[R-390] Spectrum Analyzer

[2002tii]

Barry wrote:

>how does the AGC affect the bandpass, particularly its shape.  I 
>thought all the AGC did is change the gain at the various stages so 
>the shape would be the same just translated "up" or "down" depending 
>on the signal.

AGC doesn't affect the passband shape, but it affects your VIEW of 
the passband as you sweep through it.  Imagine a perfect AGC with 
infinite slope and a threshold 100 dB below peak output -- as soon as 
your sweep signal entered the edge of the bandpass (-100 dB), the IF 
output would pop all the way to full output and remain there until 
the sweep left the -100 dB bandpass at the other edge.  It would 
appear to you that your IF had a perfectly square shape (flat top 
with an infinite cutoff rate) the width of the IF's -100 dB 
bandwidth.  So, instead of the actual funky-looking, slope-shouldered 
IF response of, say, 4 kHz at the -6 dB points, you would see an 
amazing, perfectly rectangular response say 15 kHz wide (or wherever 
the -100 dB points of your IF filter is).

A real-life AGC won't distort your results to quite that degree, but 
receiver AGCs do have lots of gain and very flat slopes, so they 
distort the apparent passband shape significantly.  Also, their 
response is dynamic (and generally asymmetrical, attack to release) 
so the envelope distortion will depend on how fast you sweep in 
relation to the AGC time constants.

It's just like trying to measure the frequency response of an audio 
amplifier with a speech compressor in the measurement chain.  Bad idea.

If you are puzzled why the AGC does that when you sweep, but not when 
you listen:  Actually, it does do that when you listen, with some 
particular signals.  With phone (or other wide-ish signals), or in a 
noisy band, the AGC responds to whatever is loudest and everything 
else is related to that according to its relative received signal 
strength and the IF passband shape.  But when there is a quiet band 
with just one CW signal (like your sweep tone, which is at just one 
frequency at any particular time), that signal DOES pop up to pretty 
much full output when it is still well down the IF response skirt, 
doesn't change much as you tune through it, and stays at pretty much 
full output until it is well down the IF response skirt on the other side.

Best regards,

Don


Copyright (c) 2011.  Not for redistribution