Cryogenic Systems

Cryogenic LNA

Q: How cold does a cryogenic LNA need to be, and why not just go colder?

Most cryogenic LNAs operate at the 4 K stage of a liquid-helium dewar or a Gifford-McMahon or pulse-tube cryocooler. Modern InP HEMT amplifiers show diminishing returns below about 4 to 6 K: the noise temperature flattens because residual hot-electron noise, gate leakage, and parasitic resistances stop scaling with physical temperature. Cooling further to the 10 to 100 mK range of a dilution refrigerator is done for qubit readout, but the LNA itself usually still sits at 4 K because its DC power dissipation (typically 1 to 10 mW) would overwhelm the millikelvin cooling budget. SiGe HBT LNAs are often run at 15 to 20 K where they reach 10 to 20 K noise temperature with much lower DC power, which suits multi-channel arrays.

Q: Why do HEMTs outperform GaAs and silicon LNAs when cooled?

InP and GaAs HEMTs have very high electron mobility in their two-dimensional electron gas channel, and that mobility rises sharply as the device is cooled, which lowers the equivalent input resistance that sets the minimum noise temperature. The intrinsic transconductance also improves, so gain per stage goes up. Silicon MOSFETs freeze out their carriers and degrade at cryogenic temperatures, while SiGe HBTs improve but plateau higher than InP HEMTs. A good InP HEMT can reach 2 to 5 K noise temperature across 4 to 12 GHz at 4 K physical temperature, versus 20 to 40 K for the same device at room temperature. The trade-off is that HEMTs need careful gate-bias and electrostatic-discharge handling.

Q: How do I measure the noise temperature of a cryogenic LNA accurately?

The standard method is the cold/hot attenuator Y-factor technique. A calibrated input is switched between a load at a known cold temperature (often the 4 K stage itself) and a warmer reference, and the ratio of output powers (the Y-factor) is recorded. The noise temperature follows from Te = (Thot minus Y times Tcold) divided by (Y minus 1). At these levels the dominant uncertainty is the physical temperature of the reference loads and the loss of the input cabling and any cryogenic isolator, each of which adds its own thermal-noise contribution. Practitioners de-embed cable and connector loss carefully and often verify against a second variable-temperature load.

/kry-uh-JEN-ik el-en-AY/

Cooling a low-noise amplifier to 4 to 20 K collapses the thermal noise of its active device, and the result is a cryogenic LNA: a first-stage amplifier whose noise temperature drops from tens of kelvin at room temperature to just a few kelvin when chilled. The dominant technology is the indium-phosphide HEMT, whose electron mobility rises steeply as it cools, with SiGe HBTs used where lower DC dissipation matters. A well-designed InP HEMT LNA reaches 2 to 5 K noise temperature with 30 to 40 dB of gain across 4 to 12 GHz, which is why these amplifiers sit at the receiver front end of every major radio telescope, deep-space ground station, and superconducting-qubit readout chain.

Category: Cryogenic Systems

Physical Temp: 4 to 20 K

Noise Temp: 2 to 5 K (InP HEMT)

Why Cooling a Front-End Amplifier Pays Off

The sensitivity of a receiver is set almost entirely by its first amplifier, because the Friis cascade formula divides every later stage's noise by the gain ahead of it. The noise an amplifier adds is conveniently expressed as an equivalent input noise temperature, and a large part of that figure comes from the thermal agitation of carriers in the transistor channel. That agitation scales with the device's physical temperature, so lowering the junction from 290 K toward 4 K removes most of the thermal contribution and exposes the much smaller intrinsic and hot-electron terms. A room-temperature HEMT that adds 30 K of noise can drop below 5 K once it is bolted to the 4 K plate of a cryostat.

Cooling does more than freeze out thermal noise. The electron mobility of an InP or GaAs HEMT channel rises by roughly a factor of two to three as the device approaches liquid-helium temperature, which lowers the channel resistance that sets the minimum noise temperature and simultaneously raises transconductance and gain. The amplifier also needs less DC bias current to reach a given gain, which matters because every milliwatt dissipated at the 4 K stage must be pumped out by a cryocooler whose efficiency at that temperature is poor. Practical InP LNAs are biased at 1 to 10 mW total to stay within the cooling budget while still delivering 30 dB or more per module.

