Cryogenic Systems

Cryogenic Noise Temperature

Q: How is cryogenic noise temperature measured?

The standard technique is the Y-factor method using a calibrated noise source or two loads at known physical temperatures. The amplifier sees a hot load (for example a matched termination at 295 K or a calibrated diode) and a cold load (a termination immersed in liquid nitrogen at 77 K or liquid helium at 4 K). The ratio of measured output powers is Y, and the equivalent noise temperature is Te = (Thot - Y times Tcold) divided by (Y - 1). At cryogenic levels the load-temperature uncertainty and connector loss dominate the error budget, so cryogenic noise calibration often uses a cooled matched load inside the same Dewar as the device under test, with thermometry traceable to within plus or minus 0.1 K. Some labs use the cold-attenuator method, placing a temperature-controlled attenuator ahead of the amplifier to present a precisely known noise temperature.

Q: Why does cooling an amplifier lower its noise temperature?

Two mechanisms combine. First, the thermal (Johnson-Nyquist) noise generated by the lossy elements ahead of and inside the transistor scales directly with physical temperature, so cooling a 290 K resistive loss to 15 K cuts that contribution by roughly 20 times. Second, in a HEMT the channel electron mobility rises sharply as the lattice cools, increasing transconductance and reducing the device gate and drain thermal noise. A GaAs or InP HEMT LNA with a 35 K noise temperature at room temperature can fall to 4 to 8 K when cooled to 12 to 15 K. The improvement saturates near 10 to 15 K because residual hot-electron noise and intervalley scattering no longer decrease, which is why most cryogenic LNAs target the 4 to 20 K stage of a closed-cycle cryocooler rather than millikelvin temperatures.

Q: What noise temperature is achievable for a cryogenic LNA at X-band?

State-of-the-art InP HEMT low-noise amplifiers reach 3 to 6 K equivalent input noise temperature across roughly 8 to 12 GHz when cooled to 12 to 15 K, corresponding to a noise figure of about 0.04 to 0.09 dB. At 4 K bath temperature the very best devices have demonstrated below 2 K over narrow bands, approaching the quantum noise limit of h times f divided by k, which is 0.48 K per GHz (about 4.3 K at 9 GHz when expressed in the standard convention). Production deep-space and radio-astronomy receivers typically specify 4 to 10 K including input flange and bias-tee losses. SIS mixers and parametric amplifiers can go lower still but require narrower bandwidths and more complex bias and pump arrangements.

/kry-uh-JEN-ik noiz TEM-per-uh-chur/

When an RF amplifier is physically cooled to a few kelvin, its equivalent input noise temperature collapses far below the room-temperature value, and this cryogenically reduced figure is what receiver designers specify as the cryogenic noise temperature. Two effects drive it down: the Johnson-Nyquist thermal noise of the input losses scales directly with physical temperature, and the channel mobility of a cooled HEMT rises sharply, boosting transconductance and cutting intrinsic device noise. A practical InP cryogenic LNA that measures roughly 35 K at ambient can reach 4 to 8 K at a 12 to 15 K cold stage, with the best narrowband devices approaching 2 K near 4 K bath temperature. These single-digit noise temperatures are what make radio astronomy arrays, deep-space telemetry links, and quantum-computer readout chains possible.

Category: Cryogenic Systems

Typical T_e: 2 to 10 K

Cold stage: 4 to 20 K

Why Cooling Drives Noise Temperature Toward the Quantum Limit

Equivalent noise temperature T_e expresses the internal noise of a two-port as the physical temperature of a matched resistor that would deliver the same available noise power at the input. For an amplifier dominated by thermal sources, that internal noise tracks physical temperature almost linearly, so a stage of 1 dB input loss that contributes about 75 K at 290 K contributes only about 4 K when cooled to 15 K. This is the central reason flagship receivers immerse their first amplifier inside a closed-cycle cryocooler: the front-end stage sets the system noise floor through the Friis cascade, and every kelvin removed there propagates straight into sensitivity.

