fil e s,” IEEE In e na ional
Symposium on Ci cui s and Sys ems, pp.3103-3106, New Yo k, July 2007.
J. M. D. Pe ei a, P. M. B. S. Gi ao, A. M. C. Se a, “An FFT-Based Me hod o E alua e
and Compensa e Gain and O se E o s o In e lea ed ADC Sys ems,” IEEE
T ansac ions on Ins umen a ion and Measu emen , Vol. 53, no. 2, pp. 423-430, Ap il
2004.
A. Pe aglia and S.K. Mi a, “Analysis o misma ch e ec s among A/D con e e s in a
ime-in e lea ed wa e o m digi ize ,” IEEE T ansac ions on Ins umen a ion and
Measu emen , Vol. 40, no. 5, pp. 831-835, Oc . 1991.
K. Poul on e al., “A 20 GS/s 8b ADC wi h a 1 MB memo y in 0.18 µm CMOS,” in IEEE
In . Solid-S a e Ci cui s Con . (ISSCC) Dig. Tech. Pape s, Vol. 1, pp. 318–496, 2003.
M. B. Romdhane, P. Loumeau, “Analog o Digi al Con e sion speci ica ions o Ul a
Wide Band ecep ion,” P oceedings o he Fou h IEEE In e na ional Symposium on
signal P ocessing and In o ma ion Technology, pp. 157-160, 2004.
H. H. Slim, P. Russe , “Digi al Au oma ic Calib a ion Me hod o a Time-In e lea ed
ADCs Sys em used in Time-Domain EMI Measu emen Recei e ,” IEEE In e na ional
Symposium Elec omagne ic Compa ibili y (EMC 2011), pp. 476-479, 2011.
Y. R. Sun, “Gene alized Bandpass Sampling Recei e s o So wa e Radio,” Doc o al
Disse a ion, Royal Ins i u e o Technology, School o In o ma ion and Communica ion
Technology (ICT), S ockholm, Sweden, 2006.
F. S el o, M. B. Vahid a and M. B andolini, “Recon igu able Si RF Recei e F on -
Ends o Mul is anda d Radios,” Eu opean Con e ence on Wi eless Technology (EuWiT
2008), pp. 33-36, 2008.
M. Tamba, A. Shimizu, H. Munaka a, T. Komu o, “A Me hod o Imp o e SFDR wi h
Random In e lea ed Sampling Me hod,” In e na ional Tes Con e ence 2001, pp. 512-
520, 2001.

R. G. Vaughan, N. L. Sco and D. R. Whi e “The Theo y o BandpassSmpling,” IEEE
T ansac ions on Signal P ocessing, Vol. 39, pp. 1973-1984, Sep. 1991.
Y. Vande pe en, W. Dehaene, G. Leus, “A Flexible Low Powe Subsampling UWB
Recei e Based on Line Spec um Es ima ion Me hods,” IEEE In e na ional Con e ence
on Communica ions, pp. 4694-4699, 2006.
N. Vun, A. B. P emkuma , “ADC Sys ems o SDR Digi al F on -End,” P oceedings o
he Nin h In e na ional Symposium on Consume Elec onics (ISCE 2005), pp. 359-363,
2005.
258 Appendix C: Publica ions
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
9.5.3 Con e ence communica ions
“High F equency Analog- o-Digi al Con e sion Based on
Subsampling”, XXIV Con e ence o Design o Ci cui s and In eg a ed
Sys ems (DCIS’2009), 2009.
High equency Analog- o-Digi al Con e sion based on
subsampling
José Ramón Ga cía Oya#, An onio Ju ado Díez*, Fe nando Muñoz Cha e o#, An onio To alba Silgado#
#Depa amen o de Ingenie ía Elec ónica, Uni e sidad de Se illa
c/ Camino de los Descub imien os s/n 41092 Se illa, Spain
1[email p o ec ed]
3[email p o ec ed]
4[email p o ec ed]
*AT4 wi eless
c/ Se e o Ochoa 2 29590, Málaga, Spain
2[email p o ec ed]
Abs ac — The ocus o his wo k is he implemen a ion o a
Analog- o-Digi al Con e e Sys em using echniques based on
subsampling. I s main objec i e is o imp o e he ea u es o a
ecei e used o wideband communica ions educing he
numbe o elemen s o he sys em wi h a highe lexibili y and
esolu ion. This p oposed sys em is based on comme cial de ices,
mainly a Low Ji e and Wideband Sample&Hold and a High
Resolu ion In e media e- equency Analog- o-Digi al Con e e .
Keywo ds— Analog- o-Digi al Con e e , ENOB, Ji e ,
Sample&Hold, So wa e De ined Radio, SNDR, The mal Noise
I. INTRODUCTION
In gene al, he analog building blocks de e mine he
sensi i i y and selec i i y o a ecei e . The Analog- o-Digi al
Con e e (ADC) is becoming an ex emely impo an block
o he ecei e a chi ec u e, because he place o he ADC in o
he ecei e a chi ec u e ma ks which unc ions a e
implemen ed wi h analog ci cui y and wha unc ionali y is
done in he digi al signal p ocesso .
Nowadays he e is a end o inc easing he esolu ion and
speed o he ADC o a ecei e so ha i is possible o place i
close he an enna. Thus he analog on end is g ea ly
simpli ied and he lexibili y o he ecei e is imp o ed.
The cu en s a e o he a o high- equency ADCs does
no allow he use o he ADC di ec ly in he RF domain o ge
he pa adigm o he So wa e De ined Radio (SDR). Howe e ,
many esea ch p ojec s ocus on inding new solu ions
owa ds he SDR, whe e all he analog unc ionali ies (e.g.
mixe s, il e s, ampli ie s, modula o s/demodula o s) a e
pe o med in he digi al domain. Nowadays, wi h a
con en ional Analog- o-Digi al con e sion on he heade -
ecei e , only 7-8 bi s a e ob ained wi h 3 GS/s Analog- o-
Digi al Con e e s.
The basic speci ica ions o he implemen ed sys em in his
pape a e illus a ed wi h he Table I:
TABLE I
SPECIFICATIONS OF THE IMPLEMENTED SYSTEM
Analog Inpu F equency
DC-3 GHz
Signal Bandwid h
20 MHz
Sampling F equency
400-500 MHz
Resolu ion
9 bi s
In his pape an a chi ec u e based on subsampling is
p esen ed in o de o imp o e he lexibili y o a
communica ion ecei e , educing he numbe o analog
componen s. Using echniques based on subsampling, like
hose shown in his pape , we implemen an Analog- o-Digi al
Con e sion Sys em wi h a esolu ion highe han 8 bi s and
wi h a maximum o 3,1 GHz o cen e equency o he RF
signals and close o 9 bi s wi h a maximum cen e equency
o 2 GHz inpu app oxima ely.
This pape is o ganized as ollows: sec ion II in oduces
subsampling and heo e ical concep s on how o ob ain he
op imal subsampling equency and o e alua e e ec o he
main non-ideali ies ha will limi he implemen ed sys em.
These non-ideali ies help us o jus i y he choice o
componen s in sec ion III. In sec ion IV, expe imen al issues
p esen he a ained esul s abou noise and dis o ion. The
pape inishes wi h conclusions in sec ion V.
II. THEORETICAL STUDIES ON SUBSAMPLING
A. Concep o subsampling and op imal equency choice
In his sec ion we s udy he way o calcula e op imal
subsampling equency, using he signal bandwid h (BW) and
i s ca ie equency ( c), in o de o a oid aliasing and o
main ain he copy gene a ed be ween - s/2 and s/2 and he
eplicas as a as possible o he desi ed signal [1-6].
He ea e , we will use he ollowing no a ions (see Fig.1):
 s: subsampling equency
 BW: signal bandwid h
 c: ca ie equency
 B: c+BW/2
c
BW
s
s
Fig. 1 Subsampling concep (m=3)
The minimal sampling equency is es ablished by he
Nyquis Theo em, s > 2B. Howe e , we can a oid aliasing
wi h less s when exp ession (1) is ue:
1
22




m
BW
m
BW c
s
c
(1)
m is an en i e numbe whose meaning is he numbe o
copies o he o iginal signal ha appea s in he ange [0, c-
BW/2]. The maximum numbe o copies needed o a oid
aliasing is calcula ed by he exp ession (2):
)
2
(
max BW
BW
loo m c

