scieee Science in your language
[en] (orig)

Design for embedded testing of an LNA

Abstract

In-circuit testing methodologies are required to tackle the evaluation of embedded radio-frequency circuits. This paper presents design considerations for the test circuitry proposed to implement a methodology for on-chip testing a low-noise amplifier. A previously reported test technique consists on applying to the LNA a sequence of stimuli with different amplitudes, and on measuring the output amplitude for each input level. The obtained set of coordinates (Vin, Vout) allows finding the 3rd order polynomial that best fits the LNAs transfer function. The LNA input voltages which lead to the 1dB compression (P1dB) and third-order intercept (IP3) points are then calculated after the polynomial coefficients. The work presented herein addresses the implementation of this method, according the scheme shown in Figure 1, i. e., the design of a variable amplitude oscillator, of the switch to connect it to the LNA input, as well as of the RMS-DC converter to measure the LNAs output power. Their development is based on criteria seeking to minimize power consumption and simplicity. Another design driving aspect addresses the facility of controlling the test operation and of observing the output measures using digital or low frequency signals, making it easier to interface this test scheme with general purpose testers. In spite of the simplicity of the circuits being proposed, good measurement results can be obtained. This concerns namely the RMS-DC converter, which is based on a simple half-wave rectifier. Anyway, simulation results for 1dB compression and third-order intercept points show a good agreement with the expected ones. Alternative, eventually more accurate, RMS-DC converters or received signal strength indicator circuits would provide more accurate results at the cost of a much higher complexity and power consumption. Being a controlled oscillator available at the LNA input, the blocks placed after the LNA along the receiver chain can also be tested in sequence.

Read accessible full text

Design for embedded testing of an LNA

Author: José Machado da Silva,António Pinho,José Silva Matos
Year: 2005
Source: https://repositorio-aberto.up.pt/bitstream/10216/71523/2/70209.pdf
N
Abs ac —
In-ci cui es ing me hodologies a e equi ed o
ackle he e alua ion o embedded adio- equency ci cui s.
This pape p esen s design conside a ions o he es ci cui y
p oposed o implemen a me hodology o on-chip es ing a
low-noise ampli ie . The scheme being p oposed includes an
oscilla o as he s imulus gene a o , he e-design o he LNA
inpu swi ch o es con ollabili y, and an RMS-DC con e e
o measu e he LNA’s ou pu powe . Simula ion esul s o
1dB comp ession and hi d-o de in e cep poin s show a good
ag eemen wi h he expec ed ones.
Index Te ms — RF, es abili y and es echniques, low noise
ampli ie .
I.
I
NTRODUCTION
ew on-chip es me hodologies a e equi ed o ackle
he di icul ies ound when es ing he new
gene a ion o in eg a ed sys ems comp ising adio-
equency (RF) ci cui s. This is pa icula ly ue a
p oduc ion s ages whe e he es equi emen s a e no as
exhaus i e as hose equi ed a he de eloping phase. Lowe
es equipmen cos is enabled by pa i ioning es esou ces
be ween on-chip and ex e nal equipmen . Howe e , new
highly s uc u ed design o es abili y (D T) app oaches a e
necessa y o imp o e accessibili y o deeply embedded
co es including he RF ones [1].
Di e en me hodologies ha e been p oposed add essing
he es ing o RF on -ends [2 – 7]. These can be di ided in
wo g oups o app oaches:
- in oduc ion o es esou ces wi hin each one o he
anscei e blocks
- ea ing he en i e RF on -end as a single block seen
om he base-band inpu and ou pu elemen s, a e
c ea ing a loop-back signal pa h by connec ing he PA
ou pu in o he LNA inpu .
The echniques in he i s g oup allow diagnosing which
blocks in a RF on -end a e de ec i e, and o p opaga e es
signals wi hou being a ec ed by he o he blocks inse ed
in he signal pa h, bu imply a highe o e head in es
ci cui y.
The wo k p esen ed he ein has been pa ly suppo ed by he Po uguese
go e nmen , Fundação pa a a Ciência e a Tecnologia, and ca ied-ou
unde he amewo k o p ojec s NanoTEST (2A702 - MEDEA+, phase 2)
and Ne wo k o Excellence IST-1-507893-NOE TARGET (Top Ampli ie
Resea ch G oups in a Eu opean Team).
