A custom spectrum analyzer was built to measure the noise and gain of the amplifiers. This spectrum analyzer first amplifies the signal from the device under test (DUT). A computer controlled local oscillator (LO) selects a frequency to mix down, and a low pass filter sets the resolution. The computer reads the DC voltage off a power detector after it sets the frequency. This way, the computer can step through frequency to read out the spectrum.
The input to the device under test is usually a variable temperature terminator. For room temperature testing, this terminator can be dunked in liquid nitrogen. For cryogenic testing, it can be heated with a power resistor.
A simplified block diagram of the entire setup is shown below.
In the rack shown below, the volt meter is on top, the test box in the middle, and the computer controlled sweep oscillator on bottom.
To calibrate the spectrum analyzer, the output power is measured as a function of frequency for at least two terminator temperatures. A line is fit to the plot (shown below), with the slope giving the gain of the analyzer and the x intercept giving the noise temperature.
The amplifier gain and noise are found the same way as the calibration, except the noise will include some component from the spectrum analyzer, and the gain will be the total gain of the system.
The gain and noise of the bare amplifier are now easy to calculate. The bare amplifier gain is
Gamp = Gtotal / GcalThe amplifier noise is
Tamp = T'amp - (Tcal / Gamp)
I found the huge dynamic range of the spectra being measured to be a challenging aspect of getting this test setup to give accurate measurements. The lowest power measured is for a 4.2 Kelvin calibration load, and the highest for a room temperature 35 dB amplifier. Thus the input dynamic range is about 55 dB. On the one hand, the box needs enough gain to measure the 4.2 Kelvin calibration. Yet it must not compress when measuring a room temperature amplifier. Even relatively small compression would have a noticeable affect on the Y factor. A computer-controlled step attenuator was placed after the mixer, which helps, but isn't perfect. When designing a piece of test equipment such as this, it is crucial to calculate the power at every point in the signal path for both the maximum and minimum input powers, and to make sure that no amplifiers begin to compress, that the power detector remains in a linear range, and that the noise temperature of the test box doesn't change as the variable attenuator is changed.