I've been playing with trying to make rough noise figure measurements
on the cheap and have a couple of questions:

(1) Are there any issues with the following setup and procedure for
making relative noise figure measurements (e.g. comparing two
receivers) ?

____________ __________ ________
| | | | | |
|Uncalibrated| | Step | |Receiver| Audio Out
| Noise |----|Attenuator|------| Under |-------*----o (0 dB)
| Source | | | | Test | _|_
|____________| |__________| |________| | |
| | 2.7k
|_|
|
*----o (-3 dB)
_|_
| |
| | 6.65k
|_|
_|_
///
Procedu
(a) For receiver 1 connect a high impedance AC voltmeter to the audio
output marked "0 dB". Record voltage with noise source off.
(b) Turn noise source on and measure AC voltage at "-3 dB" output.
Adjust step attenuator to get same voltage as in step (a). Record
step attenuator setting.
(c) Repeat steps (a) and (b) for receiver 2.
(d) The difference in noise figure between the two receivers is the
same as the difference in attenuator settings recorded in (b) and (c).
For example if the attenuation for receiver 1 is 10 dB and for
receiver 2 is 12 dB, then receiver two has a noise figure which is 2
dB less than that of receiver 1.

Assuming this is OK we move on to question 2:

(2) To avoid the expense of a calibrated noise source, I wonder if the
repeatability from unit to unit of simple low noise amplifier circuits
(perhaps a MAR-6 ?) is good enough to allow one to be used as a noise
figure standard, at least for fairly rough measurements at HF and VHF.
I am hoping that accuracies of +/- 1 to 2 dB might be achievable.
Has anyone measured the NF performance of simple MMIC amps at HF & VHF
? Or looked into noise figure repeatability ?

73,
Steve VE3SMA