HTML
-
The terahertz (THz) regime, loosely defined as 0.1 THz–10 THz, is a unique frequency band in observing cold and dark universe, which is invisible in the optical/near-infrared regime.[1] The Purple Mountain Observatory of Chinese Academy of Sciences is leading the development of a terahertz spectrometer (also known as High-Sensitivity Superconducting Detection Module, HSTDM[2]) onboard China’s Space Station Telescope (CSST) as one of five CSST’s scientific instruments. The THz spectrometer mainly observes terahertz spectral lines in the frequency range of 0.41 THz–0.51 THz (e.g., atomic carbon CI line at 0.492 THz). Its receiver frontend is mainly composed of two NbN-based superconductor-insulator-superconductor (SIS) mixers[3] and two following intermediate-frequency (IF, 0.1 GHz–1.1 GHz) low-noise amplifiers, which are cooled down to approximately 8 K by a two-stage pulse-tube cryocooler for space applications.[4] As the cooling capacity of the two-stage pulse-tube cryocooler is limited, the adopted CLNAs should have power dissipation of no more than 5 mW, yet with a noise temperature below 10 K.
As is well known, cryogenic InP-based high-electron-mobility-transistors (HEMTs) have been widely used in radio astronomical receivers because of their state-of-the-art noise performance.[5] However, their operating frequency is typically above 1 GHz and power dissipation is above 5 mW. Recent studies demonstrate that SiGe-based HBTs perform well at frequencies as low as 0.1 GHz and at temperatures down to 4 K.[6–8] Furthermore, HBTs appear superior to HEMTs in power dissipation.[6,9] SiGe-based HBT CLNAs are thus of particular interest for space applications and large complex systems such as multibeam/phased-array receivers and quantum computers. We chose SiGe-based HBTs to develop 0.1 GHz–1.1 GHz CLNAs for our THz spectrometer. Through this development, we firstly build a small-signal and noise model at low temperatures and then study the noise and gain dependence upon DC power dissipation at different temperatures.
-
Small-signal and noise models are usually established separately in designing low-noise amplifiers. For HBTs, a combined small-signal and noise model has been proposed,[10,11] as shown in Fig. 1. It can be seen clearly that there are mainly thermal noises from uncorrelated base/collector/emitter physical resistances (i.e., vn,b, vn,c, and vn,e) and partially correlated base and collector shot noises (i.e., in,b and in,c).[12] Note that when the operating frequency is less than fT/3, (fT is transistor’s cutoff frequency) this correlation is rather limited and can be ignored.[13] As we adopt packaged transistors like industry standard SOT343 in our design, an emitter parasitic inductor Le is introduced to take account of the packing effect that is considerable especially for a large relative bandwidth like 0.1 GHz–1.1 GHz. The parasitic inductance Le is primarily responsible for the decrease in amplifier gain with increasing operating frequency. The small-signal and noise parameters can be extracted from the measured RF and DC characteristics at different temperatures, respectively.
Once the critical model parameters have been defined, the radio frequency performance of the amplifier can be simulated using the small signal and noise model discussed previously. It is worth noting that some parameters exhibit less sensitivity to ambient temperature fluctuations. Consequently, these parameters can be determined from the DC and RF characteristics measured at room temperature. They include the base resistance rb, collector resistance rc, emitter resistance re, and the parasitic collector-base capacitance Ccb.
Before determining all the model parameters, we estimated the minimal noise temperature Tmin associated with the operating temperature and bias current based on the transistor’s DC characteristics.[9,10] It can be evaluated with a simpler expression as
where Ta is the ambient temperature, β = Ic/Ib is the current gain, ncx = Ic/(GmVT) is the extrinsic collector current ideality factor, Gm = dIc/dVbe is the extrinsic transconductance, Ic is the collector current, VT = kTa/q is the thermal voltage, k is Boltzmann constant, and q is the magnitude of charge of an electron. This simplified noise-temperature expression is valid at frequencies that are much lower than the transistor’s cutoff frequency. Obviously, the current gain β and transconductance gm, which are temperature-dependent, can be characterized through the DC measurements of the transistor at temperatures from 290 K to 4 K.