J. Machado da Sil a, Gab iel Pinho, and J. S. Ma os a e wi h
Uni e sidade do Po o, Faculdade de Engenha ia, and INESC Po o,
Campus FEUP, Rª D . Robe o F ias nº 378, 4200-465, Po o, Po ugal
(phone:+351225081400; ax:+351225081443; e-mail: {jms, gab iel,
jsm}@ e.up.p ).
On he o he hand, he ones in he second g oup minimize
he ci cui y o e head bu don’ allow iden i ying which
block(s) a e beha ing inco ec ly. Fu he mo e, some
a chi ec u es and seconda y e ec s, such as an inc ease in
powe consump ion, may impai he use o loop-back
echniques [7].
In [8] echniques o measu e di e en pe o mance
cha ac e iza ion pa ame e s o he low-noise ampli ie
(LNA) we e p esen ed, which ely bo h on he inclusion o
es auxilia y esou ces and on eusing he unc ionali y o
al eady exis ing blocks.
One o hose echniques consis s on applying o he LNA
a sequence o s imuli wi h di e en ampli udes and
measu ing he ou pu ampli ude o each inpu le el. The
ob ained se o coo dina es (V
in
, V
ou
) allows inding he 3
d
o de polynomial ha bes i s he LNA’s ans e unc ion,
i.e., coe icien s a
0
, a
1
, a
2
and a
3
in (1).
...
3
3
2
210
++++=
inininou
aaaa
νννν
(1)
The LNA inpu ol ages which lead o he 1dB
comp ession (P
1dB
) and hi d-o de in e cep (IP
3
) poin s a e
hen calcula ed a e he polynomial coe icien s using,
espec i ely,
3
1
20
1
1
1
3
4
10 a
a
a
A
dB










−=
(2)
and
3
1
3
4
a
a
A
IP
=
(3)
LNA
Mixe
PA
Mixe
T/R
O
LO
Powe
RMS-DC
IF
A
RSSI
S
T
Fig. 1. T anscei e a chi ec u e wi h LNA’s es in as uc u e.
Design o Embedded Tes ing o a LNA
José Machado da Sil a, Gab iel Pinho, José S. Ma os
Acei e pa a ap esen ação na XX Con e ence on Design o Ci cui s and In eg a ed Sys ems, Lisboa, No emb o 2005
To embed his es scheme in he ci cui a ypology such
he one p esen ed in Fig.1 can be used, whe e an ampli ude
con olled oscilla o is equi ed o gene a e he di e en
ampli ude s imuli.
The local oscilla o (LO) used in he up-con e e
(mixe ) and he powe ampli ie (PA) i sel , can be used o
p o ide his unc ionali y, p o ided a loop-back scheme
exis s o connec he PA ou pu o he LNA inpu , and a
scheme o con ol he signal ampli ude is a ailable. In his
case he PA’s beha iou has o be aken in o accoun in he
cha ac e iza ion o he es s imulus, and e en ually he non-
linea i ies o some PA con igu a ions may impai he
iabili y o his solu ion due o he highe di icul y in
ob aining low ampli ude s imuli. This would imply also
highe powe consump ion du ing es ing.
To measu e he LNA’s ou pu powe , loga i hmic
ampli ie s could be used o allow o a wide dynamic
a ia ion o he signal powe o be ep esen ed wi hin a
limi ed ol age ange. The ecei ed signal s eng h indica o
(RSSI) block exis ing in ypical ecei e s’ a chi ec u es
associa ed o he in e media e equency (IF) ampli ie can
also be ( e)used o his pu pose. Bu in his case ei he he
mixe beha iou is also included in he measu e pa h, o a
swi ching mechanism has o be included o bypass i .
Howe e , as he in e media e equency is usually much
smalle han he LNA’s cen e equency, his app oach can
be comp omised.