The second contribution comes from the active device itself. In a high-electron-mobility transistor, lattice cooling reduces phonon scattering, so carrier mobility and transconductance g_m climb while the gate-induced and drain thermal noise drop. The minimum noise temperature of a HEMT roughly follows T_min ≈ f / f_T × T_phys scaled by a device fitting factor, which explains why the same chip improves by five to ten times between 290 K and 15 K. Below roughly 10 K the benefit flattens, because residual hot-electron and intervalley noise no longer fall, so designers rarely chase millikelvin baths for the LNA itself.

There is also a hard floor. The quantum (zero-point) limit sets the smallest meaningful added noise of a phase-insensitive linear amplifier at T_q = hf / k, equal to 0.048 K per GHz. At 9 GHz this is about 0.43 K, so a 3 to 5 K X-band cryogenic LNA already operates within roughly an order of magnitude of fundamental physics, which is why further gains demand parametric or SIS approaches rather than simply colder HEMTs.

Governing Noise-Temperature Equations

Y-factor noise temperature (hot/cold loads):
T_e = (T_hot − Y × T_cold) / (Y − 1), Y = P_hot / P_cold

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

Friis cascade (front end dominates):
T_sys = T₁ + T₂ / G₁ + T₃ / (G₁G₂) + …

Quantum (zero-point) limit:
T_q = hf / k ≈ 0.048 K × f_GHz

Where Y = power ratio, T_hot/T_cold = load physical temperatures (K), F = noise factor, G = available gain (linear), h = Planck constant, k = Boltzmann constant, f = frequency. Example: HEMT LNA cooled to 15 K with F = 0.07 dB gives T_e ≈ 290 × (10^0.007 − 1) ≈ 4.7 K.

Cryogenic Front-End Noise Performance by Technology

Front-end device	Band	Cold stage	Noise temp T_e	Equivalent NF	Typical use
InP HEMT LNA	4 to 12 GHz	12 to 15 K	3 to 6 K	0.04 to 0.09 dB	Radio astronomy, VLBI
GaAs HEMT LNA	1 to 8 GHz	15 to 20 K	5 to 12 K	0.07 to 0.18 dB	Deep-space telemetry
SIS mixer	100 to 700 GHz	4 K	10 to 50 K	near quantum-limited	Submm astronomy
Parametric amplifier (JPA/TWPA)	4 to 12 GHz	0.01 to 0.1 K	0.2 to 1 K	quantum-limited	Qubit readout
Uncooled HEMT LNA (reference)	4 to 12 GHz	290 K	30 to 60 K	0.4 to 0.8 dB	Terrestrial links

Common Questions

Frequently Asked Questions

How is cryogenic noise temperature measured?

The Y-factor method is standard: the amplifier views a hot load (often 295 K) and a cold load (77 K liquid nitrogen or 4 K liquid helium), giving Y = P_hot / P_cold and T_e = (T_hot − Y × T_cold) / (Y − 1). At cryogenic levels, connector loss and load-temperature uncertainty dominate, so a calibrated cold load is placed inside the same Dewar with thermometry good to about ±0.1 K. The cold-attenuator technique is an alternative.

Why does cooling an amplifier lower its noise temperature?

Two effects stack. The Johnson-Nyquist noise of input losses scales with physical temperature, so cooling a 290 K loss to 15 K cuts that part by roughly twentyfold. Separately, a cooled HEMT gains carrier mobility and transconductance, lowering its intrinsic device noise. A LNA near 35 K at ambient can reach 4 to 8 K at 12 to 15 K. The gain flattens below about 10 K as residual hot-electron noise stops falling.

What noise temperature is achievable for a cryogenic LNA at X-band?

Modern InP HEMT LNAs reach 3 to 6 K from roughly 8 to 12 GHz at a 12 to 15 K stage, an equivalent NF of about 0.04 to 0.09 dB. At a 4 K bath the best narrowband parts dip below 2 K, nearing the quantum limit of 0.048 K per GHz (about 0.43 K at 9 GHz). Fielded receivers usually specify 4 to 10 K once input flange and bias-tee losses are included.

Related Terms

← Cryogenic LNA Cryogenic Switch →