A G-Band SiGe BiCMOS LNA With an Area Efficient Built-In Temperature Compensation Circuit and Robust to TID Radiation

Received 3 September 2024, accepted 15 September 2024, date of publication 23 September 2024, date of current version 3 October 2024. Digital Object Identifier 10.1109/ACCESS.2024.3465850 A G-Band SiGe BiCMOS LNA With an Area Efficient Built-In Temperature Compensation Circuit and Robust to TID Radiation ALVARO URAIN 1, (Student Member, IEEE), DAVID DEL RIO 1,2, (Member, IEEE), CLARA I. LUJÁN-MARTÍNEZ 3, (Senior Member, IEEE), MIKKO KANTANEN 4, (Member, IEEE), AND ROC BERENGUER 1, (Senior Member, IEEE) 1Tecnun, University of Navarra, 20018 Donostia-San Sebastian, Spain 2Ceit-Basque Research and Technology Alliance (BRTA), 20018 Donostia-San Sebastian, Spain 3Department of Electronic Engineering, University of Sevilla, 41004 Sevilla, Spain 4VTT Technical Research Centre of Finland, 02044 Espoo, Finland Corresponding author: Alvaro Urain ([email protected]) This work was supported by Spanish Ministry of Science and Innovation under Grant PID2019-109984RB-C44 (milliRAD Project) and Grant PID2023-146064OB-C33 (Harmonic-Heal Project). ABSTRACT This paper presents a BiCMOS low-noise amplifier (LNA) operating at G-band (140 to 220 GHz) that is robust against harsh operation conditions, namely large temperature variations and radiation exposure. A SiGe technology (IHP’s 0.13-µm SG13G2) has been used for the amplifier design, due to its suitability for millimeter wave (mmW) frequencies and its built-in tolerance to total ionizing dose (TID) radiation. To address the impact of temperature variations on HBTs, the negative temperature dependence of the gain is compensated with a positive temperature dependent collector current, generated by the proposed on-chip and compact (0.0035 mm2) biasing circuit. It has a negligible degradation on the LNA performance, which shows a measured performance comparable with the state of the art with a minimum noise figure (NF) of 9 dB and a gain of 18.1 dB at room temperature, centered around 195 GHz and with a 3-dB bandwidth of 25 GHz. The measured gain and noise figure variations over a 20C to 80C temperature range are ±1.1 dB and ±1.5 dB, respectively. Finally, an experiment to study the TID radiation response of mmW LNAs is carried out. Fabricated samples are exposed to 250 krad(Si), 1.5 Mrad(Si) and 2 Mrad(Si), exhibiting a variation smaller than 1 dB in both gain and noise figure. INDEX TERMS Millimeter wave (mmW), SiGe, G-band, LNA, temperature compensation, total ionizing dose, TID, radiation, space electronics, rad(Si). I. INTRODUCTION The development of new advanced Earth and deep space observation systems, such as radiometers, has increased the demand for more reliable, more resilient, and cheaper spaceborne electronics. Radiometers are instruments that are used to obtain information of the atmosphere or planets by measuring the power (Blackbody Radiation) emitted by gases, such as humidity and ice information from H2O particles at 183 GHz and 200 GHz respectively. Due to the low power level of the received signal, keeping constant The associate editor coordinating the review of this manuscript and approving it for publication was Photos Vryonides . and stable performance of the radiometer electronics is crucial. However, this becomes a challenge due to the harsh environment of space [1], where radiation or temperature can significantly degrade the performance and reliability of electronic components. For example, space radiation, in the form of the interaction of small particles (electrons, neutrons...) with transistors, leads to different degrading effects that can be categorized into two main groups: single event effects (SEE) and cumulative effects. SEE are well known for their critical damage on digital circuits [2] and also in RF circuits, where fast current pulses generated by transients (SET) in amplifiers or mixers can propagate through the receiver [3],[4]. Cumulative effects, especially 138180 ⌦2024 The Authors. This work is licensed under a Creative Commons Attribution-NonCommercial-NoDerivatives 4.0 License. For more information, see https://creativecommons.org/licenses/by-nc-nd/4.0/ VOLUME 12, 2024 A. Urain et al.: G-Band SiGe BiCMOS LNA With Temperature Compensation and TID Robustness Total Ionizing Dose, are characterized by the generation of traps and leakage currents that degrade the performance of devices, therefore having their most impact on RF and mmW circuits. In the last years many works have focused on the mitigation of SEE damage on digital circuits and great effort has been made to minimize its impact on system performance [5]. However, these techniques cannot be directly applied to RF circuits to mitigate TID effects. Some hardening techniques have been proposed for RF circuits [6],[7], however, compensation techniques can have a negative impact, particularly at millimeter wave (mmW) frequencies. Therefore special care must be paid when introducing them. SiGe BiCMOS technologies have become a good candidate for mmW circuits in harsh radiation environments, as SiGe HBTs have a built-in tolerance to total ionizing dose radiation due to device structure [8]. However, the constant radiation exposure can still favour the generation of trapped charges in the emitter-base spacer oxide that may result in unwanted recombination currents that change the current gain, , of bipolar transistors [8]. Therefore, the impact of these unwanted leakage currents on the Gummel curves of HBT transistors has been widely studied in the last few years [8],[9],[10],[11],[12]. These works have concluded that while base currents are affected, the general behaviour remains stable, especially when the transistor is biased close to the optimal transition frequency, ft[8]. Although these works demonstrate the TID tolerance of standalone transistors, it is still necessary to study how these small changes are translated into RF circuits. Some experiments below 70 GHz have verified the TID robustness of SiGe amplifiers up to a few Mrad(Si) (less than 1 dB of gain variation) [13],[14],[15], but, their gain and NF degrade when increasing the TID level above 10 Mrad(Si) [14]. Furthermore, degradation is observed at lower TID levels when the biasing scheme is compromised by radiation. In [6] degradation was observed already at 200 krad(Si) when different bias points were applied. And in [16] a gain degradation of 3 dB at 1 Mrad(Si) was observed, due to the combined effect of the nFET biasing circuit and the displacement damage of proton exposition [17]. Finally, a PTAT current generator was studied in [18], which exhibited TID sensitivity already at a few hundreds of krad(Si). This shows that although amplifier transistors might be robust to radiation, it is also important to include the biasing circuit in the TID tests since the degradation in the biasing circuit might also degrade the amplifier performance. Finally, these studies should also be extended to mmW frequencies where working close to the technology limit could affect the TID robustness. On the other hand, the development of new silicon-based SiGe BiCMOS technologies and the increasing research of recent years have also demonstrated the suitability of SiGe HBT-based circuits at mmW frequencies. They have shown very promising results [19] and have emerged as an alternative candidate to III-V compounds, as they offer significant cost savings and enhanced scalability. However, their main drawback is that the high temperature dependence of HBT transistor characteristics makes it challenging to design mmW integrated circuits robust to temperature variations. For example, the gain of a mmW amplifier without temperature compensation can suffer variations of 4 dB per stage over 40 C to 125 C[20],[21]. Some works have focused on the temperature stability of critical blocks such as PAs [22],[23] and LNAs [24],[25] below 100 GHz, but literature is scarce above 100 GHz, where the effects of temperature can be more severe. Working close to or above half the ftof the technology limits the maximum gain per stage, which prevents the use of compensation schemes that introduce a penalty in the maximum achievable gain. This forces to move towards multistage topologies as the operating frequency increases. The consequence is that the gain degradation due to the temperature of each stage is added and can result in an overall variation of more than 10 dB [26],[27]. This can completely jeopardize the system’s performance if not addressed correctly [28], making it compulsory to compensate for the effect in applications with large temperature variations, such as space. In this work, the impact of the harsh conditions on SiGe amplifiers is studied and mitigated. A temperature compensated 195 GHz LNA is presented, achieving a stable NF of 9 dB (±1.5 dB) and a stable gain of 18 dB (±1.1 dB) over a wide temperature range of at least 20 C to 80 C. This temperature stability is achieved with an on-chip and low-area biasing circuit of just 0.0035 mm2without the need for external or additional circuitry, such as temperature sensors, DACs or microcontrollers. A TID experiment is also proposed to study the response to TID of the LNA and biasing circuit and to check if the robustness of SiGe technologies is compromised by the high operation frequency. The LNA shows less than 1 dB of variation in gain and NF when exposed to up to 2 Mrad(Si) of TID radiation. To the best of the authors’ knowledge, this is the first SiGe LNA that addresses both temperature stability and radiation robustness above 100 GHz, covering the increasing research interest in providing robust