One o he ad an ages b ough in by he second g oup o
es app oaches desc ibed be o e elies on he es ing
in e ace being made mainly a he baseband equency, and
hus a oids in e e ing and dealing di ec ly wi h high
equency signals. In o de o include a RF es
me hodology wi hin he global es s a egy o he in eg a ed
sys em, using he same es in as uc u e, i is desi able ha
obse a ion and con ol be made a low equencies.
The wo k p esen ed he ein add esses he implemen a ion
o he me hod p esen ed in [8], acco ding he scheme shown
in Fig.1, i. e., he design o a a iable ampli ude oscilla o
(VAO), o he swi ch o connec i o he LNA inpu , as well
as o he RMS-DC con e e o measu e he LNA’s ou pu
powe .
Sec ion II o he pape in oduces he LNA o be es ed.
Sec ions III o V desc ibe he oscilla o , he swi ch, and he
RMS-DC con e e o be included in he ci cui o es ing
pu poses. P elimina y simula ion esul s a e p esen ed in
sec ion VI. Finally concluding ema ks a e highligh ed in
sec ion VII.
II.
T
HE
LNA
Fig. 2 shows he scheme o he LNA o be es ed. I is a
cascode LNA wi h induc i e sou ce degene a ion [9]
designed wi h a 0.18 µm MOS echnology, uned a
435MHz, a equency used in he s anda d Eu opean
spec um o gene al eleme y and elecommand
applica ions.
M
1
M
2
M
3
L
1
L
2
R
2
C
2
C
1
C
3
L
3
R
3
i
o
V
DD
I
p
Fig. 2 – MOS LNA used in he simula ions.
I s nominal cha ac e is ics a e: inpu e lec ion coe icien
S
11
≈ -35 dB, powe gain S
21
≈ 15dB (Fig. 3), inpu 1dB
comp ession poin P
i1dB
= -10 dBm, and hi d-o de
in e cep ion poin IP
3
= 0.18 dBm.
Fig. 3 – S11 and S21 pa ame e s o he LNA.
One o he c i ical aspec s in RF on-chip implemen a ions
conce ns he implemen a ion o spi al induc o s on silicon,
bo h in e ms o he maximum ealisable induc ance alue
and quali y ac o (Q). Values o L om 1nH o 20nH wi h
Q=5 a e common wi h con en ional echnologies [9]. Q
alues up o abou 16 ha e been epo ed using newe
echnologies which p o ide a highe numbe o me al le els,
o equencies abo e 2GHz [10].
In he p esen wo k he model shown in Fig. 4 is used o
he induc o s conside ing Q alues a ound 5. L
S
and R
S
a e,
espec i ely, he spi al induc ance and he se ies esis ance,
C
S
models he eed- h ough pass c ea ed be ween he wo
e minals due o di ec capaci i e coupling, C
OX
is he oxide
capaci ance be ween he spi al and he silicon subs a e, and
R
Si
ep esen s subs a e losses. Al hough pa e ned g ound
shielding [10] can be used o minimize hese losses, i was
decided o include hem anyway.
L
S
C
S
p
i
R
S
C
OX
C
OX
R
Si
R
Si
p
o
Fig. 4 – Induc o model used in he simula ions.
III.
A
N
A
MPLITUDE
C
ONTROLLED
O
SCILLATOR
To implemen he es s imulus gene a o a scheme based on
he Colpi s oscilla o is p oposed, as i is one o he uned
schemes used in RF ci cui s due o i s simplici y. Fig. 5
shows he scheme o he oscilla o a e being imp o ed
wi h acili ies o con ol he ampli ude o he ou pu signal,
on/o swi ching (M
3
, M
4
), and ou pu bu e ing (M
5
).
The ou pu ol age o he Colpi s oscilla o is, in a i s
app oach, p opo ional o he p oduc o bias cu en (I
B
)
and he ne ank equi alen esis ance R
eq
(Fig. 5.b). This
esis ance is ac ually de e mined by he induc o ’s se ies
pa asi ic esis ance and he o e all equi alen esis ance
seen a he LC ank e minals.
R
eq
L
C
o
2I
B
sin
ω
a)
b)
M
1
M
2
M
3
L
1
R
V
C
2
o
V
DD
V
B
C
1
M
4
M
5
V
C
V
C
I
B
Fig. 5 – Scheme o he Colpi s based a iable ampli ude
oscilla o .