(2)
loo (x) is he en i e numbe nea es o x y less han x. The
las exp ession es ablishes he limi s o s, he op imal
equency in his ange. Conc e ely, he op imal alue o a oid
aliasing is one ha p oduces a copy on s/4. This equency
equals:
odd
c
sm
4

(3)
modd is an en i e odd numbe mo e han 1:
 wi h modd = 5,9,13, … he e is no spec al in e sion
 wi h modd = 3,7,11, … he e is spec al in e sion
B. Main non-ideali ies
A gene al scheme o he implemen ed ecei e is shown in
he Figu e 2. I s main ad an age is i s simplici y, elimina ing a
la ge amoun o componen s in he adi ional he e odyne
s uc u e. Howe e , he speci ica ions o he Sample&Hold
a e much mo e es ic i e han in a adi ional ecei e . This
de ice will be he mos c i ical in ou sys em because i
p ocesses high equency signals.
Fig. 2 Schema ic o he ecei e implemen ed
The main non-ideali ies p oduced in Sample&Hold a e he
ollowing:
1) Ji e : Ideally, he inpu signal is sampled in equal
equency s in e als. Ne e heless, hese in e als a e
di e en due o ji e [7-9]. This ji e p oduces an inc emen
o he o al noise, hus limi ing he e ec i e numbe o bi s
(ENOB).
Ji e is p oduced by wo di e en sou ces: he phase noise
associa ed o he oscilla o and he ape u e ji e o he
Sample&Hold. A a i s app oxima ion we can conside hese
wo sou ces o ji e as non-co ela ed Gaussian s ochas ic
p ocesses.
Ape u e ji e o a Sample&Hold depends on he changes
o he h eshold ol age acco ding o he inpu ol age so ha
i s ea u e is dependen on he signal. This sec ion shows how
ji e a ec s Signal o Noise Dis o ion Ra io (SNDR) in he
Sample&Hold ou pu . The ocus is he es ablishmen o he
maximum allowed ji e s anda d de ia ion depending on
inpu equency and he esolu ion speci ica ions.
When he inpu is a sinusoidal signal
like
 
( ) sin 2 in
y A


, SNDR is de e mined by he
exp ession (4):










caseo he
e
N
A
SNDR
in
in
in
_:
)1(2
1
12:
4
1
2
222
2
222
2








(4)
Whe e
N

is he a e age powe noise and


is he ji e
s anda d de ia ion. To deduce he las exp ession, he spec al
densi y was in eg a ed be ween 0 and s/2.
To ob ain mo e ealis ic alues o he e ec o he ji e we
pe o med simula ions (by MATLAB® So wa e) while
conside ing only he noise in he signal bandwid h is
in eg a ed (20 MHz) and he ji e p esen ed by a s ochas ic
p ocess wi h an a e age o ze o and s anda d de ia ion


.
The equa ion (3) has been used o choose he op imal
subsampling equency
s
.
These simula ions e i y how he allowed ji e is lowe o
highe inpu equencies. Figu e 3 illus a es hese conclusions
o he ollowing cases:
 c=4GHz, s=484.84MHz
 c=2GHz, s=470.59MHz
 c=1GHz, s=444.44MHz
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
x 10-12
40
45
50
55
60
65
70
75
80
Des iación ípica de ji e (seg)
SNDR (dB)
c=2GHz, s=470.59MHz, BW=20MHz
c=1GHz, s=444.44MHz, BW=20MHz
c=4GHz, s=484MHz, BW=20MHz
Fig. 3 Ji e e ec depending on inpu equency
2) O e lapped The mal Noise: a ypical Sample&Hold
p oduces kT/C noise. To simpli y he noise analysis we
conside a sampling model as i is illus a ed in Figu e 4 [10].
This scheme consis s o an inpu band pass signal (
in
V
), wi h
an associa ed noise (
in
N
), which is il e ed and la e applied
o a Sample&Hold modelled by a swi ch and capaci o . Due o
he mul iple o e lapping p oduced by subsampling, many
pa s o he signal spec um will be inside o he band o
in e es .
4kTRon Ron
Cs
s
Vin+Nin
NBW=Bneq
Zo<<Ron
Vou (nTs) + Nou
Fig. 4 Scheme o noise analysis
The equi alen swi ch esis ance has a whi e noise powe
spec al equal o
ON
kTR4
and i is il e ed by he ans e
unc ion:
SON C R
H

21
1
)( 

(5)
The esul is a il e ed whi e noise wi h a powe equal o
S
C
kT
, whose alue is independen o he esis ance and
subsampling equency.
We assume:
SON
sCR

2
1

(6)
Then, due o he subsampling p ocess, we can app oxima e
all noise powe o e lapped be ween 0 and
2
s
, as illus a ed
in Figu e 5.
s/2 s 2 s 3 s 4 s
12
1
log),(
son
ou CRj
N

Fig. 5 Subsampling e ec o e he Sample&Noise
Al hough he o al noise powe is no dependen on
subsampling equency, noise loo educes i his equency
g ows. Conc e ely, noise loo is educed 3 dB i subsampling
equency is doubled. Thus, i is con enien o choose a
maximum subsampling equency o dis ibu e noise loo
h oughou he en i e Nyquis band.
3) O he conside a ions in choosing Sample&Hold:
 I s bandwid h should be maximized in o de o no il e
GHz signals.
 Al hough i has a high bandwid h, he Sample&Hold
should be able o sample equencies close o MHz, due
o hese alues a e he maximum sample equencies o
high esolu ion comme cial A/D con e e s.
 I s ou pu ENOB can be limi ed by linea i y, so we
should s udy i s THD and SFDR.
III. USED COMPONENTS
The heo e ical s udy ha was b ie ly explained in sec ion
II allows ixing he speci ica ions o he main building blocks
o he sys em, i.e., he Sample&Hold and he ADC.
A e a s udy on comme cial componen s, we decided o
use an ex e nal Sample&Hold o he A/D Con e e since an
in e nal Sample&Hold bandwid h is limi ed o 3 GHz
app oxima ely wi h a esolu ion a ound 7-8 bi s. Howe e ,
when using an ex e nal Sample&Hold, i is possible o ob ain
a highe esolu ion o a wide bandwid h.
The chosen Sample&Hold is he Inphi 1821 TH [11], wi h
he ollowing ea u es:
 Wide bandwid h (18 GHz).
 Wide equency ange (0-6 GHz) o 10-bi linea i y.
 The mal noise is no an obs acle (SNR>60 dB)
 Low phase noise, wi h SNR > 60 dB in he ange 0-3,2
GHz.
 Capabili y o sample a he in e es ed equency (a ound
500 MS/s)
This equi emen (500 MS/s) is gi en by he maximum
sampling equency o comme cial 10-bi A/D con e e s,
since using a maximum sampling equency is con enien in

