Cryogenic Termination
How a Cooled Reference Load Sets the Noise Floor
In any cryogenic receiver, the lowest noise the system can ever reach is bounded by the physical temperature of whatever the first amplifier looks back into. A cryogenic termination exists to make that source as cold as possible. The available noise power delivered by a matched resistor into a bandwidth B is P = kTB, where k is the Boltzmann constant of 1.381 × 10−23 J/K. At 290 K a 50-ohm load delivers about −174 dBm/Hz; cooling the same load to 4 K drops that to roughly −192 dBm/Hz, an 18 dB reduction in radiated noise. That difference is exactly why radio astronomy front ends, dark-matter axion searches, and quantum processors push their reference loads down to the 4 K stage and below.
Getting the resistive film genuinely cold is harder than bolting the connector body to a cold plate. At low temperature the electron-phonon coupling that carries heat out of the resistive layer weakens sharply, scaling roughly as T5, so a chip resistor dissipating even microwatts of residual RF power can float its electron temperature tens of millikelvin above its substrate. Designers mitigate this with large thermalization pads, gold-plated copper bodies, generous bolt torque on indium or annealed-copper gaskets, and by limiting incident power. The result is judged not by the plate temperature but by the equivalent noise temperature the load actually presents to the first stage.
The second job of a cryogenic termination is impedance integrity across a 250 K temperature swing. Thin-film resistor alloys such as tantalum nitride and nichrome are chosen for their small temperature coefficient of resistance, so a load trimmed to 50.0 ohms at room temperature typically stays within a fraction of an ohm at 4 K. The practical limits on match come instead from differential thermal contraction at the connector interface and from the dielectric-constant shift of the substrate, both of which a careful designer manages with matched-CTE materials and a cryogenic calibration of the test fixture.
Governing Relationships
Pn = k T B → Pn/B (dBm/Hz) ≈ −198.6 + 10·log10(T)
Reflected Power from Return Loss:
|Γ| = 10(−RL/20), VSWR = (1 + |Γ|) / (1 − |Γ|)
Noise of a Lossy Cold Component (loss L > 1):
Te = (L − 1) × Tphys
Where k = 1.381 × 10−23 J/K, T = physical temperature in K, B = bandwidth in Hz, Γ = reflection coefficient, RL = return loss in dB. Example: at T = 4 K, Pn/B ≈ −192 dBm/Hz; at RL = 20 dB, |Γ| = 0.1 and VSWR ≈ 1.22.
Termination Performance by Cold Stage
| Cold Stage | Typical Temp | Equiv. Noise (kTB) | Return Loss (DC-18 GHz) | Typical Use |
|---|---|---|---|---|
| Room reference | 290 K | −174 dBm/Hz | > 25 dB | Bench calibration hot load |
| Cryocooler 1st stage | 40 to 70 K | ≈ −182 dBm/Hz | > 22 dB | Line thermalization, spare ports |
| Cryocooler 2nd stage | 4 K | ≈ −192 dBm/Hz | > 20 dB | Radiometer cold load, LNA input |
| Still / cold plate | 100 to 800 mK | ≈ −204 dBm/Hz | > 18 dB | Detector array reference |
| Mixing chamber | 10 to 20 mK | < −215 dBm/Hz | > 15 dB | Qubit readout source impedance |
Frequently Asked Questions
How cold does a cryogenic termination actually get, and why does that matter?
The resistive element heat-sinks to a cold stage: the 4 K plate of a pulse-tube or Gifford-McMahon cooler, or the still, cold, or mixing-chamber plate of a dilution fridge at 800 mK, 100 mK, or 10 mK. What counts is the temperature of the resistive film itself, not the plate it bolts to. Weak electron-phonon coupling at low temperature lets a poorly thermalized resistor float tens of millikelvin above its mount. Since available noise power is kTB, every residual kelvin adds noise; at 4 K the load delivers about 4 K of equivalent noise, while a well-thermalized millikelvin load approaches the floor qubit readout needs.
Why not just use a room-temperature 50-ohm load and a long cable into the cryostat?
A 290 K load radiates about 290 K of noise back up the line, roughly 70 times the noise of a 4 K load, and that noise swamps the cold LNA or qubit. The long cable also adds insertion loss whose physical temperature appears as extra noise through the loss-times-temperature product, plus thermal gradients and microwave resonances. The whole purpose of cryogenic measurement is to lower the source temperature seen at the device, so the reference load itself must be cold and thermalized directly at the cold stage.
How do return loss and impedance match change when a termination is cooled to 4 K?
Tantalum nitride and nichrome films have a small temperature coefficient of resistance, so a load trimmed to 50.0 ohms at 295 K usually stays within a fraction of an ohm at 4 K, shifting return loss by only 1 to 3 dB. Good parts hold better than 20 dB return loss (VSWR under 1.22) to 18 GHz and 15 dB through 40 GHz. The real cold-temperature limits are differential thermal contraction at the connector and substrate dielectric shift, and measuring it needs a cryogenic VNA calibration because the cables also contract and change loss.
What is the difference between a cryogenic termination and a cryogenic attenuator?
A termination is a one-port device that absorbs incident power and reflects almost none; it is the cold reference load at the end of a line. A cryogenic attenuator is a two-port part, typically 3, 6, 10, or 20 dB, that sits inline to cut signal level and to thermalize a coaxial line at each stage. Both use thin-film resistive technology and both must be heat-sunk, but the attenuator sets a controlled cold noise temperature on a through path while the termination defines the end-of-line reference. Qubit input lines chain attenuators at successive stages and use terminations to cap unused ports.