Va ying I
B
would p o ide a mo e linea ampli ude
a ia ion hen ha ob ained a ying R
eq
, howe e , i was
ound p e e able o keep I
B
cons an and o in oduce a
means o con ol R
eq
as a wide a ia ion could be a ained,
namely down o lowe ol age le els. This ea u e is
p o ided by elemen R
V
in Fig. 5.a), which basically is a
ansis o ope a ing in linea mode con olled by i s ga e
ol age. Being he numbe and ampli udes o he se o
s imuli le els o be applied de e mined, elemen RV can
ac ually be implemen ed by a s ack o ansis o s digi ally
con olled. This way a, mo e con enien , digi al in e ace is
allowed.
T ansis o s M
3
and M
4
ope a e in swi ching mode and a e
used o swi ch-o he oscilla o when he ci cui ope a es in
no mal mission mode. This acili y p o ides bo h o a oid
powe consump ion and o c ea e a high impedance ou pu
in no mal mode ( he unc ionali y o swi ch S
T
in Fig. 1).
T ansis o M
5
p o ides ou pu bu e ing.
The dynamic ange equi ed o he oscilla o ’s ampli ude
is de e mined by he LNA’s gain comp ession
cha ac e is ic. The di e en oscilla o ’s ou pu ol age
ampli udes should be dis ibu ed wi h uni o m inc emen s
below and abo e he alue leading o he 1dB comp ession
poin , so ha he polynomial (1) ha bes i s he LNA’s
ans e unc ion can be ound accu a ely. Fo he LNA
unde conside a ion he P
i1dB
co esponding inpu ol age
alue is abou 200 mV (-10dBm @ 50 Ω) peak- o-peak.
Fig. 6 – Oscilla o ’s lowes and highes ou pu le els.
Fig. 6 shows he oscilla o ou pu ol age a he lowes
and highes le els which co espond o abou -21dBm and
-2dBm, espec i ely, obse ed a he LNA inpu .
The a iable ampli ude acili y could ha e been ob ained
main aining he oscilla o ope a ing wi h cons an
ampli ude, ollowed by a a iable gain ampli ie . This
would lead o a la ge and, e en ually, mo e powe hung y
and noisie ci cui .
IV.
LNA
I
NPUT
S
WITCHING
The ansmi / ecei e swi ches used in anscei e s ci cui s
a e ypically implemen ed using disc e e componen s such
as GaAs MESFETs and PIN diodes. Designs using CMOS
pass ansis o s and ansmission ga es ha e also been
p oposed [11], bu ha e he disad an age o p esen ing
highe inse ion loss and nonlinea i y. These handicaps can
be o e come i ansis o s a e no inse ed in he signal pa h,
and a e ins ead used in associa ion wi h passi e de ices. A
swi ch based in his p inciple is p oposed in [12].
Howe e , o equencies below 1GHz, as Q is lowe , he
induc ance se ies esis ance is esponsible o a highe
inse ion loss which deg ades he S
21
pa ame e . To
minimize his e ec a swi ch based on a se ies esona o is
used [13].
L
1
C
1
C
2
O
IM
M
1
M
2
PA
LNA
L
2
VAO
Fig. 7 – An enna swi ch.
Fig. 7 shows he scheme o his swi ch, oge he wi h he
in e ace wi h he VAO and he PA’s ou pu swi ch (IM
block p o ides o impedance ma ching wi h he an enna).
In ecei ing mode bo h M
1
and M
2
a e o , leading L
1
and
C
2
o c ea e a se ies esona o which gi es pa h o he
an enna’s signal and whose inpu and ou pu impedances
ma ches he 50 Ω equi emen . In his case he oscilla o ’s
ansis o M
4
is on, leading M
5
o be o and p esen ing a
high impedance ou pu .
