[Collins] Collins 62S-1 Transverter Downconversion Gain??

[Dr. Gerald N. Johnson]

On 9/21/2010 7:22 PM, Carl wrote:
>
><SNIP>
>
> The 8970 doesnt have to work that way to my knowledge but I'd have to
> drag the manual out of the file cabinet to be sure. In the case of the
> A4 I wanted the true system NF and that included taking measurements
> thru the mechanical filters. It was calibrated at the bandwidth of the
> image reject filter following the external mixer which should be
> sufficient. At HF I wont loose sleep over a dB anyway.

My preamp lost 4 dB gain and a couple dB NF. There is no calibration 
that works correctly for the 8970 with narrower devices under test than 
its 4 MHz IFs. Yes it will calibrate, but it gives wrong answers because 
of its assumption of the bandwidth of the noise contributions. There is 
no other way to get correct NF other than the oldest fashioned 3 dB rise 
or y factor with everything narrow band. And noise in a 2 kHz filter has 
to be averaged a long time to get a steady value, plus any common meter 
that is peak reading calibrated in RMS is driven batty by narrow band 
noise. The next time I do that I plan to use a true RMS meter that works 
by heating a resistor with the input and reading the temperature rise. 
Truly true RMS, not affected by crest factor unless the input amplifier 
gets over driven on noise peaks. But I'll have to watch the meter for a 
minute at a time eyeballing an average value, unless I hook a much 
larger capacitor in parallel with the meter to do that averaging. Then 
I'll just have to watch for the meter to stop moving as it oozes to the 
reading in a slow exponential curve with an ever declining slope.
>
> I got rid of that 5722 setup but still have a NIB tube but Id be more
> likely to use SS these days if I wanted to return to that method.

The beauty of the 5722 is that noise is directly known from the value of 
the terminating resistor and the DC plate current. It needs no 
calibration. Its predictable. The rub of anything solid state is that 
while simpler to use, the noise is not predictable and is not a function 
of the diode current. Raising the current may raise or lower the noise, 
so it has to be measured and calibrated at a constant level. It has the 
best bandwidth, especially if a small geometry device is used like the 
base-emitter junction of a microwave bipolar transistor. Both sources 
tend to change the noise head output impedance between on and off and 
that impedance change can change the gain of a typically mismatched 
MOSFET LNA which violates one of the tenants of NF measuring, presuming 
the LNA gain is the same noise on or noise off. There is were the good 
solid state noise head has a 15 to 30 dB attenuator good from DC to 10s 
of GHz to keep the LNA from noticing the changes in impedance. This 
isn't a problem at HF or noise figures over a few dB.
>
>
73, Jerry, K0CQ, Technical Adviser to the Collins Radio Association.