o de o educe he o e lapping noise e ec s. Conc e ely, he
A/D con e e chosen is E2V AT84AS001 [12].
IV. EXPERIMENTAL RESULTS
The ecei e sys em implemen ed is illus a ed in Figu e 6.
In his sys em he GHz inpu signal is gene a ed and con e ed
o di e en ial o be sampled by he Sampled&Hold. By he
o he side, a unique clock signal is gene a ed o he
Sample&Hold and he ADC ha is con e ed o di e en ial
signal oo, wi h baluns a lowe equencies. We used he
e alua ion boa ds o hese de ices connec ed by cables. Also
DC blocks and bias ee a e used o block he DC signal and o
adap he impedances.
This sys em can be used o illus a e some subsampling
e ec s s udied in he p e ious sec ions, like he o e lapping
he mal noise wi hin band signal, which is dependen on he
subsampling equency. This e ec is illus a ed in Figu e 7,
whe e we use he op imal equency so ha hi d o de
ha monics a e o e lapped wi h he signal (as i is deduced
om he heo e ical s udies), and his way we only measu e
he noise e ec (SNR) in eg a ed in he signal bandwid h,
because he second o de ha monics will be he u hes
possible o he desi ed signal. Signal ampli ude is equal o
-1 dBm and noise is in eg a ed in 20 MHz.
Signal
Gene a o
SMIQ
Rohde
&
Schwa z
Balun
ZAPDJ-2
Cable N-SMA
DC-Block
DC-Block
Cable SMA-SMA
Cable SMA-SMA
S&H
1821TH
Inphi
Clock
Gene a o
E8257D
Agilen Adap.
SM-SF
Balun
ZFSCJ-2-4
Cable
SMA-SMA
Balun
ZFSCJ-1-2
Balun
ZFSCJ-1-2
Cable SMA-SMA
Cable SMA-SMA
DC-Block
DC-Block
IN+
IN-
CLK+ CLK-
ADC
AT84AS001
e2
Cable SMA-SMA
Cable SMA-SMA
DC-Block
DC-Block
Cable SMA-SMA
Cable SMA-SMA CLK+
CLK-
Bias-Tee
Bias-Tee
Fil o
250
Fil o
250
Adap.
SM-SF
Adap.
SM-SF
Cable
SMA-SMA
Cable
SMA-SMA
Cable SMA-SMA
Cable SMA-SMA
OUT+
OUT-
OUT
D11-D0
12
Logic
Analyze
16760-A
Agilen
Fig. 6 A/D con e e sys em implemen ed
Fig. 7 Measu emen o he e ec o o e lapping noise (ENOB s. he
di e en Subsampling Op imal F equencies)
Ano he in e es ing measu emen is he in luence o he
ji e , which is mo e c i ical a highe equencies and hence
he main limi a ion in he implemen ed sys em. This e ec is
illus a edin Figu e 8, using he op imal subsampling
equency immedia ely less han 500 MHz.
Fig. 8 Ji e e ec measu emen (ENOB s. Ca ie equency)
The o al esolu ion o his sys em is illus a ed in Figu e 9,
using he op imal subsampling equency immedia ely lowe
han 500 Mhz. The inal esul is an A/D con e e sys em ha
con e s signals up o 3,1 GHz wi h mo e han 8 e ec i e bi s.
200
250
300
350
400
450
500
0
1
2
3
4
5
6
7
8
9
s (MHz)
enob
1000
1500
2000
2500
3000
3500
8.2
8.3
8.4
8.5
8.6
8.7
8.8
8.9
c (MHz)
enob
Fig. 9 ENOB o he implemen ed sys em s. Caa ie F equency
As an example, he ou pu spec um is illus a ed in Figu e
10, o a 3 GHz inpu equency and a 480,2 MHz op imal
subsampling equency.
103104105106107108109
-20
0
20
40
60
80
100
120
Hz
dBm
BW=20MHz
Fig. 10 Ou pu Spec um o a 3 GHz Analog Inpu Signal
V. CONCLUSIONS
In he p esen documen we ha e discussed he cu en
limi a ions o he esolu ion o A/D con e e s, which is an
obs acle o he use o he digi al signal p ocesso di ec ly in
RF. In his pape an Analog- o-Digi al Con e e Sys em
using echniques based on subsampling has been p esen ed.
We ha e in oduced he subsampling concep and i s main
non-ideali ies, like ji e o o e lapping he mal noise. The
heo e ical s udy has been used o design a sys em ha
demons a es expe imen ally ha he pe o mance o
subsampling based sys em.
Expe imen al ea u es o he p oposed ecei e sys em ( o
signals wi h a 20 MHz bandwid h) a e:
 ENOB is a ound 9 bi s o 1-2 GHz inpu equency.
 ENOB is mo e han 8 bi s up o 3,1 GHz inpu equency.
The e o e, he expe imen al esul s place he subsampling
based ecei e as an al e na i e o he ypical ecei e
a chi ec u es wi h an enhanced econ igu abili y and
p og ammabili y.
ACKNOWLEDGMENTS
This wo k has been de eloped wi hin he scope o he
TelMAX P ojec and is pa ially unded by CDTI –Cen o
pa a el Desa ollo Tecnológico e Indus ial-, o he Spanish
Minis y o Science and Inno a ion, unde he INGENIO 2010
P og am / CENIT call.
REFERENCES
[1] R. G. Lyons, Unde s anding Digi al Signal P ocessing, Uni ed S a es:
P en ice Hall, 2001.
[2] M. A. I. Mos a a, S. Embabi, M. C. Fe nando and W. C. Chan, Ch.
Go e JR, ―Subsampling RF Recei e ‖ U.S. Pa en 01811614, Dec. 5,
2002.
[3] R. G. Vaughan, N. L. Sco and D. R. Whi e ―The Theo y o Bandpass
Smpling‖ IEEE T ansac ions on Signal P ocessing., ol. 39, pp. 1973-
1984, Sep. 1991.
[4] Y. R. Sun, ―Gene alized Bandpass Sampling Recei e s o So wa e
Radio‖ Doc o al Disse a ion, Royal Ins i u e o Technology, School o
In o ma ion and Communica ion Technology (ICT), S ockholm,
Sweden, 2006
[5] M. Neg ei os, E. Schule , L. Ca o and A. A. Susin, ―Tes ing RF
Signal Pa hs Using Spec al Analysis and Subsampling‖ in P oc.
SBCCI, 2003.
[6] Y. Vande pe en, W. Dehaene and G. Leus ―A Flexible Low Powe
Subsampling UWB Recei e Based on Line Spec um Es ima ion
Me hods‖ Communica ions, 2006 IEEE In e na ional Con e ence on
Volume 10, Page(s):4694 – 4699, June 2006
[7] J. Ca , ―Clocking High-Speed A/D Con e e ‖, Maxim Applica ion
No e 1558, Jan. 2007.
[8] ―Design a Low-Ji e Clock o High-Speed Da a con e e s‖, Na ional
Semiconduc o Applica ion No e 1558, No . 2001.
[9] R. S ephens, ―The Rules o Ji e Analysis‖, Agilen Technologies
Applica ion No e.
[10] S. Ka onen, ―Cha ge-Domain Sampling o High F equency Signals
wi h Embedded Fil e ing‖ hesis, Facul y o Technology, Depa men
o Elec ical and In o ma ion Enginee ing, Uni e si y o Oulu, Finland,
Jan. 2006.
[11] ―1821TH 18 GHz Bandwid h 2GS/s THA da a shee ‖, Inphi, Wes lake
Village, Cali o nia, Uni ed S a es.
[12] ―12-bi 500 Msps ADC AT84AS001‖, E2V, Sain Eg è e Cedex,
F ance.
500
1000
1500
2000
2500
3000
3500
0
1
2
3
4
5
6
7
8
9
10
Fin (MHz)
enob
264 Appendix C: Publica ions
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
“Mul iple Clocking High Analog- o-Digi al Con e sion Based on
Subsampling”, XXVI Con e ence o Design o Ci cui s and In eg a ed
Sys ems (DCIS’2011), 2011.
405
Mul iple Clocking High Analog- o-Digi al Con e sion based on Subsampling
José Ramón Ga cía Oya1, An onio Ju ado Díez2, Fe nando Muñoz Cha e o1, An onio To alba Silgado1,
Fe nando J. Má quez Lasso 1, En ique López-Mo illo 1
1 Depa amen o de Ingenie ía Elec ónica, Uni e sidad de Se illa c/ Camino de los Descub imien os s/n 41092 Se illa, Spain
oya@g e.esi.us.es | munoz@g e.esi.us.es | o [email p o ec ed] | e nando.ma [email p o ec ed] | [email p o ec ed]
2 AT4 wi eless, c/ Se e o Ochoa 2 29590, Málaga, Spain
ajdiez@a 4wi eless.com
0XOWLSOH&ORFNLQJ+LJK$QDORJWR'LJLWDO&RQYHUVLRQ
EDVHGRQ6XEVDPSOLQJ
-RVp5DPyQ*DUF D2 D$QWRQLR-XUDGR' H])HUQDQGR0XxR]&KDYHUR$QWRQLR7RUUDOED6LOJDGR)HUQDQGR-
0iUTXH]/DVVR(QULTXH/ySH]0RULOOR
'HSDUWDPHQWRGH,QJHQLHU D(OHFWUyQLFD8QLYHUVLGDGGH6HYLOOD
F&DPLQRGHORV'HVFXEULPLHQWRVVQ6HYLOOD6SDLQ
R D#JW H HVLXVHVIPXQR]#JW H HVLXVHVWRUUDOED#XVHVIHUQDQGR PD UTXH]#JL H HVLXVHV
HQULTXH ORSH]#JL H HVLXVHV 
$7ZLUHOHVV
F6HYHUR2FKRD0iODJD6SDLQ
DM GL H]#DW ZLUHO HVV FRP

