CHARACTERISTICS OF BROADBAND INP HFET MILLIMETER-WAVE AMPLIFIERS AND THEIR APPLICATIONS IN RADIO ASTRONOMY RECEIVERS(1)
Abstract--Recent developments in ultra-low-noise, cryogenically-cooled, heterostructure field-effect transistor (HFET's)
receivers for frequencies up to 110 GHz are reviewed. Design and examples of the realization of InP HFET receivers in the
frequency range 18 to 110 GHz are described. Applications to ultra-low-noise radio astronomy receivers, as well as
broadband continuum radiometers, are discussed.
Keywords: field-effect transistor, high-electron-mobility transistor, heterostructure field-effect transistor, noise, noise modeling,
low-noise amplifiers, radiometers, radio astronomy
1. INTRODUCTION
In the early 1970's, the ultra-low-noise receiving systems employed mainly solid-state masers, cryogenically-cooled parametric
amplifiers (or converters) and Schottky diode mixers. At the end of that decade, the advances in the technology of GaAs FET's,
combined with cryogenic cooling, made the noise performance of GaAs FET amplifiers competitive with the performance of
parametric amplifiers [1], [2]. Also, a new mixing element, superconductor-insulator-superconductor (SIS) junction, capable of
quantum limited detection, was developed [3]-[5].
The progress in the performance of FET and HFET amplifiers was quite dramatic: from state-of-the-art noise temperature of 20
K at 4.75 GHz in 1980 [1] to the state-of-the-art noise temperature of 15 K at 43 GHz in 1993 [6] at the ambient temperature of 18
K. This last result was achieved with InP lattice matched HFET's. Since 1993 the progress in the noise performance has not been
significant. However, the technology of InP devices has matured and allowed for system insertion. The hybrid "chip and wire"
amplifiers have been demonstrated up to frequencies of 110 GHz and successfully used in several instruments for radio astronomy
research. These include: Very Large Array (VLA), Very Large Baseline Array (VLBA), Microwave Anisotropy Probe (MAP) and
several ground-based instruments for the investigation of cosmic microwave background. A number of MMIC designs were
demonstrated up to 140 GHz (for example, [16]) with even higher frequencies under development. In radio astronomy
instrumentation, HFET receivers now compete in performance with masers and SIS mixer/HFET IF amplifier tandems for
frequencies below 120 GHz. At frequencies above 120 GHz up to about 1 THz, SIS mixers demonstrate the best noise performance.
Above 1 THz, cooled Schottky diode mixers and hot electron bolometer (HEB) mixers provide the lowest noise temperatures [8].
2. LOW-NOISE HFET'S: STATE-OF-THE-ART
For years, device technologists in their quest for low-noise FET (HFET) were guided by the relation developed by Fukui [10].
where Tmin is the minimum noise temperature, f is the frequency, To is the standard temperature 290 K, fT is the intrinsic cut-off
frequency, gm is the transconductance, rg and rs are gate and source parasitic resistances, respectively, and Kf is the fitting factor,
assuming values between 1.2 and 2.5. A corresponding approximate expression for the minimum noise temperature of a FET chip,
based on the noise model developed by the author [11], can be written as:
where Tg and Td are equivalent gate and drain temperatures, respectively; gds is the drain-to-source conductance; and rt = rs + rg + rgs, where rgs is the intrinsic gate resistance. Tg is approximately equal to the physical temperature Ta of a device.
The comparison of both expressions demonstrates that the technological progress towards increasing the values of fT and reducing the values of rg and rs would result in lower noise temperatures (noise figures) as explained by both. The open question is whether the improvements in the noise performance of HFET's, be it conventional, pseudomorphic or lattice-matched, were only a consequence of increased values of intrinsic cut-off frequencies and reduced values of parasitic resistances, or were they also a result of the reduced value of the product Td gds at optimal bias conditions [9].