For BFP842ESD SiGe transistors, figure 2 illustrates the dependence of the current gain β and extrinsic transconductance Gm on the collector current at 290 K and 4 K. Based on the trade-off between gain and noise, an optimal collector current of 4 mA was selected. At this bias point, the temperature-dependent behaviors of β and Gm are plotted in Fig. 3, with additional details provided in Table 1. The sharp increase of the two key parameters at cryogenic temperatures leads to a significant reduction in the noise contribution of the transistors. Based on the measured values of the transistor current gain β and transconductance Gm at different temperatures, the minimum noise temperature Tmin was calculated, as depicted in Fig. 4. It can be clearly seen that this type of transistor exhibits excellent noise characteristics even at low collector currents, thereby making it suitable for the development of our 0.1–1.1 CLNAs particularly with low power dissipation.
After evaluating the transistors’ cryogenic noise characteristics, we can select some transistors suitable for cryogenic applications and then extract small-signal model parameters. An open-collector method was utilized to determine the emitter resistance re. By setting the collector current to zero, the collector–emitter voltage Vce was characterized as a function of the applied base current Ib. The value of re was then derived according to re = dVc/dIb. To obtain the collector resistance rc, the DC source was configured in current–source mode, enabling precise control over the current Ic. Firstly, the collector voltage Vc was measured as a function of the base current Ib under a fixed Ic. Secondly, the relationship between Vc and Ic was established under a fixed base current Ib. It should be noted that calibrating the resistance of the DC test line was essential for this measurement. The value of rc can be obtained by
where rp is the resistance of the DC test line. As for rb, we can assume rb ≈ rc, since the noise of the transistors is mainly the shot noise caused by the PN junction current at the ambient temperature close to absolute zero. The base-emitter junction admittance gbe defined as gbe = dVb/dIb, can also be extracted through DC measurement. To determine the baseemitter junction capacitance Cbe and the collectorbase parasitic capacitance Ccb, the through-reflect-line (TRL) de-embedding calibrations were carried out for the measurement of S parameters at low temperatures. Note that for the extraction of the collector-base capacitance Ccb, we measured the S parameters under off-bias conditions, ensuring that the junction was reverse-biased. In this case, both the transconductance gm and the base-emitter admittance gbe are almost zero, leaving only the parasitic capacitance of the transistor. Subsequently, the S-parameters of the transistor were converted into Y-parameters to determine the value of Ccb, which is given by
The base-emitter capacitance Cbe, which is dependent on the transistor’s bias, can be determined in terms of the active bias S-parameters. Firstly, the S-parameters are converted into Z-parameters to eliminate the contributions of resistances rb, rc, and re. Secondly, the Z-parameters are transformed into Y-parameters from which Cbe is extracted using the following equation:
At this point, two key parameters remain to be determined: the transconductance gm and the emitter parasitic inductor Le. The relationship between the external transconductance Gm and the intrinsic transconductance gm is expressed as Gm = gm/(1+gmre). We can have gm ≈Gm when the emitter resistance re is small. The emitter parasitic inductor Le can be determined by fitting the cryogenic forward transmission coefficient S21 of the amplifier circuit.
The small-signal model for the BFP842ESD transistor was established through the parameter extraction process described above, as listed in Table 2. With this model, its cryogenic scattering parameters were simulated. For the evaluation of the noise characteristics of the transistor, the two shot noises were calculated based on a base bias current Ib = 2 μA and a collector current Ic = 4 mA, respectively. The thermal noise contributions from the three resistive elements at ambient temperature Ta were calculated straightforwardly. This approach enabled an accurate estimation of both noise and gain characteristics, thereby facilitating the design and optimization of cryogenic low noise amplifiers. The comparative analysis of the simulated and measured results, with respect to the amplifier’s gain and noise temperature, is presented in Section 4.