circuits for harsh environments at mmW and sub-THz frequencies. The rest of the article is organized as follows: Section II discusses the amplifier architecture; Section III emphasizes the temperature compensation; Section IV describes the performed radiation experiment; Section Vvalidates the approach with measurement results; while Section VI concludes the article. II. AMPLIFIER DESIGN Working at the edge of the technology (above half the ft) leads to an inherently lower available gain and higher noise levels, which results in amplifiers with a performance that narrowly meets the specifications for the target application [29]. Therefore, even small variations caused by harsh conditions of space, such as radiation [6],[16] and temperature [20], [24],[25], can potentially make the amplifier not to meet VOLUME 12, 2024 138181 A. Urain et al.: G-Band SiGe BiCMOS LNA With Temperature Compensation and TID Robustness the required performance. Consequently, it is necessary to include gain boosting and noise reduction techniques to maximize its performance and to make it more robust to the degradation caused by radiation and temperature. In this work, we present a 3-stage single-ended cascode LNA, shown in Fig. 1, that addresses both gain and NF optimization to minimize the impact of variations caused by harsh conditions of space. FIGURE 1. Schematic of the proposed 3-stage LNA. G-Band SiGe amplifiers at mmW frequencies have shown high gains that can go up to 21 dB [30], but this is only achievable with multistage topologies. Several studies have compared common emitter (CE) and cascode topologies’ performance, and have concluded that 3 stages in cascode configuration show a better gain, NF and power consumption trade-off than an amplifier with 5 CE stages [31],[32],[33]. Additionally, reducing the number of stages also simplifies the circuit and reduces the number of performance variation contributors, as gain and NF variations of each stage are concatenated (according to Friis equation). HBT transistors with a fixed area of 0.07 ⇥0.9 µm2and adjustable number of fingers are used. Sizes are incremental with each stage to decrease the output reactance of the stage, and, therefore, to have more relaxed interstage matching networks [32] with the penalty of reducing the Maximum Achievable Gain (MAG) for low current values. Consequently, in order to boost the lost gain, common base (CB) transistor sizes are smaller than the common emitter ones [33], in such a manner that for the same collector current the CB works closer to the peak ft and provides higher gain. Other gain boosting techniques such as inductive base peaking can be applied but it has been discarded due to their high risk of strong instabilities [34]. Noise reduction is performed in the first stage by introducing a shunt transmission line, TLn, between collector and emitter of transistors Q1 and Q2, which resonates out connection parasitics and improves NF performance [35],[36]. Transistors are current biased with current mirrors instead of a direct connection to voltage sources. Current biasing provides more control of the RF parameters of HBTs, as it allows to fix a collector current, Ic, and avoids the exponential sensitivity to Vbe changes. The nominal collector currents of each stage are selected after the biasing study presented in Fig. 2. It relates the bias current of a cascode stage with the theoretical maximum gain (MAG) and minimum Noise Figure (NFmin) that it can provide. For more realistic MAG and NFmin values, EM simulations of transistor core connections and supply and ground planes have been included. Fig. 2also shows the impact of the transistor size selection of each stage (where NQi is the number of emitters of each transistor of the schematic of Fig. 1) both in the power consumption and the gain boosting. The first stage should provide sufficient gain to suppress the noise contribution of the following stages (according to Friis equation), but it also should have low NF, as it sets the baseline for the overall amplifier’s noise figure. Therefore, the bias current of the first stage is selected at 1.8 mA as it shows the best compromise between sufficient gain and low NF. Although the contribution to the overall NF of the second stage is decreased thanks to the gain of the first stage, the noise will slightly increase due to the second stage. Thus, the bias current of the second stage (4 mA) is not selected to maximize gain, but again to find a compromise between gain and NF. Finally, the noise contribution of the third stage is negligible, thus the bias current is selected at 5.7 mA to optimize gain and power consumption. FIGURE 2. (a) Maximum available gain and (b) minimum NF biasing study for each stage including transistor sizes and EM simulations (points represent the selected biasing