It is well known that the minimum noise figure of a multistage amplifier with sufficient gain is determined not only by the device
noise figure, but also by its associated gain. A more proper figure of merit is the minimum noise measure which determines the
lowest noise temperature (noise figure) of an amplifier made of an infinite number of stages. It is, therefore, a very useful tool in
establishing the limits of performance of practical amplifiers. An example of the dependence of minimum noise temperature Mmin
versus frequency for the state-of-the-art AlInAs/GaInAs/InP HFET's is shown in Fig. 1 [13]- [16].
3. DESIGN AND PERFORMANCE OF LOW-NOISE, MILLIMETER-WAVE CRYOGENIC AMPLIFIERS
Design of cryogenically-coolable amplifiers requires knowledge of the signal and noise properties of FET's at cryogenic
temperatures. In one of the first studies of cryogenic properties of GaAs FET's [19], it was observed that the changes in the
measured S-parameters upon cryogenic cooling could be mostly explained by the changes in the device transconductance gm and
drain-to-source conductance gds. A large number of later papers devoted to the investigation of the signal and noise properties of
FET's and HFET's at cryogenic temperatures confirm this early observation (for example, [6], [7], [11], [12], [17]-[22]).
This observation, together with the noise model of [11], provide sufficient information to allow for computer-aided design of
cryogenic amplifiers with optimal, according to some criterion, noise bandwidth performance. The criterion could be different for
different applications: for example, minimum value of the maximum noise temperature across a given bandwidth, minimum value
of the average noise temperature over a given bandwidth, etc. Usually an amplifier has to satisfy other requirements, as the input
and output return loss, unconditional stability, minimum gain and gain flatness, etc., which can also be reliably investigated in a
CAD process. As a result, there are no "set-in-stone" design rules; yet the following observations could prove valuable. In similarity
with an impedance matching, the "noise matching" of a FET can be done perfectly only at a single frequency. Consequently,
wideband "noise matching" is similar to wideband matching of a complex impedance to a resistive load. For a given frequency
range, there is a particular gate width of a FET (HFET) which allows for the largest possible relative bandwidth. A proper choice
of the inductance in series with the gate (for example, the inductance of the gate bond wire) usually results in the simplest broadband
"noise matching" network. Also, a proper value of the source inductance allows one to address other requirements (stability, input
return loss, gain flatness) without penalty in the noise temperature of a multistage amplifier.
The NRAO Central Development Laboratory has developed a number of InP HFET millimeter-wave amplifiers in the 18 to 110
GHz range for applications in radio astronomy receiving systems [6], [7], [17], [18], [27], [28].
An example of noise and gain characteristics and the comparison with the model prediction for a room temperature InP HFET, six-stage, W-band amplifier is shown in Fig. 2. The devices have gate dimension .1 x 50 m and are biased at Vds = 1.0 V and Ids = 5
mA. For the purpose of modeling, the equivalent circuit given in [17] is used. The noise model of [11] is assumed for noise
computation with Tg = 297 K and Td = 1500 K. All other resistors and lossy lines (1.7 dB/inch at 75 GHz) are assumed to be at Ta
= 297 K. An example of noise and gain characteristics and the comparison with the model prediction for a cryogenic amplifier is
shown in Fig. 3. The transistors were biased at Vds = .9 V and Ids = 3 mA in the first two stages and Ids = 5 mA in the last three
stages. For the model data, the only changes from room temperature were: Tg = Ta = 27 K and Td = 500 K and about 20% increase
in transconductance gm.
A summary of noise performance of HFET receivers cooled to approximately 20 K and a comparison with typical performance of
SIS mixer receivers cooled to 4 K is presented in Fig. 4. Fig. 4 also shows the best performance of InP HFET receivers expected
at any frequency for current state-of-the-art devices. This graph is based on minimum noise measure of a state-of-the-art, .1 µm gate
length InP HFET, shown in Fig. 1, and includes a modest correction to the receiver noise temperature estimated from typical losses
of matching circuits, waveguides, horns, dewar windows, etc. A broadband HFET receiver design can attain the noise temperature
equal to the minimum noise measure only at a single frequency, a property clearly illustrated in Fig. 4. The examples of HFET
receiver noise temperature, shown in Fig. 4, demonstrate that for a typical rectangular waveguide bandwidth, the average noise
temperature is approximately equal to the value determined by the minimum noise measure at the highest frequency within the band.