$EVWUDFW²7KLV SDSHU SUHVHQWVD'DWD $FTXLVLWLRQ 6 VWHPIRU
WHVWLQJ ZLGHEDQGPXOWLVWDQGDUG UHFHLYHUVXVLQJWHFKQLTXHV
EDVHG RQVXEVDPSOLQJ :LWKWKH6RIWZDUH 'HILQHG 5DGLR
SDUDGLJPLQPLQGWKHXOWLPDWHJRDOLVWKHGHVLJQRIDQ$QDORJ
WR'LJLWDOFRQYHUVLRQV VWHPZKLFKFRQILJXUHGE VRIWZDUHDWWKH
5DGLR )UHTXHQFLHV FDQ EHDSDUW RIORZFRVW FHUWLILFDWLRQ
HTXLSPHQW,WVPDLQ REMHFWLYH LVWR LPSURYH WKH IHDWXUHVRI D
SUHYLRXVO SUHVHQWHGGDWD DFTXLVLWLRQ PRGXOH WKDWXVHV D
FRPPRQFORFNUHIHUHQFH 7KLVSURSRVHG V VWHPLV EDVHG RQ
FRPPHUFLDO GHYLFHVPDLQO  D /RZ -LWWHU DQG :LGHEDQG
6DPSOH+ROG DQG D +LJK5HVROXWLRQ ,QWHUPHGLDWHIUHTXHQF 
$QDORJWR'LJLWDO&RQYHUWHU

.H ZRUGV²$QDORJWR'LJLWDO&RQYHUWHU(12%-LWWHU
6DPSOH+ROG6RIWZDUH'HILQHG5DGLR61'57KHUPDO1RLVH
,,1752'8&7,21
1RZDGD VGXHWRWKHDFFHSWDQFHRIZLUHOHVVWHFKQRORJLHV
LQWKHFRQVXPHUPDUNHWPRELOHSKRQHVSHUVRQDOFRPSXWHUV
79VHWV HWF WKHUH LVDODUJHQXPEHU RIGLIIHUHQW
FRPPXQLFDWLRQVWDQGDUGV$VDFRQVHTXHQFHWKHUHLVDWUHQG
WRGHVLJQ PXOWLVWDQGDUGUHFHLYHUV>@SUHIHUDEO  ORZ FRVW
XVLQJRQO DIHZFRPSRQHQWV
,QRUGHUWR JHW WKH SDUDGLJPRI WKH 6RIWZDUH'HILQHG
5DGLR 6'5 WRLQFUHDVH WKH UHVROXWLRQ DQG VSHHG RIWKH
$QDORJWR'LJLWDO&RQYHUVLRQVWDJHLWLVFUXFLDOWREHDEOHWR
SODFHLWFORVHUWKHDQWHQQD>@7KXVWKHDQDORJIURQWHQGLV
JUHDWO  VLPSOLILHGDQGWKHIOH[LELOLW  RIWKHUHFHLYHULV
LPSURYHGGXHWRWKHIDFWWKDWPRUHDQDORJIXQFWLRQDOLWLHVHJ
ILOWHUVPL[HUVDUHSHUIRUPHGLQWKHGLJLWDOGRPDLQ
6XEVDPSOLQJ UHFHLYHUV XVH D 6DPSOH+ROG6+WR
SURGXFHORZIUHTXHQF UHSOLFDV RIWKH5) VLJQDO EDVHG RQ
SDVVEDQGVDPSOLQJWHFKQLTXH7KHUHIRUHWKHPDLQDGYDQWDJH
RIWKHV VWHPEDVHGRQVXEVDPSOLQJLVLWVIOH[LELOLW EHFDXVH
LWLVSRVVLEOHWRSURFHVVPRVWRIWKHVLJQDOGLJLWDOO $OVRWKH
QXPEHURIDQDORJEXLOGLQJEORFNVLVORZHUDQGZHFDQXVHDQ
$QDORJWR'LJLWDO&RQYHUWHU$'&ZKRVHVSHFLILFDWLRQVDUH
PRUHUHOD[HG IRU D YHU ZLGH LQSXW IUHTXHQF  EDQG 2QWKH
RWKHU KDQG VRPHRI WKH PDLQGLVDGYDQWDJHVRI WKH V VWHP
EDVHG RQVXEVDPSOLQJDUHWKHGHPDQGLQJ VSHFLILFDWLRQV
UHTXLUHGIRUWKH6+ZLGHLQSXWEDQGZLGWKDQGORZDSHUWXUH
MLWWHUDQGWKHKLJKQRLVHILJXUHGXHWRWKHHIIHFWRIWKHIROGHG
WKHUPDOQRLVHLQWKHEDQGRILQWHUHVW
7KHFXUUHQWVWDWHRIDUWRIKLJKIUHTXHQF $'&VGRHVQRW
DOORZ WKH XVH RIWKH$'&GLUHFWO  LQWKH5) GRPDLQ ,QD
SUHYLRXVZRUN >@ ZHSUHVHQWHGDQ $QDORJWR'LJLWDO
&RQYHUWHUV VWHP EDVHG RQVXEVDPSOLQJ WKDWLPSURYHG WKH
FXUUHQWVSHFLILFDWLRQVIRUWKHFRQYHQWLRQDO FRQYHUVLRQ
ELWVZLWK  *6V$'&V RQ WKH KHDGHUUHFHLYHU REWDLQLQJD
UHVROXWLRQKLJKHUWKDQELWVDQGZLWKDPD[LPXPRI*+]
RIFHQWHUIUHTXHQF RIWKH5)VLJQDOV
,QWKHODVW HDUVWKHVSHFLILFDWLRQVRIWKHVWDWHRIDUWKDYH
EHHQLPSURYHGPDNLQJ LW SRVVLEOH WR REWDLQDUHVROXWLRQ
DURXQGELWVIRU*6V$'&V>@
7KHUHIRUHZH SURSRVHGWR LPSURYH WKH UHVROXWLRQ RIRXU
V VWHPXVLQJ WZR GLIIHUHQW RSWLPDO VDPSOLQJIUHTXHQFLHV LQ
RUGHUWRUHGXFHWKHHIIHFWRIWKHIROGHGWKHUPDOQRLVHDQGWKXV
LQFUHDVLQJWKHWRWDOUHVROXWLRQ
7KHEDVLFVSHFLILFDWLRQVRIWKHLPSOHPHQWHGV VWHPLQWKLV
SDSHUDUHLOOXVWUDWHGLQ7DEOH,
7$%/(,
63(&,),&$7,2162)7+(,03/(0(17('6<67(0
$QDORJ,QSXW)UHTXHQF '&*+]
6L
J
QDO%DQGZLGWK 0+]
6DP
S
OLQ
J
)UH
T
XHQF
6+*+]
6DP
S
OLQ
J
)UH
T
XHQF
$'&0+]
5HVROXWLRQ!ELWV

:LWKDVLJQDO EDQGZLGWK RI 0+]WKHSURSRVHG $'&
V VWHPDFKLHYHV EHWZHHQDQGELWVRI UHVROXWLRQIRUD
SURJUDPPDEOHFDUULHUIUHTXHQF UDQJLQJIURPXSWR*+]
% SURSHU VHOHFWLRQ RIWKHFHQWHUIUHTXHQF DQGVLJQDO
EDQGZLGWKWKLVGDWDDFTXLVLWLRQV VWHPFDQEHXVHGLQPRVWRI
SUHVHQWZLUHOHVVVWDQGDUGV
7KLVSDSHULV RUJDQL]HGDV IROORZV VHFWLRQ ,, UHYLHZV
VXEVDPSOLQJ DQG WKHRUHWLFDO FRQFHSWVRQ KRZ WR REWDLQ WKH
RSWLPDOVXEVDPSOLQJ IUHTXHQF DQGWRHYDOXDWHHIIHFWVRIWKH
PDLQQRQLGHDOLWLHVWKDW ZLOOOLPLW WKH LPSOHPHQWHGV VWHP
7KHVHQRQLGHDOLWLHVKHOSXVWRMXVWLI  WKH FKRLFHRI WKHQHZ
VDPSOLQJIUHTXHQF SODQLQWKLVVHFWLRQDQGWKHFKRLFHRIWKH
FRPSRQHQWVLQVHFWLRQ,,,,QVHFWLRQ,9H[SHULPHQWDOLVVXHV
SUHVHQW WKH DWWDLQHG UHVXOWV DERXW QRLVH DQG GLVWRUWLRQ7KH
SDSHUILQLVKHVZLWKFRQFOXVLRQVLQVHFWLRQ9