currents). Finally, Fig. 2also highlights the sensitivity of the amplifier performance to variations in their biasing conditions: current shifts higher than 10% result in an overall gain variation of more than 1 dB. As TID can have a negative 138182 VOLUME 12, 2024 A. Urain et al.: G-Band SiGe BiCMOS LNA With Temperature Compensation and TID Robustness FIGURE 3. EM modelling of the G-Band amplifier using Momentum. impact on the bias circuit [18], in this work voltage references of the bias circuit are left as tuning knobs to compensate for any possible variation introduced by radiation. The 3D view of the layout of the proposed amplifier is shown in Fig. 3. Matching networks are realized with custom-designed transmission lines with Z0=85 and MOM capacitors (0.064 fF/µm2). Supply-ground planes are designed in an interleaved sandwich manner to increase the decoupling capacitance and minimize their impact. Transistor connections have been optimized with careful layout design to avoid possible instabilities due to unwanted couplings or gain decrease due to undesired emitter degeneration. The electromagnetic (EM) behaviour of all passive components is captured with the 2.5D Momentum tool and then cosimulated in Cadence Spectre RF with the active components to extract the post-layout performance. The proposed amplifier achieves a simulated maximum gain of 17.8 dB at 186 GHz with a 3dB BW of 30 GHz and a minimum NF of 10.7 dB at room temperature, which is comparable to the state of the art. The impact of temperature variations before applying any compensation was also studied at simulation level. Fig. 4 shows that the gain variation goes up to 10.5 dB and the NF variation to 4 dB from 20C to 80C. This temperature dependence of the gain completely degrades the amplifier performance, which becomes unacceptable for most systems and shows the need to introduce compensation methods to maintain a stable gain along a wide temperature range. III. TEMPERATURE COMPENSATION Space scenarios under radiation usually also suffer from temperature variations, which result in gain and NF degradation. Temperature effects on HBT transistors are usually compensated with biasing schemes [20],[25] that self-adjust the bias currents depending on the temperature. However, special care must be paid to the bias circuit, as it could be susceptible to radiation [16],[18] and therefore induce bias changes that result in gain and NF degradation (as seen in Fig. 2). Consequently, the proposed bias circuit should not only compensate temperature but also show TID robustness. FIGURE 4. Simulated LNA performance from 20 C to 80 C (a) gain and (b) NF. HBT transistors are a good candidate for this purpose, as they show a high TID robustness. However, the characteristics of SiGe HBT transistors are susceptible to temperature changes, which have a direct impact on their RF performance [37],[38]. In fact, the voltage gain, Avcascode, VOLUME 12, 2024 138183 A. Urain et al.: G-Band SiGe BiCMOS LNA With Temperature Compensation and TID Robustness of a single cascode amplifying stage is approximated [39] as Avcascode(1/Tk)⇡gm1[R0//ZL]⇡qIc kTk ZL,(1) with gm1, the transconductance of the transistor; R0, the output resistance of the cascode stage; ZLthe load impedance (input impedance of following stage); q, value of the elementary charge; and k, the Boltzmann constant. Eq. (1) shows that the gain decreases with temperature, Tk, for a constant collector current, Ic, due to the inverse dependence on temperature of the transconductance, gm(1/Tk)=Ic VT =q kTk Ic,(2) with VT, the thermal voltage. Eq. (1) only captures first order effects, but it can be deduced from it that a positive temperature dependent (PTAT) collector current Ic(Tk), can compensate the complementary dependence (CTAT) of gm. To analyze more complex dependencies and second order effects, transistor level simulations have been used. Fig. 5shows that for a given MAG value there is an optimal Ic for each temperature. This confirms the necessary positive temperature dependence of Ic deduced from Eq. (1), but instead of a PTAT behaviour, some 2nd order effects are obtained due to the oversimplification of Eq. (1). However, due to physical effects, the transition frequency of bipolar transistors also decreases at higher temperature, as shown in Fig. 6, which results in an inherent MAG reduction and a minimum NF increase at higher temperatures [40]. In order to force a specific biasing current, we propose to use a current-based solution instead of voltage-based biasing with additional circuitry, such as temperature sensors, DACs and microcontrollers, that would add more area and power consumption. The proposed solution achieves the desired positive temperature dependent current by introducing 4 resistors to a cascode current mirror, as shown in Fig. 7, where HBT