The detailed characteristics of many of the amplifiers used in the construction of the laboratory receivers of Figure 4 will be
discussed during presentation. This discussion will include examples of complex gain match over a waveguide band frequency
range. For example, a gain match of ± 2.5 dB and a phase match of ± 25 degrees can be routinely obtained in NRAO hybrid designs
at W-band.
4. RECEIVERS FOR RADIO ASTRONOMY
The general concept of a compact, HFET receiver for radio astronomy application has been outlined by Weinreb et al.[22] and
several examples are given therein. The noise temperature of state-of-the-art receivers versus frequency can be estimated from the
data shown in Fig. 4. In the examples of Fig. 4, only the noise contribution of a dewar window and that of a horn at cryogenic
temperature are responsible for the difference between the amplifier and receiver noise temperature. In practical receivers, an
addition of different other components (polarizer, orthomode transducer (OMT), hybrids, circuits for the injection of the calibration
signals) may be needed. These, depending on their physical temperature, may further degrade the performance of the systems, both
in noise and bandwidth.
Recently, the modest cooling requirements, large instantaneous bandwidth, and very low noise of HFET millimeter-wave amplifiers
have spurred the interest in their use as broadband continuum radiation detectors, especially for cosmic microwave background
radiation measurements (for example, [23], [24]). The radiometric sensitivity of such receivers may be limited by random gain
fluctuations of broadband HFET amplifiers.
By design, the voltage response of a radiometer in the Rayleigh-Jeans limit is a linear function of the incident noise power
where kb 1.4 x 10-23 J/K, Tsys is the system temperature (the sum of the receiver and antenna noise temperatures), and rf is the
effective RF bandwidth. Sufficient RF gain and detector responsivity must be present for the first-stage noise to dominate the
contributions from the square-law detector, with responsivity R, and subsequent post-detection electronics.
In the presence of gain fluctuations, the following parameterization for the radiometer's spectral density V (in Volts/rt-Hz) may be employed:
where 0 is the demodulation efficiency and g2(f) is the radiometer's total noise spectral density of the gain fluctuations as a
function of video frequency. A simple power law dependence with frequency is observed for the gain fluctuation in HEMT's g2(f)
(1 Hz/f) where g2 is the amplitude and 0.9 is the spectral index of the spectral density [25]. After calibration, it is
convenient to express the sensitivity in terms of the spectral noise density T in mK/rt-Hz.
For a direct detection receiver, 0 = 1. From (4), one notes for such a receiver in the narrow bandwidth limit, the ratio of spectral density to the mean signal output approaches a constant magnitude . In the wide bandwidth limit, the radiometric offset and gain stability determine the effective spectral density V/Vdc = (1Hz/f) /2 g. The so-called "knee" or "corner" frequency fknee ~ (rfg2/2)1/, occurs at the transition between these two regions.
An example of the output power spectra of a radiometer built with amplifiers similar to those of Figs. 2, 3 and 4 is shown in Fig.
5. The largest bandwidth of this radiometer was 20 GHz, but smaller bandwidth could be selected by bandpass filters (comp. Fig.
5). The system temperature was measured with a feed horn viewing a hot/cold load through a mylar vacuum window. The measured
receiver noise was 300 ± 10 K at room temperature and 80 ± 10 K when cooled to a physical temperature of 50 K. The measured
receiver sensitivity was 3 mK/rt-Hz at an ambient temperature of 300 K and 0.8 mK/rt-Hz at 50 K for video frequencies >> 10 kHz
and an effective RF bandwidth of 20 GHz. For both bandwidths, 3 GHz and 20 GHz , the noise at a video frequency of 100 kHz
was within ~ 5% of the magnitude computed from the measured system temperature and effective bandwidth. The measured
sensitivities at room temperature are the best reported for a continuum receiver in the 3 mm atmospheric window. At 50 K, the
observed receiver sensitivity is competitive with the cryogenic performance of sub-Kelvin bolometric detectors or SIS mixers with
a wideband IF. Obviously, the sensitivities limited by bandwidth and the system noise temperature can be attained only in a proper
receiver design, as, for example, Dicke receiver with sufficiently fast switching, correlation receiver, etc. [23]-[27].