Fig. 1. Powe spec um a he inpu ( op) and he ou pu (bo om) o a
nonlinea sys em
The uni e sal o mula o ind alid sampling equencies in
he p esence o ha monics is gi en by [3]:
sksk ni j ni )1(
111 ++<≤+ (4)
Being i 1 and j 1 wo ha monics o 1 and nk = loo ((j 1-i 1)/ s).
An algo i hm o ind he ange o alid subsampling
equencies o mul iband sys ems is p esen ed in [4]. F om
he gene al equa ions ob ained in [4], and conside ing he
pa icula case o dual band sys em, he maximum eplica
o de o he lowe band (n1) mee s he ollowing equa ion:
⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛
−+−
≤
⎟
⎟
⎠
⎞
⎜
⎜
⎝
⎛
=
))()((2 2211
11
1
LULU
L
s
L
loo
loo n (5)
Whe e L1 and U1 a e he low and he high limi s o he
lowe band and L2 and U2 a e he low and he high limi s o
he uppe band. Knowing 2= R1 1, eplica o de s o he uppe
band (n1) mee he ollowing cons ain :
)()( 111211 RnR loo nnR loo +≤≤ (6)
The eigh possible anges o dual band applica ions a e
lis ed in [4]. Thus, he inal sampling anges will be gi en he
ollowing exp ession:
hmdcmdimddb FFFFF ∩∩∩= (7)
Whe e F is he in e sec ion o all he alid anges calcula ed
om (4) and (5), Fdb, Fimd, Fcmd and Fhmd a e he alid
sampling equency se s o he undamen al signals,
in e modula ion, c oss modula ion and ha monic dis o ion,
espec i ely.
In o de o ind F an algo i hm has been de eloped and
w i en in MATLAB sc ip . This algo i hm calcula es hese
anges and he loca ion o he eplicas whe e he inpu
pa ame e s a e he undamen al equencies, he numbe o
ha monics and he signal bandwid h. As an example, Table I
shows he h ee i s alid anges immedia ely lowe han 2
GHz o he undamen als signals a 1.82 and 2.4 GHz,
conside ing i e ha monics and a signal bandwid h equal o 25
MHz. The subsampled spec um is illus a ed in Fig. 2 o a
sampling equency equal o 2 GHz, showing ha he e is no
o e lapping be ween signals.
TABLE I
VALID SAMPLING FREQUENCIES BELOW 2 GHZ
Lowe F equency Bound
(MHz)
Uppe F equency Bound
(MHz)
1995 2000
1837.5 1978.33
1801.67 1802.5
0100 200 300 400 500 600 700 800 900 1000
F equency (MHz)
1 - 1820.00 MHz @ 180.00 MHz
2 - 2400.00 MHz @ 400.00 MHz
(-1 1 + 1 2) - 580.00 MHz @ 580.00 MHz
(3 1 + -2 2) - 660.00 MHz @ 660.00 MHz
(-2 1 + 2 2) - 1160.00 MHz @ 840.00 MHz
(2 1 + -1 2) - 1240.00 MHz @ 760.00 MHz
(-1 1 + 2 2) - 2980.00 MHz @ 980.00 MHz
(3 1 + -1 2) - 3060.00 MHz @ 940.00 MHz
(-2 1 + 3 2) - 3560.00 MHz @ 440.00 MHz
(2 1 + 0 2) - 3640.00 MHz @ 360.00 MHz
Fig. 2. Subsampled spec um o 1.82 and 2.4 GHz inpu equency
IV. OPTIMIZATION OF THE RECEIVER ARCHITECTURE
In o de o op imize he p oposed ecei e , se en inpu
equencies ha e been selec ed o s udy he selec i e
combina ions o di e en dual band applica ions. These
chosen s anda ds a e WCDMA (V) a 880 MHz, GSM-DCS a
1.82 GHz, WCDMA (I) a 2.12 GHz, Blue oo h a 2.4 GHz,
WiMAX a 3.5 and 5.8 GHz, and 802.11a a 5.2 GHz.
Since he main ocus is o co e he maximum numbe o
s anda ds, i is manda o y o use a S&H be o e he ADC in
o de o ha e enough analog bandwid h. The S&H om Inphi
wi h pa numbe 1821TH has been selec ed o his wo k,
because o i s high inpu analog bandwid h (up o 18 GHz),
minimum ape u e ji e (50 s) and a maximum clock
equency equal o 2 GHz.
The i s s udied scena io is based on high esolu ion ADC
wi h a high sampling equency o educe he olded noise
e ec . Wi h his ocus in mind he selec ed ADC was a 12-bi
ADS5400 om Texas Ins umen s wi h maximum clock
equency o 1 GHz. Using a sampling equency o almos 1
GHz, i is possible o co e all he dual band applica ions, as
illus a ed in Fig. 3 (Case 1), whe e he meaning o axis x is
illus a ed in Table II. Using as e e ence a ypical SNR o he
ADC equal o 58 dB, he heo e ical SNR o each dual band
applica ion was calcula ed om (2) and (3).
Ano he op ion is o use a highe esolu ion ADC, like he
14-bi ADS5474 om Texas Ins umen s (Case 2 in Fig. 3).
This de ice was selec ed because i s maximum sampling
equency is 400 MHz and, he e o e, he olded noise would
only be a ound 4 dB highe han Case 1.
TABLE II
DUAL BAND APPLICATIONS AND AXIS X CORRESPONDENCE
X axis 1 2 3 4 5 6
Inpu
F eq.
(GHz)
0.88-
1.82
0.88-
2.12
0.88-
2.4
0.88-
3.5
0.88-
5.2
0.88-
5.8
X axis 7 8 9 10 11 12
Inpu
F eq.
(GHz)
1.82
-2.12
1.82-
2.4
0.88-
3.5
1.82-
5.2
1.82-
5.8
2.12-
2.4