transistors are selected over MOS transistors to avoid performance loss due to TID radiation impact on biasing schemes [16]. In a conventional cascode current mirror, as base-emitter voltages of Q1and Q0can be considered equal, the collector current of the amplifying stage, Ic, is proportional to the size ratio between Q1and Q0transistors, n, Ic=Is1exp ✓Vbe1 VT(Tk)◆=nIs0exp ✓Vbe0 VT(Tk)◆=nI0,(3) with Isi, the reverse saturation current of the HBT transistor; and I0, the collector current of transistor Q0. The negative effect of VT(1/Tk) is first counteracted by having positive temperature dependent base-emitter voltages. As base currents, ib, increase with temperature due to the strong inverse dependence of the current gain [41], this is achieved by introducing base resistors with opposite temperature dependencies, Rb0(Tk) and Rb1(1/Tk). This FIGURE 5. (a) Simulated maximum achievable gain (MAG) of a single cascode stage at 195 GHz for different temperatures and (b) the needed bias current for a stable gain over temperature. FIGURE 6. Simulated transition frequency, ft(GHz), of a single transistor at different temperatures. results in a collector current Ic, Ic(Tk)=nexp ✓ib0(Tk)Rb0(Tk) VT(Tk)ib1(Tk)Rb1(1/Tk) VT(Tk)◆I0 Ic(Tk)=n0(Tk)I0,(4) with n0=nexp ✓ib0(Tk)Rb0(Tk) VT(Tk)ib1(Tk)Rb1(1/Tk) VT(Tk)◆.(5) However, this solution is still not sufficient to compensate the negative dependence of gm(1/Tk) and it becomes necessary to force a positive temperature dependence when generating the I0as well. 138184 VOLUME 12, 2024 A. Urain et al.: G-Band SiGe BiCMOS LNA With Temperature Compensation and TID Robustness FIGURE 7. Proposed temperature compensation biasing scheme. As both Q1and Q0are diode connected in series, it can be considered that they share an equal current value, I0=Iref for big values. The positive temperature dependence behaviour is achieved with a temperature stable reference voltage, Vref , and a CTAT resistor, Rref (1/Tk), I0(Tk)=Iref (Tk)=Vref V0 Rref (1/Tk),(6) with V0as the diode voltage drops of Q01 and Q0, which have a simulated low temperature dependence that can be neglected compared to the resistance dependence. This PTAT behaviour of I0(Tk) can be further increased by subtracting a CTAT current that is generated with a ground PTAT resistor, Rgnd (Tk), resulting into I0(Tk)=Iref (Tk)Ignd (1/Tk) I0(Tk)=Vref V0 Rref (1/Tk)Va(1/Tk) Rgnd (Tk),(7) with Vathe diode voltage drop of Q0. Finally, when the base resistance effect (4) and the I0(Tk) generation (7) are combined, a highly positive temperature dependence is achieved, that is defined as, Ic(Tk)=n0Iref (Tk)Ignd (1/Tk),(8) with, n0=exp ✓ib0(Tk)Rb0(Tk) VT(Tk)ib1(Tk)Rb1(1/Tk) VT(Tk)◆ Iref (Tk)=Vref V0 Rref (1/Tk) Ignd (1/Tk)=Va(1/Tk) Rgnd (Tk).(9) The used unsalicided (Rref and Rb1) and salicided (Rgnd and Rb0) poly-silicon resistances show a temperature behaviour as follows R(Tk)=R(T0)(1 +↵(TkT0)).(10) with temperature coefficient values, ↵ref ,↵gnd ,↵b0and ↵b1, that are set by the technology. The values of Rb0and Rb1 are 10 kand 800 respectively, and are selected to find a compromise between low noise (with higher Rb1) and high temperature dependence (with higher Rb0/Rb1ratio). The values of Rgnd (Tk) and Rref (1/Tk) are 600 and 820 and are obtained by approximating the expression of Ic(Tk)(9) to one of the ideal positive temperature dependent currents of Fig. 5b. The temperature compensation is demonstrated in Fig. 8. The proposed solution reduces the gain variation from 10.5 dB to 2 dB with a penalty in the 3 dB bandwidth that is reduced from 30 to 25 GHz. Temperature compensation across corners is presented in Fig. 9. While the compensation is not compromised at the best case scenario (bcs), the performance needs to be restored at the worst case one (wcs). Therefore, the Vref of the last stage is left as a tuning knob to recover gain and any compensation loss due to process variations and simulation inaccuracies. FIGURE 8. Simulated gain of the LNA at different temperatures after the temperature compensation. Dashed curve is the simulated gain performance before compensation. FIGURE 9. Simulated gain variation at central frequency of 185 GHz for different corners. The voltage references of the proposed bias circuits are also left as tuning knobs to compensate possible current shifts introduced by TID radiation. As explained with Fig. 2, gain and noise performance are highly dependent on the collector current of each stage and Eq. (8) and (9) show that the bias generation is sensitive to variations of base currents and diode voltage drops. Although previous studies have reported low ibvariations due to leakage currents induced by TID VOLUME 12, 2024 138185 A. Urain et al.: G-Band SiGe BiCMOS LNA With Temperature Compensation and TID Robustness radiation, it is still necessary to verify its robustness, given the significant impact these variations could have after the exponential operation shown in Eq. (9). IV. TEST DEFINITION TO STUDY THE AMPLIFIER’S ROBUSTNESS TO TOTAL IONIZING DOSE The previous section covered the impact on SiGe amplifiers of temperature changes that can be experienced in harsh environments and proposed a compact and efficient solution to mitigate its effect on LNAs. However, device reliance should not only be verified against temperature but also against radiation. For example, satellite electronics at LEO, GEO and Highly Elliptical Orbits are exposed to radiation levels that can go up to 100 krad(Si) depending on their target orbit and inclination [42],[43],[44]. The radiation exposure is further increased in deep space exploration missions (like Jupiter’s moon, Europa, exploration) that suffer from higher TID levels and also require long-term reliability [45],[46], thus demanding radiation verifications up to 2 Mrad(Si) levels (H radiation dose level indicated at the European Space Components Coordination (ESCC) Basic Specification No. 22900 [47]). Many studies have shown that SiGe HBTs under the mentioned TID radiation levels present very small variations thus proving its robustness [8],[9],[10],[11],[12]. But only a few works have analyzed how these small changes are translated to amplifier performance at lower frequencies [6], [13],[14],[15],[16]. Therefore, it is still necessary to expand these studies to mmW frequencies, where circuits are working close to or above half their fT. Furthermore, previous experiments [6],[16],[18] showed the negative impact of biasing schemes on TID robustness. Thus, it becomes compulsory to verify with a radiation experiment the impact of the proposed biasing circuit. To test the reliability of the proposed SiGe amplifier, a total of 12 samples organized into groups of 3 are exposed directly to the beam, as shown in Fig. 10. Each group is exposed to different levels of TID: •Group A =252 krad(Si) (19 hours) •Group B =1.5 Mrad(Si) (114 hours) •Group C =2 Mrad(Si) (157 hours) •Group D =tracking group Group A’s TID level is chosen to test the reliability of electronic components for Earth orbital missions, whereas groups B and C are exposed to higher TID levels to test long-term reliability and deep space exploration conditions. To have a deeper understanding of the radiation impact, an extra set of samples is also irradiated (group D) as a tracking group. It is exposed to a total TID level of 2 Mrad(Si) and it is DC characterized at intermediate levels (same as groups A and B) to monitor and study the progression of the effects. Finally, the control group is composed of 6 unirradiated samples and is used to compare and find performance variations. The radiation experiment is performed at the National Center of Accelerators (CNA) of Spain in collaboration FIGURE 10. (a) Co-60 irradiator and (b) group of samples exposed to different TID radiation levels. with the University of Sevilla, where samples are exposed to different photon radiation levels with a Co-60 irradiator (model Gammabeam™X200 from Best Theratronics), shown in Fig. 10. The dies are located 46 cm from the source in order to receive a radiation rate of 13.3 krad(Si) per hour. This rate is within the indicated range in the European Space Components Coordination (ESCC) Basic Specification No. 22900 for High Dose Rate (HDR) tests (dose rate within 3.6 to 36 Krad(Si)/hour). Fig. 11 shows the flow diagram of the proposed experiment to test the robustness of the SiGe amplifier. First, a DC- characterization of the control group is performed before radiation exposure. This test verifies if the DC current signals respond correctly to DC voltage changes, and it is performed without any RF input or output signal, just with DC probes and by progressively increasing reference voltages and extracting the current consumption curves of each gain stage. Before radiation exposure, an RF characterization is also performed to analyze the gain and NF performance of the control group. The next step after irradiation is to perform a DC-characterization in less than 1 hour from the extraction 138186 VOLUME 12, 2024 A. Urain et al.: G-Band SiGe BiCMOS LNA With Temperature Compensation and TID Robustness FIGURE 11. Radiation test diagram. moment. The supply current results are then compared with the pre-irradiation measurements to check for any radiation effect. As shown in Fig. 2, this verification after irradiation is crucial, as any significant variation in the bias current could result in an RF performance degradation. The DC characterization of the control group (unirradiated) showed a DC variation of around 10% between samples due to process variations and measurement uncertainties. Therefore, in this work, DC supply current variations higher than 10% are associated