5. ACKNOWLEDGMENTS
The contributions of L. Nguyen of Hughes Research Labs, and J. Webber, A. R. Kerr, S.-K. Pan, N. Bailey, S. Thacker, W.
Lakatosh, W. Wireman, R. Harris of NRAO are gratefully acknowledged.
6. REFERENCES
[1] S. Weinreb, "Low-noise, cooled GASFET amplifiers," IEEE Trans. Microwave Theory Tech., vol. 28, pp. 1041-1054, Oct. 1980.
[2] S. Weinreb et al., "Ultra low-noise, 1.2-1.7 GHz, cooled GASFET amplifiers," IEEE Trans. Microwave Theory Tech., vol. 30, pp. 849-853, June 1982.
[3] P. L. Richards et al., "Quasi-particle heterodyne- mixing in SIS tunnel junctions," Appl. Physics Lett., 34(5), pp. 345-347, 1979.
[4] J. R. Tucker, "Quantum-limited detection in tunnel junction mixers," IEEE J. of Quantum Electron., vol. QE-15, pp. 1234-1258, 1979.
[5] A. R. Kerr and S.-K. Pan, "Some recent developments in the design of SIS mixers," Int. J. Infrared & Millimeter Waves, vol. 11, no. 10, pp. 1169-1187, Oct. 1990.
[6] M. W. Pospieszalski et al., "Millimeter-wave, cryogenically-coolable amplifiers using AlInAs/GaInAs/InP HEMT's," in Proc. 1993 IEEE MTT-S Int. Microwave Symp., Atlanta, GA, pp. 515-518, June 1993.
[7]. M. W. Pospieszalski, "Ultra-low-noise receivers for the 1 to 120 GHz frequency range," (invited) in Proc. of 23rd European Microwave Conf., Madrid, Spain, pp. 73-79, Sept. 1993.
[8] Proc. of 7th Int. Symp. on Space Terahertz Tech., Charlottesville, VA, March 1996.
[9] L. D. Nguyen, L. E. Larson and U. K. Mishra, "Ultra-high-speed modulation doped field effect transistors," in Proc. IEEE, vol. 80, pp. 494-518, April 1992.
10] H. Fukui, "Optimal noise figure of microwave GaAs MESFET's," IEEE Trans. Electron Devices, vol. ED-26, pp. 1032-1037, July 1979.
[11] M. W. Pospieszalski, "Modeling of noise parameters of MESFET's and MODFET's and their frequency and temperature dependence," IEEE Trans. on Microwave Theory and Tech., vol. 37, pp. 1340-1350, Sept. 1989.
[12] M. W. Pospieszalski, "Cryogenically-cooled, HFET amplifiers and receivers: state-of-the-art and future trends," in Proc. 1992 Int. Microwave Symp., Albuquerque, NM, pp. 1369-1372, 1992.
[13] K. H. G. Duh et al., "A super low-noise .1 µm T-gate InAlAs-InGaAs-InP HEMT," IEEE Microwave Guided Wave Lett., vol. 1, pp. 114-116, May 1991.
[14] S. E. Rosenbaum et al., "AlInAs/GaInAs on InP HEMT low-noise MMIC amplifiers," in Proc. 1991 Int. Microwave Symp., pp. 815-819, Boston, MA, June 1991.
[15] P. D. Chow et al., "W-band and D-band low noise amplifiers using 0.1 micron pseudomorphic InAlAs/InGaAs/InP HEMT's," in Proc. 1992 Int. Microwave Symp., pp. 807-810, Albuquerque, NM, June 1992.
[16] P. M. Smith, "Status of InP HEMT technology for microwave receiver application," (invited), in Proc. 1996 Int. Microwave Symp., San Francisco, CA, pp. 5-8, June 1996.