2 4 6 8 10 12
40
45
50
55
60
65
SNR (dB)
Case 1: @ 1GHz
Case 2: @ 400 MHz
Case 3: @ 2GHz
1GHz
Case 4: @ 2GHz
400MHz
Fig. 3. Expec ed SNR o single and mul iple clock a chi ec u es
Howe e , as shown in Fig. 3, his op ion is less lexible,
because i is no possible o ind any sampling equency
lowe han 400 MHz o he i s h ee scena ios. In o de o
imp o e he SNR wi hou losing lexibili y, wo s eps
subsampling app oach is p oposed, whe e he sampling
equency o S&H was se a a ound 2 GHz and he sampling
equency o ADC a a ound 1 GHz (Case 3 in Fig. 3).
Al hough his a chi ec u e imp o es he SNR by
app oxima ely 3 dB om (3) in espec o Case 1, i could be
necessa y o implemen a second subsampling p ocess and,
he e o e, a new olded noise e ec will be added.
The las op ion is o use a mul iple clock a chi ec u e
employing he ADS5474 (Case 4 in Fig. 3) and a i s
sampling equency a ound 2 GHz. Theo e ically he SNR is
imp o ed a ound 3 dB in espec o Case 3. In his case, due o
he second subsampling p ocess, olded noise e ec s mus be
added as well.
Fo he es o combina ions o equencies he cu es
p esen he same endency, being possible o co e all he
scena ios. Howe e , since cases 2 and 4 p esen he bes
esul s abou SNR he nex s ep will be o co e all he dual
band applica ions o hese cases. The p oposed solu ion is o
use a bank o band-pass il e s be ween he S&H and he ADC.
This solu ion will be applied o Case 4, because i has mo e
lexible a chi ec u e, wi h a highe numbe o a ailable alid
anges. Using his solu ion, some ha monics will be emo ed
and he lexibili y o he ecei e will be inc eased.
The solu ion is based on wo il e s, whose band-pass anges
a e [0-400] and [400-800] MHz. The maximum sampling
equency was selec ed in o de o ha e bo h undamen al
eplicas in each ange. The selec ed il e co esponds o he
highe o hese wo equencies (Case 5 in Fig. 4).
Ano he solu ion is o ix a unique BP il e o all he
applica ions (Case 6 and 7 in Fig. 4). In hese cases i is
possible o co e almos all he s anda ds wi h only one o
hese il e s, wi hou conside ably educing he esolu ion.
Al hough in o de o maximize he lexibili y and he SNR, he
op imal a chi ec u e is like he one illus a ed in Fig. 5, o a
mo e conc e e applica ion o mo e elaxed SNR speci ica ions
a single BP il e could be used in o de o educe he
complexi y o he sys em.
0 2 4 6 8 10 12
48
50
52
54
56
58
60
62
64
66
SNR (dB)
Case 5: en elope o Cases 6 and 7
Case 6: il e in [0-400] MHz
Case 7: il e in [400-800] MHz
Case 4: @ 2 GHz & 400MHz wi hou il e s
Case 2: @ 400MHz wi hou il e s
Fig. 4. Expec ed SNR o di e en a chi ec u es based on BP il e s
Fig. 5. Op imized a chi ec u e based on mul iple clock and BP il e s
V. CONCLUSION
In his pape an op imiza ion o a dual band mul i-s anda d
ecei e based on subsampling in a non linea en i onmen
has been p esen ed. Subsampling echniques ha e been
selec ed due o he simplici y ha hese sys ems p esen in
compa ison o adi ional ecei e s. Howe e , ecei e s based
on subsampling ha e addi ional sou ces o noise whose
minimiza ion has been he main ocus o his wo k om he
s udy o di e en a chi ec u es. Mo eo e , dual band sys ems
in nonlinea scena ios ha e an addi ional p oblem because o
he o e lapping o ha monics. An e icien algo i hm has been
de eloped in o de o ind he alid sampling equencies,
inc easing he lexibili y o he ecei e and co e ing he
maximum numbe o dual band applica ions o di e en
communica ion s anda ds.
REFERENCES
[1] R. Vaughan, N. Sco , and D. Whi e, “The Theo y o Bandpass
Sampling,” IEEE T ansac ions on Signal P ocessing, ol. 39,
no. 9, pp. 1973-1984, Sep embe 1991.
[2] J. Mi ola, “The So wa e Radio A chi ec u e,” IEEE
Communica ions Magazine, ol. 33, no. 5, pp. 26-38, May 1995.
[3] C. H. Tseng, “A Uni e sal Fo mula o he Comple e Bandpass
Sampling Requi emen s o Non Linea Sys ems,” IEEE
T ansac ions on Signal P ocessing, ol. 57, no. 10, pp. 3869-
3878, Oc obe 2009.
[4] C. H. Tseng, and S. C. Chou, “Di ec Downcon e sion o
Mul iband RF Signals Using Bandpass Sampling,” IEEE
T ansac ions on Wi eless Communica ions, ol. 5, no. 1, pp. 72-
76, Janua y 2009.
[5] M. B. Dadi, and R. Bouallegue, “On he RF Subsampling
Con inuous-Time ΣΔ Downcon e sion S age o Mul is anda d
Recei e s,” In e na ional Con e ence on Compu e Enginee ing
and Technology (ICCET), ol. 6, pp. 167-171, June 2010.
274 Appendix C: Publica ions
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
9.5.4 Publica ions pa ially ela ed wi h his hesis
“An 8-bi 19 MS/s low-powe 0.35 µm CMOS pipelined ADC o
DVB-H”, In eg a ion, he VLSI Jou nal, 2012
An 8-bi 19 MS/s low-powe 0.35 mm CMOS pipelined ADC o DVB-H
B. Palomo
n
, F. Mun
˜oz, R.G. Ca ajal, J.R. Ga cia, F. Ma quez
Depa men o Elec onic Enginee ing, Uni e si y o Se ille, Spain
a icle in o
A icle his o y:
Recei ed 6 May 2011
Recei ed in e ised o m
25 Oc obe 2011
Accep ed 26 Oc obe 2011
A ailable online 4 No embe 2011
Keywo ds:
Pipelined ADC
CMOS analog in eg a ed ci cui s
Low powe
Low ol age
Opamp-sha ing
abs ac
This pape p oposes an 8b 19 MHz CMOS pipelined analog- o-digi al con e e (ADC) o DVB-H.
In o de o educe he powe consump ion a combina ion o echniques has been used, such as op-amp
sha ing, low-powe amplifie s wi h gain boos ing and an agg essi e capaci o scaling. The p o o ype
ADC ab ica ed in 0.35 mm CMOS demons a es a maximum di e en ial nonlinea i y (DNL) o 0.63 leas
significan bi (LSB) and a maximum in eg al nonlinea i y (INL) o 0.58 LSB wi h a peak signal- o-noise-
and-dis o ion a io (SNDR) and spu ious- ee dynamic ange (SFDR) o 42.76 and 51.57 dB a 19 MHz.
The ADC wi h an ac i e a ea o 4.78 mm
2
consumes less han 4 mW a he men ioned sampling
equency.
&2011 Else ie B.V. All igh s ese ed.
1. In oduc ion
Du ing he las ew yea s much e o has been de o ed
owa ds he educ ion o he supply powe o mixed signal CMOS
sys ems. This is p ima ily due o he inc easing impo ance o
ba e y-powe ed elec onics, and he con inued down-scaling o
de ice sizes. Pipelining has been accep ed as one o he bes
app oaches o implemen high-speed medium- o-high esolu ion
analog- o-digi al con e e s wi h minimum powe consump ion.
Digi al ideo b oadcas ing (DVB) sys em becomes e y a ac-
i e o applica ions in wi eless mobile communica ion de ices,
such as lap op compu e s, mobile phone and ehicles [1].
Recen ly, digi al ideo b oadcas ing-handheld (DVB-H) has made
i possible o deli e b oadcas ele ision o o he mul imedia
se ices o a mobile o handheld de ice [2]
The block diag am o a DVB analog on -end is shown in Fig. 1.
The une selec s he channel con e ing he OFDM RF signal o a
fi s in e media e equency a ound 35 MHz a e which i is
bandpass fil e ed by a SAW-fil e s age. The SAW fil e is ollowed
by a con ollable gain amplifie (AGC) in o de o adap he signal
le el. Finally he esul ing signal is con e ed in o a digi al signal
using an ADC.
The p oposed ecei e implemen s a subsampling echnique as
his is he mos e ficien solu ion om a powe consump ion
poin o iew. This echnique pe o ms, a he same ime, he
mixing and sampling p ocess, aking ad an age o he band
olding inhe en o he sampling p ocess.
Fo an in e media e equency alue o 34 MHz and he
maximum signal bandwid h defined in he DVB s anda d
(8 MHz), he op imum alue o he sampling equency is nex