In es mode bo h M
1
and M
2
a e on, leading L
1
and C
1
o
o m a pa allel ank esona ing a he ansmi ing equency
ha c ea es an open ci cui (ac ually 230 Ω). (The same
happens when he PA ope a es in ansmi ing mode.) When
he swi ch is open he oscilla o can be swi ched on o apply
he es s imulus o he LNA inpu . As he swi ch’s ou pu
impedance is essen ially capaci i e, induc o L
2
is included
o adap his impedance o he oscilla o .
Fig. 8 shows S
11
and S
21
pa ame e s o LNA including
he swi ch in he ecei ing mode. The swi ch exhibi s an
inse ion loss o 0.94 dB in on mode and 10.5 dB isola ion
in o mode.
Fig. 8 – S11 and S21 LNA’s pa ame e s, including he swi ch.
Fig. 9 illus a es he wa e o ms obse ed a he an enna
and LNA inpu s, showing ha low dis o ion and low
in e e ence a e in oduced by he inpu swi ch, and he
VAO.
Fig. 9 – Wa e o ms a he an enna and LNA inpu s.
V.
RMS-DC
C
ONVERTER
Fig. 10 shows he scheme o he RMS-DC con e e being
p oposed o de ec he LNA’s ou pu ms ol age. Basically
i is a hal -wa e ec i ie using a diode-connec ed P-channel
MOS ansis o s in a common N-well. The bias ol age V
B
is se o allow measu emen s o be pe o med e en o he
lowes ol age le els. As he es is pe o med using a
known wa e o m i is easy o ela e he measu ed DC
ol age wi h he ac ual ms ol age alue.
RMS
-
DC
ou pu
C
1
V
DD
V
B
LNA
ou pu
M
C
2
R
Fig. 10 – Schema ic o he RMS-DC con e e .
The measu ed ms alue would no be co ec i he LNA
ope a ion eaches deep dis o ion, bu , o pe o m he
p oposed es i is no necessa y o use o high ol age
le els. Fo he highes ol age le els being used, he LNA
ou pu wa e o m s a s o comp ess i s o he nega i e
hal -cycles. Due o his ac , o ob ain a mo e accu a e
measu e, he nega i e hal -cycles a e cap u ed connec ing
he ec i ie ’s e e ence o V
DD
. A ull-wa e ec i ie would
equi e a mo e complica ed ci cui .
VI.
T
EST
S
IMULATION
R
ESULTS
Using he es ci cui y p esen ed be o e, a sequence o 11
equally spaced ol age le els spanning he en i e oscilla o
dynamic ange a e gene a ed con olling p ope ly he
oscilla o . Fig. 11 shows he LNA’s powe ans e
unc ions ob ained obse ing bo h he LNA’s (*) and he
RMS-DC (•) ou pu ol ages, oge he wi h he i s o de
ideal cha ac e is ic (–). I can be seen ha he alues
ob ained wi h he RMS-DC con e e lead o lowe ou pu
powe s o he highe inpu ol ages. This is due, as said
be o e, o he ac ha nega i e cycles s a dis o ing be o e
he posi i e ones. Anyway, a he poin o 1dB comp ession
his e ec is s ill small allowing a measu e o P
1dB
o be
ob ained wi h a -0.5dB e o . Would he posi i e cycles be
de ec ed ins ead, he cu e gi en by he RMS-DC con e e
would be abo e he LNA’s ue cu e wi h a la ge e o .
Fig. 11 – LNA’s powe ans e cha ac e is ic ob ained obse ing
i s ou pu di ec ly (*), using he RMS-DC de ec o (•), and
conside ing only he linea ideal cha ac e is ic.
VII. C
ONCLUSIONS
Ci cui s o s imuli gene a ion, swi ching, and powe
de ec ion a e p oposed o implemen a me hod o es a
LNA o 1dB comp ession and hi d-o de in e cep ion
poin s. Thei de elopmen is based on c i e ia seeking o
minimize powe consump ion and simplici y. Ano he
design d i ing aspec add esses he acili y o con olling
he es ope a ion and o obse ing he ou pu measu es
using digi al o low equency signals, making i easie o
in e ace his es scheme wi h gene al pu pose es e s.
