To measure receiver noise figure, normally you need a calibrated RF noise source. I had been using a cable-TV amplifier with known noise figure and gain. With the input terminated, the output noise level above thermal is the sum of the two values, which was simple enough. But I wasn't sure the level was stable so I decided to build a dedicated noise source.
The circuit reverse-biases a 2N2369 base-emitter junction to generate avalanche noise. Because the noise level varies somewhat with junction current, instead of a load resistor I used a 2N5484 JFET arranged as a constant-current source. The current is equal to IDSS, 3.9 mA for the 2N5484 I selected. 2N2369 VEB is 5.9 V. It remains constant for any supply voltage above 7.2 V. Output impedance is 50Ω. For 75Ω, change the 56Ω resistor to 82Ω.
I built the source on a female N-connector to directly mate with a variable RF attenuator, which wasn't needed for the Y-factor method I finally decided to use. A trimpot adjusts the output level. Originally I had intended to set it for a round ENR value, but I discovered that a setting of 322Ω yielded the same return loss with the power on or off. This makes it unnecessary to substitute a 50Ω load for the source during noise figure tests. I measured a return loss of 29 dB with a power splitter that yields 30 dB for a 50Ω load.
To calculate ENR I measured the output noise level with my HP 141T/
May 20, 201388–108 MHz