o 19 MHz. A ending o he Mobile and Po able DVB-T Radio
Access In e ace (MBRAI) om EICTA, he SNR o he demodula-
ion p ocess should be 27 dB in he wo s case. An 8 bi ADC
achie es he specifica ion, including a secu i y ma gin in o de o
an icipa e in e e ing componen s influence.
This pape is o ganized as ollows: he pipelined ADC 1.5-bi
pe s age a chi ec u e is shown in Sec ion 2.Sec ion 3 enume a es
and de ails he low powe echniques applied o he ADC in o de
o achie e such a low consump ion. Sec ion 4 desc ibes he ci cui
implemen a ion and he measu emen esul s. The pape is
concluded in Sec ion 5.
2. ADC A chi ec u e
A 1.5-bi -pe -s age a chi ec u e has been used in he pipelined
ADC because i shows bo h he lowe powe consump ion and
smalles a ea compa ed wi h a chi ec u es based on highe
esolu ion s ages. The powe e ficiency o he 1.5 bi configu a ion
[3–5] ely on ha he amplifie s ope a e a a low closed-loop gain
leading o a bes se ling ime o minimum powe consump ion.
A block diag am o he pipeline 1.5-bi /s age a chi ec u e is
shown in Fig. 2. I consis s o a cascade o se en s ages. Each s age
esol es wo bi s wi h a sub-ADC, sub ac s he con e ed alue,
which only can ake alues V
e
,V
e
o 0 (whe e V
e
is di e -
en ial e e ence ol age), om i s inpu s, and amplifies he
Con en s lis s a ailable a SciVe se ScienceDi ec
jou nal homepage: www.else ie .com/loca e/ lsi
INTEGRATION, he VLSI jou nal
0167-9260/$ - see on ma e &2011 Else ie B.V. All igh s ese ed.
doi:10.1016/j. lsi.2011.10.003
n
Co esponding au ho . Tel.: þ34 954487472; ax: þ34 954487373.
E-mail add esses: [email p o ec ed] (B. Palomo),
[email p o ec ed] (F. Mun
˜oz), [email p o ec ed] (R.G. Ca ajal),
[email p o ec ed] (J.R. Ga cia), [email p o ec ed] (F. Ma quez).
INTEGRATION, he VLSI jou nal 45 (2012) 222–227
276 Appendix C: Publica ions
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
“Compac low-powe implemen a ion o con inuous- ime ΣΔ
modula o s”, In eg a ion, he VLSI Jou nal, 2012.
Compac low-powe implemen a ion o con inuous- ime
SD
modula o s
E. Lo
´pez-Mo illo
a
, F. Mun
˜oz
a,
n
, A. To alba
a
,F.Ma
´ quez
a
, I. Rebollo
b
, J.R. Ga cı
´a-Oya
a
a
Elec onic Enginee ing Depa men , Escuela Supe io de Ingenie os, Uni e si y o Se illa, Camino de los Descub imien os s/n, 41092 Se illa, Spain
b
Fa sens S.L., Pa que Tecnolo
´gico de San Sebas ia
´n, Paseo Mikele egi 54, Plan a 0—Oficina 1, 20009 San Sebas ia
´n, Spain
a icle in o
A icle his o y:
Recei ed 11 Ap il 2012
Recei ed in e ised o m
27 Sep embe 2012
Accep ed 2 Oc obe 2012
Keywo ds:
Analog-digi al con e sion
Sigma-del a modula ion
Low powe
abs ac
This pape p esen s a low-a ea con inuous ime (CT) sigma–del a (
SD
) modula o implemen a ion
based on a local eedback. The p oposed s uc u e p o ides a e y low impedance node wi hou he
need o classical op-amps, which leads o a educ ion in powe and a ea consump ion. Two e sions o a
con en ional fi s -o de CT
SD
modula o p o o ype ha e been ab ica ed wi h he pu pose o
e alua ing he idea. The modula o equi emen s ha e been se o a passi e RFID ag wi h sensing
capabili y applica ion, so ha achie ing minimum ac i e a ea and e y low powe consump ion a e he
main objec i es o he p esen ed design. Expe imen al esul s o he fi s e sion o he modula o
show 8 bi s o E ec i e-Numbe -O -Bi s (ENOB) in a 25 kHz signal bandwid h wi h 7 mW o powe
consump ion. The p oposed implemen a ion has also shown o be e y obus agains supply ol age
and bias cu en a ia ions. A second app oach has also been designed, using he same p inciple o
ope a ion, in o de o inc ease he inpu ol age ange wi hou any powe consump ion penal y a he
expense o dec easing he inpu impedance and s ingily inc eased a ea. This second app oach shows
9 bi s o ENOB in he same signal bandwid h wi h a powe consump ion o 4.35 mW. A Figu e O
Me i (FOM) o 0.267 pJ/s a e has been achie ed wi h a o al a ea consump ion (wi hou pads) o
110 mm125 mm in a 0.35 mm CMOS echnology.
&2012 Else ie B.V. All igh s ese ed.
1. In oduc ion
RFID (Radio F equency Iden ifica ion) sys ems ha e been
adi ionally used o iden ifica ion and acking applica ions,
eplacing he classic ba codes in se e al applica ions such as
supply chain managemen , in en o y con ol in wa ehousing,
ai po baggage con ol and manu ac u ing.
The RFID sys em is made up o wo main blocks called
ansponde ( ag), no mally embedded in a label, and he eade .
RFID ags can ei he be passi e o ac i e. An ac i e ag akes he
ene gy om a ba e y, so ha i can ansmi s longe dis ances
and uses mo e sophis ica ed signal p ocessing. A passi e ag
sca enges he ene gy om he elec omagne ic field emi ed by
he eade . As i does no need any ba e y i can be smalle and
cheape han an ac i e ag and wi h unlimi ed li e ime. In e es
o he passi e applica ions is g owing due o he high ab ica ion
and main enance cos s o ac i e ags.
Combining senso s wi h passi e RFID ags opens he way o
new applica ions o RFID in consume elec onics, au omo i e,
medicine and heal hca e. As he passi e RFID senso nodes a e
powe ed by ene gy sca enging, ul a-low powe consump ion
and obus ness agains p ocess a ia ion and changes in he
supply ol age a e essen ial equi emen s. In addi ion, as ypical
in mass p oduc ion applica ions, low a ea consump ion is c ucial
in o de o dec ease he ab ica ion cos .
In he design o passi e RFID ags wi h sensing capabili y, mos
o he epo ed wo ks a e ocused on an enna and RF on -end
design. Howe e , much e o is s ill needed on he senso in e -
ace, in which he ADC (Analog- o-Digi al Con e e ) is a c ucial
componen . The design o ADCs o passi e RFID sys ems is a
cu en challenge o he IC design esea ch communi y as i mus
combine low powe consump ion, small a ea and obus ness
agains powe supply a ia ions.
This pape p esen s a compac ADC implemen a ion which
con e s he signal coming om a MEMS (Mic o Elec o–
Mechanical Sys em) accele ome e , which has a high po en ial
o a a ie y o applica ions in mobile phones, lap ops, game
consoles and handheld de ices. The accele ome e is a single-
ended ou pu signal s uc u e p esen ly a ailable in a 0.35
m
m
CMOS echnology, which is s ill a eliable and cheap echnology
o MEMS. The whole sys em is in ended o be powe ed by an
UHF RFID on -end, which p o ides a 3 V nominal supply ol age
ypical o he selec ed echnology [1]. As he ADC will be
in eg a ed in he same die wi h he MEMS accele ome e , i has
been designed in he same echnology wi h a single-ended inpu .
The basic ADC specifica ions a e summa ized in Table 1.
A esolu ion o 8 bi s is equi ed in a 25 kHz signal bandwid h.
Al hough he nominal supply ol age is 3 V, he ADC mus ha e a
Con en s lis s a ailable a SciVe se ScienceDi ec
jou nal homepage: www.else ie .com/loca e/ lsi
INTEGRATION, he VLSI jou nal
0167-9260/$ - see on ma e &2012 Else ie B.V. All igh s ese ed.
h p://dx.doi.o g/10.1016/j. lsi.2012.10.001
n
Co esponding au ho . Tel.: þ34 954 481 308.
E-mail add ess: [email p o ec ed] (F. Mun
˜oz).
Please ci e his a icle as: E. Lo
´pez-Mo illo, e al., Compac low-powe implemen a ion o con inuous- ime
SD
modula o s,
INTEGRATION, he VLSI jou nal (2012), h p://dx.doi.o g/10.1016/j. lsi.2012.10.001
INTEGRATION, he VLSI jou nal ](]]]])]]]–]]]

278 Appendix C: Publica ions
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
“A No el CMOS Tunable Linea T ansconduc o Based on Quasi
Floa ing Ga e T ansis o s,” XXIII Con e ence o Design o Ci cui s
and In eg a ed Sys ems (DCIS’2008), 2008.
T. Sánchez-Rod íguez1, F. Muñoz1, Jose Ramón Ga cía1, Jose Manuel Rod íguez1,
Ma iano Jimenez-Fuen es1 and R. G. Ca ajal1
1. Escuela Supe io de Ingenie os, Depa amen o de Ingenie ía Elec ónica, Uni e sidad de Se illa,
Camino de los Descub imien os s/n, 41092 Se illa, Spain
Abs ac — A no el CMOS low ol age unable linea
ansconduc o is p esen ed. I is based on he ansconduc o
p esen ed in [1] and [2]. The ansconduc o p esen ed in [2]
employs Floa ing-Ga e T ansis o s a he inpu s age o each
in e e o he a chi ec u e p esen ed by Nau a in [1], imp o ing
i s low ol age ope a ion capabili ies. The app oach p oposed
he e employs quasi- loa ing ga e ansis o s ins ead o loa ing
ga e ansis o s, ob aining u he imp o emen s such as highe
ansconduc ance and ejec ion DC common mode ol age.
Mo eo e , a dynamic biasing echnique has been inco po a ed, in
o de o sol e one o he mayo d awbacks o p e ious
implemen a ions, he sensi i i y o he linea i y o he
ansconduc o o p ocess a ia ions. The p esen ed
ansconduc o achie es e y high speed ope a ion and i is
sui able o high equency con inuous ime il e s. I has been
designed in a 0.5 μm CMOS echnology om 1.4 V powe supply.
Index Te ms—Analog CMOS, Quasi-Floa ing Ga e
T ansis o s, T ansconduc o .
I. INTRODUCTION
HE ma ke in which we a e in ol ed is looking o high-
speed and low-powe ansconduc o ampli ie s o
po able communica ions sys ems [1]-[4]. The educ ion in
powe consump ion and he scaling down o supply ol ages
can be achie ed using speci ic low- ol age echniques as he
one p oposed in [5] and high equency ope a ion can be
ob ained using ansconduc o s wi h a educed numbe o
in e nal nodes (poles) as in [1] and [2].
The main ad an age o he ansconduc o p esen ed by
Nau a in [1] is ha he absence o in e nal nodes leads o e y
high equency ope a ion. Despi e o his a ac i e ea u e,
he e a e some d awbacks which do no allow eliable
p og ammabili y me hods, biasing schemes o o e come
p ocess a ia ions o low ol age ope a ion capabili ies:
The supply ol age mus be la ge han he sum o he
h eshold ol ages o a p-mos and a n-mos ansis o s.
P og ammabili y o he ansconduc ance is achie ed by
modi ying he supply ol age.
I is qui e sensi i e o he inpu common mode ol age
and p ocess a ia ions (any o se a he inpu appea s a
he ou pu and any a ia ion a he alue o he inpu
common mode ol age deg ades he ou pu common
mode ol age dec easing he linea i y o he
ansconduc o ).
Al hough he app oach in [2] allows simple implemen a ion
o p og ammabili y schemes and low ol age ope a ion by
using Mul iple-Inpu Floa ing-Ga e ansis o s, his is a he
cos o a educ ion o he ansconduc ance due o he
capaci i e di ide in he loa ing ga e ansis o e minals, and
he e o e, he maximum achie able wo king equency is
educed. Mo eo e , he sensi i i y o he ansconduc o o
p ocess a ia ions was s ill an open issue.
In his pape a new ansconduc o based on quasi- loa ing
ga e ansis o s is p esen ed. I sol es he p oblem o he
sensi i i y o he ansconduc o o p ocess a ia ions,
p o ides a simple p og ammabili y scheme, and inc eases he
ansconduc ance, while main aining he low ol age ope a ion
capabili ies and he e y high equency ope a ion due o he
absence o in e nal nodes.
In Sec ion II he p oposed ansconduc o a chi ec u e will
be desc ibed and i s ad an ages will be explained. Mo eo e ,
sec ion III will p o ide simula ion esul s o he
ansconduc o ha con i ms he heo e ical assump ions. As i
will be seen in his sec ion, he ansconduc o p esen s
imp o ed pe o mances in e ms o p og ammabili y,
sensi i i y o p ocess a ia ions and low ol age ope a ion
capabili ies while main aining noise, linea i y and powe
consump ion igu es.
The OTA has been laid-ou and sen o ab ica ion in a
s anda d 0.5 μm CMOS echnology, so as expe imen al esul s
will be p o ided du ing he con e ence.
II. PROPOSED TRANSCONDUCTOR
A. Quasi-Floa ing Ga e T ansis o s
Quasi-Floa ing Ga e T ansis o s ha e ecen ly been used o
many analog ci cui s as hey p esen imp o ed pe o mances
o ci cui s in which any possible o se a he inpu ol age
cause deg ada ion o he ci cui ope a ion [5].
As i is epo ed in [6], a quasi loa ing ga e (QFG)
A No el CMOS Tunable Linea T ansconduc o
Based on Quasi Floa ing Ga e T ansis o s
T
280 Appendix C: Publica ions
SUBSAMPLING RECEIVERS WITH APPLICATIONS TO SOFTWARE DEFINED RADIO SYSTEMS
“A Ve y Low Powe 8-Bi 16MSamples/s Pipelined Con e e o
DVB-H,” XXIII Con e ence o Design o Ci cui s and In eg a ed
Sys ems (DCIS’2008), 2008.
A VERY LOW POWER 8-BIT 16MSAMPLES/S
CMOS PIPELINED CONVERTER FOR DVB-H
B. Palomo, F. Muñoz, R.G. Ca ajal, J.R. Ga cía, H. ElGmili and A. To alba
G upo de Ingenie ía Elec ónica, Uni e sidad de Se illa
Abs ac . A 2.5V, 8-bi , 16 MS/s pipeline
analog- o-digi al con e e (ADC) o DVB
applica ion and ba e y powe ed sys ems has
been implemen ed in 0.35 μm CMOS
echnology. In o de o educe he powe
consump ion a combina ion o echniques has
been used, such as op-amp sha ing, low-powe
ampli ie s wi h gain boos ing and an
agg essi e capaci o scaling. The pos -layou
simula ion shows a peak signal- o-noise-and-
dis o ion a io (SNDR) o 48.51 dB,
maximum di e en ial nonlinea i y (DNL) o
0.40 leas signi ican bi (LSB), maximum
in eg al nonlinea i y (INL) o 1.06 LSB, and a
powe consump ion o less han 4 mW.
1. INTRODUCTION
Du ing he las ew yea s much e o has been
de o ed owa ds he educ ion o he supply
powe o mixed signal CMOS sys ems. This is
p ima ily due o he inc easing impo ance o
ba e y-powe ed elec onics, and he con inued
down-scaling o de ice sizes. Pipelining has
been accep ed as he bes app oach o
implemen high-speed medium- o-high
esolu ion analog- o-digi al con e e s wi h
minimum powe consump ion.
2. 1.5-BIT/STAGE STRUCTURE
This 1.5 bi con igu a ion is pa icula ly
sui able o minimize he con e e ’s o al
powe dissipa ion [1]-[3] because he
ampli ie s ope a e a a low closed-loop gain
leading o a bes se ling ime o minimum
powe consump ion.
A block diag am o he pipeline 1.5-bi /s age
a chi ec u e is shown in Fig 1. I consis s o a
cascade o se en s ages. Each s age esol es
wo bi s wi h a sub–ADC, sub ac s he
con e ed alue om i s inpu s, and ampli ies
he esul ing esidue by a gain o wo. The las
s age o he pipeline does no need o gene a e
a esidue and, hen, i does no equi e an op-
amp. The esul ing 14 bi s a e combined wi h
digi al co ec ion o yield eigh bi s a he
ou pu o he ADC. Compa a o s o se up o
±V e /4 can be ole a ed wi hou deg ada ion o
he o e all SNDR using he men ioned
echnique.
S age 1 S age i S age 7
n1 bi s ni bi s n7 bi s
ADC i DAC i
+
Vin (i)
Dou (i)
Vcda (i) Vou (i)
V es(i)2X
S/H
Figu e1. Pipeline-ADC 1.5-bi /s age
a chi ec u e.
A ully di e en ial solu ion has been
implemen ed o maximize he powe supply
ejec ion a io (PSRR) and o minimize e en
ha monic dis o ion. A swi ched-capaci o
implemen a ion was also selec ed, which
ope a es using a non-o e lapping wo-phase
clock.
3. LOW POWER TECHNIQUES
The main con ibu ion o his pape is he
cle e combina ion o di e en powe sa ing
echniques o achie e a e y low powe
solu ion, which will be now desc ibed.
3.1. The sample and hold ampli ie
The sample and hold ampli ie (SHA) a he
inpu o he pipelined con e e usually akes
one hi d o he o e all con e e powe
consump ion [8]. The dedica ed SHA has been
emo ed in ou design and he sampling
ope a ion is pe o med by he swi ched-
capaci o esidue ampli ie (based on a
MDAC) o he i s s age. In his case, special
ca e needs o be paid o he inpu swi ches
in ol ed in he sampling p ocess, and a clock-
boos ing echnique [4] has been implemen ed
o imp o e i s linea i y. In his way a esis ance
independen on he inpu signal is achie ed in
he “on” s a e, as a cons an ol age is applied
ac oss he ga e- o sou ce e minals o he
NMOS ansis o swi ch.