[17] M. W. Pospieszalski et al., "Q- and E-band cryogenically-coolable amplifiers using AlInAs/GaInAs/InP HEMT's," in Proc. 1995 IEEE MTT-S Int. Microwave Symp., Orlando, FL, pp. 1121-1124, May 1995.
[18] M. W. Pospieszalski et al., "Millimeter-wave waveguide-bandwidth cryogenically-coolable InP HEMT amplifiers," Proc. 1997 IEEE MTT-S Int. Microwave Symp., Denver, CO, pp. 2285-1288, June 1997.
[19] C. A. Liechti and G. B. Larrick, "Performance of GaAs MESFET's at low temperature," IEEE Trans. Microwave Theory Tech., vol. 24, pp. 376-381, June 1976.
[20] K. H. G. Duh et al., "Ultra-low-noise, cryogenic, high-electron-mobility transistors," IEEE Trans. Electron Devices, vol. 35, pp. 249-256, March 1988.
[21] K. Joshin et al., "Noise performance at cryogenic temperatures of AlGaAs/InGaAs HEMT's with .15-µm T-shaped WSix gates," IEEE Trans. Electron Devices, vol. 39, pp. 515-519, March 1992.
[22] S. Weinreb et al., "Cryogenic, HEMT, low- noise receivers for 1.3 to 43 GHz range," in Proc. 1988 MTT-S Int. Microwave Symp., New York, NY, pp. 945-948, May 1988.
[23] E. J. Wollack et al., "An instrument for investigation of the cosmic microwave background radiation at intermediate angular scales," Ap. J., vol. 476, pp. 440-457, Feb. 1997.
[24] C. Bennet et al., "Microwave anisotropy probe (MAP) MIDEX proposal," NASA Goddard Space Flight Center, Greenbelt, MD, Dec. 1995.
[25] E. J. Wollack, "High-electron-mobility transistor gain stability and its design implications for wide band millimeter-wave
receivers," Rev. Sci. Instrum., vol. 66, pp. 4305-4312, 1995.
[26] N. Jarosik, "Measurement of the low frequency gain fluctuations of a 30 GHz HEMT cryogenic amplifier," IEEE Trans. Microwave Theory Tech., vol. 44, pp. 193-197, Feb. 1996.
[27] E. J. Wollack and M. W. Pospieszalski, "Characteristics of Broadband InP Millimeter-Wave Amplifiers for Radiometry," to be presented at the IEEE Int. Microwave Symp., Baltimore, MD, June 1998.
[28] J. C. Webber, A. R. Kerr, S.-K. Pan and M. W. Pospieszalski, "Receivers for the Millimeter Array," presented at SPIE's Int. Symp. on Astronomical Telescopes & Instrumentation, Kona, Hawaii, March 20-18, 1998.
Fig. 1. A minimum noise measure of a .1 m gate length AlInAs/GaInAs/InP HFET. Experimental results at room temperature from three different laboratories are also shown: "" [13]-[16].
Fig. 2. A comparison of measured gain and noise characteristics of a W-band amplifier with model prediction at room temperature. Measured noise includes the contribution of pyramidal horn and receiver (Tn 2000 K).
Fig. 3. A comparison of measured gain and noise characteristics of a W-band amplifier with model prediction at cryogenic temperature (Ta = 20 K). Measured noise temperature includes the contribution of dewar window, pyramidal horn (Ta = 20 K) and room temperature receiver (Tn 2000 K).
Fig. 4. Comparison of noise temperature of NRAO cryogenic receivers using InP HEMT amplifiers cooled to about 20 K and SIS mixer receivers cooled to 4 K (SIS mixer receiver data courtesy of A. R. Kerr and S.-K. Pan, NRAO).
Fig. 5. The power spectra of the radiometer output. The filled symbol data were taken at a physical temperature of 54 K and the indicated effective RF bandwidths. The open symbol data were taken at a physical temperature
1. An earlier version of this paper was presented at the 1997 Topical Symp. on Millimeter-Wave, Shonan Village, Japan, July 1997.