with the TID. Right after DC measurements, samples are stored at 18 C in vacuum-packed packages to minimize any unwanted recovery effect until RF characterization. Special care is taken although studies have shown that samples at room temperature within that time window do not show any recovery effect [47],[48]. Finally, samples are RF characterized in less than 7 days at the MilliLab facilities of VTT Technical Research Centre in Finland. Their performance is then compared to non-irradiated RF characterization to determine the impact of TID exposure, which is observed when gain and NF variations are higher than the uncertainties introduced by process variations between samples and by the measurement set-up, such as probe landing. As seen from measurements of the control group, these variations can go up to 1.5 dB. V. FABRICATION AND MEASUREMENT RESULTS The amplifier was manufactured in the 0.13-µm SG13G2 technology offered by IHP, with ft/fmax =300/500 GHz HBT transistors. The chip micrograph is shown in Fig. 12, which occupies an overall area of 0.71 ⇥0.77 mm2with pads. The proposed positive temperature dependence biasing scheme is implemented for the three LNA stages, and they only add an extra area of 0.0035 mm2per stage to the amplifier core circuit, which occupies an active area of 0.06 mm2. S-parameters were measured with an MPI TS2000-SE wafer probing station using a Keysight N5245A network analyzer and external WR-5 frequency extenders. NF measurements were performed using a custom-made ELVA-1 noise source and calculated with the Y-factor method [49]. FIGURE 12. Chip micrograph (0.55 mm2) with the core LNA (0.06 mm2) and the biasing circuits for each stage (0.0035 mm2). A. TEMPERATURE MEASUREMENTS The LNA has been first characterized by measuring 6 different chips showing repetitive and consistent behaviour between all of them. The probe station is equipped with an ERS AC3 thermal chuck which was used to sweep the temperature from 20 C to 80 C, and data was extracted at 7 different intermediate temperatures. Fig. 13 presents the S-Parameter and noise measurements from 20 C to 80 C. The LNA is shifted up less than 10% from simulations and is centered at 195 GHz, showing a 3-dB BW of 25 GHz with a maximum nominal gain of 18.1 dB at room temperature. The amplifier is matched to its minimum NF at the input and to 50 Ohms at the output, both showing acceptable S11 and S22 values in the frequency band. Fig. 14 shows that at 195 GHz a small gain variation across 20 C to 80 C is achieved with the proposed solution. VOLUME 12, 2024 138187 A. Urain et al.: G-Band SiGe BiCMOS LNA With Temperature Compensation and TID Robustness FIGURE 13. Measured LNA performance from 20 C to 80 C (a) gain and NF (b) S11 and S22. The gain variation in the considered temperature range is just ±1.1 dB, which confirms the simulation results, where a variation of 10.5 dB across the complete temperature range was reduced to 2 dB when the proposed bias scheme is included. The minimum NF is 9 dB at room temperature and has a variation of ±1.5 dB in the whole temperature range. Fig. 14 also shows that, as expected, a positive temperature dependent current consumption is achieved. The overall power consumption of the LNA including the three bias circuits is 35 mW. The first and second stages share the same reference voltage, while the third is used to adjust process variations, as mentioned in Section III. Linearity is not characterized by measurements, but simulations show a simulated input 1 dB Compression Point of 19.5 dBm at the center frequency with a variation of ±1.2 dB in the whole temperature range, much better than the performance without compensation, which showed a variation of more than ±4 dB in the temperature range. Table 1compares the proposed LNA with previously reported temperature compensated amplifiers at lower bands and with mmW SiGe amplifiers. The experimental results show comparable performance to mmW state of the art while achieving a similar temperature compensation to the solutions presented at lower bands. FIGURE 14. LNA temperature performance (a) gain and NF variation at the central frequency, 195 GHz, and (b) positive dependent current consumption. TABLE 1. Comparison with previously published G-Band and with temperature compensated LNAs at lower bands. B. POST-RADIATION MEASUREMENTS The aim of the radiation experiment was to analyze the impact of TID radiation on the biasing circuit and the LNA behaviour by comparing the supply current, gain and NF performance of samples under different TID radiation dose levels. A DC characterization was performed right after irradiation in order to detect any radiation effect on samples, as indicated in the flow diagram of Fig. 11 and explained in Section IV. As shown in Fig. 15, the current consumption 138188 VOLUME 12, 2024