The pe o mance o hese ci cui s is highly dependen on
he cha ac e is ics o he echnology being used, as well as
o he ope a ing equency. This is he case, namely, o he
induc o s’ quali y ac o which can ac ually be highe a
equencies highe han hose conside ed in his wo k. This
could lead, o example, o op o ano he ypology o
implemen he inpu swi ch.
In spi e o he simplici y o he ci cui s being p oposed,
good measu emen esul s can be ob ained. This conce ns
namely he RMS-DC con e e , which is based on a simple
hal -wa e ec i ie . Al e na i e, e en ually mo e accu a e,
RMS-DC con e e s o ecei ed signal s eng h indica o
ci cui s would p o ide mo e accu a e esul s a he cos o a
much highe complexi y and powe consump ion.
Being a con olled oscilla o a ailable a he LNA inpu ,
he blocks placed a e he LNA along he ecei e chain
can also be es ed in sequence.
R
EFERENCES
[1] In e na ional Technology Roadmap o Semiconduc o s, 2004
Edi ion. h p://public.i s.ne .
[2] Benoî R. Veille e and Go don W. Robe s, “A Buil -In Sel -
Tes S a egy o Wi eless Communica ion Sys ems”,
P oceedings o he In e na ional Tes Con e ence, 1995.
[3] E.A. McShane, K. Shenai, B. Blaes, “A monoli hic RF
mic osys em in SOI CMOS o low-powe ope a ion in
adia ion-in ense en i onmen s“ P oceedings o he IEEE
Ae ospace Con e ence, Vol. 5, pp. 421-429, Ma ch 2000.
[4] D. Lupea, U. Pu sche, H.J. Jen schel, “Spec al Signa u e
Analysis – BIST o RF F on -Ends”, Ad ances in Radio
Science, pp. 155-160, Vol.1, 2003.
[5] R. Voo aka anam, S. Che ubal, A. Cha e jee, “A Signa u e
Tes F amewo k o Rapid P oduc ion Tes ing o RF
Ci cui s”, P oceedings o he Design Au oma ion and Tes
Con e ence in Eu ope, 2002.
[6] Je zy Dab owski, “BiST Model o IC RF-T anscei e F on -
End”, 2003, P oceedings o he 18 h IEEE In e na ional
Symposium on De ec and Faul Tole ance in VLSI Sys ems
(DFT’03), 2003.
[7] S. Oze , C. Olgaa d, A. O ailoglu, “Tes abili y Implica ions in
Low-cos In eg a ed Radio T anscei e s: A Blue oo h Case
S udy”, P oceedings o he In e na ional Tes Con e ence,
2001.
[8] G. Pinho, J. Machado Sil a, H. Mendonça, and J. S. Ma os, "A
Tes Me hodology o Compu e Typical LNA Cha ac e iza ion
Pa ame e s ", P oceedings o he XVIII Con e ence on Design
o Ci cui s and In eg a ed Sys ems, No embe 2004.
[9] Thomas H. Lee, The Design o CMOS Radio- equency
In eg a ed Ci cui s, Camb idge Uni e si y P ess, 1998.
[10] C. Pa ick Yue, S. Simon Wong, “Design S a egy o On-
Chip Induc o s o Highly In eg a ed RF Sys ems”,
P oceedings o he Design Au oma ion Con e ence, 1999.
[11] F. Huang, K. O, “A 0.5-µm CMOS T/R Swi ch o 900-MHz
Wi eless Applica ions”, IEEE Jou nal o Solid S a e Ci cui s,
ol. 36, no. 3, pp. 486-492, Ma ch 2001.
[12] Ni anjan A. Talwalka , C. Pa ick Yue, Hai ao Gan, and S.
Simon Wong, ”In eg a ed CMOS T ansmi –Recei e Swi ch
Using LC-Tuned Subs a e Bias o 2.4-GHz and 5.2-GHz
Applica ions”, IEEE Jou nal o Solid-s a e Ci cui s, Vol. 39,
No. 6, Juns 2004.
[13] T. Tokumi su, I. Toyoda, M. Aikawa, “Low Vol age High
Powe T/R Swi ch MMIC Using LC Resona o s”, IEEE
Mic owa e and Millime e s-Wa e Monoli hic Ci cui s
Symposium, pp. 27-30, 1993.