Silicon devices behave badly when cooled because dopant carriers freeze out, so cryogenic front ends rely on III-V HEMTs or, for lower-power multi-channel arrays, on SiGe heterojunction bipolar transistors. The price of this performance is operational complexity: the amplifier must survive thermal cycling, its gate is extremely sensitive to electrostatic discharge, and the whole chain (often including a cryogenic isolator ahead of it) has to be co-designed so that input cable loss does not squander the low noise figure that cooling provides.

Governing Equations

Cascade noise temperature (Friis):
T_e,sys = T₁ + T₂ / G₁ + T₃ / (G₁G₂) + …

Noise figure to noise temperature:
T_e = (F − 1) × T₀, T₀ = 290 K

Y-factor noise measurement:
T_e = (T_hot − Y × T_cold) / (Y − 1)

Where G_i = stage gain (linear), T_i = stage noise temperature (K), F = noise factor, T₀ = 290 K reference, and Y = ratio of output power for hot vs. cold input loads. Example: an InP HEMT first stage with T₁ ≈ 4 K and G₁ = 35 dB (≈3160×) makes a following 100 K room-temperature stage contribute only ≈0.03 K, so T_e,sys ≈ 4 K.

Device Technology Comparison

Device	Physical Temp	Noise Temp (typ)	Band	DC Power	Best Application
InP HEMT	4 to 6 K	2 to 5 K	4 to 12 GHz	1 to 10 mW	Radio astronomy, qubit readout
GaAs HEMT	10 to 15 K	5 to 15 K	1 to 40 GHz	5 to 20 mW	General cryogenic receivers
SiGe HBT	15 to 20 K	10 to 20 K	0.1 to 10 GHz	0.5 to 5 mW	Multi-channel detector arrays
Uncooled HEMT	290 K	20 to 60 K	1 to 100 GHz	10 to 100 mW	Commercial front ends
Parametric (TWPA)	0.01 to 0.1 K	< 1 K (quantum-limited)	4 to 8 GHz	Pump only	First-stage qubit readout

Common Questions

Frequently Asked Questions

How cold does a cryogenic LNA need to be, and why not just go colder?

Most cryogenic LNAs run at the 4 K stage of a liquid-helium dewar or a Gifford-McMahon or pulse-tube cooler. InP HEMTs show diminishing returns below about 4 to 6 K because residual hot-electron noise, gate leakage, and parasitics stop scaling with temperature. Going to the 10 to 100 mK range of a dilution refrigerator is done for qubit readout, but the LNA itself usually stays at 4 K since its 1 to 10 mW dissipation would swamp the millikelvin cooling budget; SiGe parts are often run at 15 to 20 K for low-power array work.

Why do HEMTs outperform GaAs and silicon LNAs when cooled?

InP and GaAs HEMTs have very high electron mobility in their two-dimensional electron gas, and that mobility rises sharply on cooling, lowering the equivalent input resistance that sets minimum noise temperature while raising gain. Silicon MOSFETs freeze out their carriers and degrade, and SiGe HBTs improve but plateau higher than InP. A good InP HEMT reaches 2 to 5 K noise temperature across 4 to 12 GHz at 4 K, versus 20 to 40 K for the same part at room temperature. The cost is careful gate-bias and electrostatic-discharge handling.

How do I measure the noise temperature of a cryogenic LNA accurately?

The standard approach is the cold/hot Y-factor technique: the input is switched between a load at a known cold temperature (often the 4 K stage) and a warmer reference, and the ratio of output powers gives Y, from which T_e = (T_hot − Y × T_cold) / (Y − 1). The dominant uncertainty is the physical temperature of the reference loads and the loss of input cabling and any isolator, each of which adds its own thermal-noise term, so practitioners de-embed cable and connector loss and cross-check against a variable-temperature load.

Related Terms

← Cryogenic Isolator Cryogenic Noise Temperature →