System temperature (T_sys)

A radio telescope never measures a signal by itself. It measures signal plus noise, and almost all of what it measures is noise. System temperature is the total of that noise, in kelvin, referred to the antenna terminals. The signal you’re after, a line or a source, shows up as a small addition on top of it, and the radiometer equation says how long you need to integrate to see that addition.

What it’s made of

• Receiver noise, T_rx. From the Friis formula: the LNA, plus small contributions from everything after it. With an LNA at the feed this is typically the largest single term for a small station.
• Sky. The cosmic microwave background contributes about 2.7 K everywhere. At 1420 MHz the Galaxy’s own continuum adds a few kelvin away from the plane and more toward it. The atmosphere adds a couple of kelvin near the zenith and more at low elevation. This is the bottom of the quiet window described in water hole.
• Spillover and sidelobes. Any part of the feed pattern that misses the dish, or any sidelobe that lands on the ground, sees something close to 290 K. Even a few percent of the beam on the ground adds tens of kelvin. A good f/D match and pointing above about 20° keep it down.
• The source. When you point at the Galactic plane, the continuum and the line add their own antenna temperature. For bright HI that’s not negligible.

Worked example

A receiver of 95 K (the LNA-at-the-feed chain worked in Friis formula), plus an assumed 20 K for sky and spillover, gives T_sys ≈ 115 K off the plane. These are example values: your own spillover depends on your feed and elevation.

For a 1.8 m dish with an effective area of 1.40 m², that system has a system equivalent flux density of 2k × 115 / 1.40 ≈ 227,000 Jy. SEFD is the flux density of a point source that would double the system noise, a single number for comparing whole telescopes. It’s why small dishes stick to bright, extended targets.

Lowering it

In order of cost-effectiveness: move the LNA to the feed and get every connector out from in front of it (LNA); stop the feed spilling past the rim; observe higher above the horizon; then buy a better LNA. A tenth of a dB of noise figure is a few kelvin, while a badly placed cable can be hundreds. The receive-chain noise calculator shows each term separately.