-
Based on the combined small-signal and noise model for the BFP842ESD transistor, a two-stage cryogenic amplifier operating from 0.1 GHz to 1.1 GHz was designed. The circuit schematic and the assembled discrete amplifier are illustrated in Fig. 5. To enhance the input impedance matching, gain flatness and stability, resistive negative feedback was implemented between the collector and emitter. It is well known that at room temperature the resistive feedback is generally considered to be an impractical option due to the increased thermal noise from the feedback resistance Rf. As the feedback resistance is cooled from 290 K to 4 K, however, its thermal noise reduces by a factor of 72. In addition, for β / Rf ≪ gm/2, the thermal noise introduced by the feedback resistance can be ignored.[14] Consequently, adding a high value feedback resistance for impedance matching does not adversely affect the noise performance of the amplifier. In our design, the use of RF-glass at the input and output ports provides superior stability under extreme thermal and mechanical conditions. Additionally, a small series resistor is introduced between amplifier stages, increasing the stability factor (K-factor) from 2.5 to 3.8 and thereby providing a larger stability margin while mitigating potential oscillations due to device aging. Furthermore, several bypass capacitors of 470 pF were placed near the transistor, while a larger capacitance of 0.1 μF was connected in parallel at the DC bias port to stabilize the transistor’s bias.
The microstrip lines of the amplifier were fabricated on a substrate Rogers 5880 with a thickness of 0.127 mm. This choice of thin dielectric material improves the isolation between the input and output ports of the circuit. It is worth noting that the transistors are mounted close to the via holes to minimize parasitic inductance and capacitance, which can significantly impact the amplifier stability at cryogenic temperatures. Furthermore, the cavity of the amplifier was carefully optimized through an isolated design of the DC and RF paths. This approach effectively suppresses potential resonances within the cavity, ensuring stable operation across the specified frequency range.
-
The noise temperature and gain of the 0.1 GHz–1.1 GHz low-noise amplifier were measured inside a 4-K cryostat. The cryogenic noise measurements were performed using the cold attenuator method with an estimated noise-temperature accuracy of ±1 K.[15,16] The measured and simulated gain and noise temperature of the amplifier at 4 K are shown in Fig. 6. It can be clearly seen that both the measured gain and noise temperature agree well with the simulation ones, validating the approximate modeling approach used for the design. As you can see, the gain curve without the parasitic inductance Le is quite different from that with Le. To mitigate the gain degradation at higher frequencies, an equalization circuit has been implemented in the IF chain, leading to a substantial improvement in the spectral characteristics of the IF output. At the ambient temperature of 4 K, the optimal noise temperature occurs at a collector voltage of 1.6 V and a current of 8 mA. Under this condition, the average noise temperature across the frequency range of 0.1 GHz–1.1 GHz is 4.5 K, with a minimum value of 3.7 K at 0.16 GHz. The amplifier has a gain exceeding 35 dB throughout the entire operating frequency range. Additionally, the cryogenic amplifier’s input and output voltage standing wave ratio (VSWR) values are both below 2, as shown in Fig. 7.
Figure 8 further demonstrates the dependence of the gain and noise temperature on the collector current at 0.5 GHz. Obviously, both the gain and noise temperature are insensitive to the collector current. Note that at a collector current of 2.6 mA, corresponding to a power dissipation of only 3 mW, the amplifier still demonstrates a gain greater than 32 dB and an average noise temperature of 6.8 K.
The amplifier underwent comprehensive environmental testing, including mechanical, thermal cycling, thermal vacuum, high-temperature life, and irradiation testing. As shown in Fig. 9, the amplifier’s gain and noise temperature remain unchanged, demonstrating its robustness for space applications.
A comparison of the amplifier’s performance against recently published state-of-the-art SiGe cryogenic low-noise amplifiers is summarized in Table 3. The proposed amplifier demonstrates a combination of low noise temperature, high gain, and low power consumption, highlighting its suitability for space applications.
-
A 0.1 GHz–1.1 GHz cryogenic low-noise amplifier (CLNA) has been developed for the terahertz spectrometer onboard China’s Space Station Telescope (CSST). A combined small-signal and noise model was proposed to facilitate the amplifier design. Utilizing HBT transistors, this CLNA achieves good noise performance with low power dissipation. At a power dissipation of merely 3 mW, it exhibits an excellent average noise temperature of 6.8 K while maintaining a gain of over 32 dB. CLNAs of this kind, with low power dissipation and yet low noise temperature, should be of particular interest to space applications and complex systems.
首页
登录
注册


DownLoad: