Since the RF envelope and the audio envelope in SSB do not correspond,
audio derived AGC tends to produce a smoother audio output. (Its
analogous to audio clipping versus SSB RF clipping, where the flat tops
of the audio clipping lead to spikes in the RF envelope, exactly the
opposite of that desired for best use of the transmitter amplifier
stages). Deriving the AGC at audio and controlling mostly gain stages
before the crystal filter and detector does lead to a problem at the
attack because there's a bit of signal past the control stages when the
AGC arrives at the control stages. Part of it is time delay in the
crystal filter, a little is time delay in the gain stages, some is in
the post detection low pass filter. Plus the fact that the AGC tends to
work on peaks and the fastest audio half cycle is going to be 0.2
millisecond from zero to peak, (2.5 KHz) but the average audio half
cycle is going to be at least 5 times longer, a millisecond. Leads to
clicks on the attack side of the incoming signals.
AGC derived from a wide IF while a narrow AF filter is in use often
leads to odd results as unwanted signals in the IF but not passing
through the narrow AF filter cause the desired signal to bounce about in
amplitude. So having the AF selectivity inside the AGC loop is a
definite benefit, except that the AF selectivity adds a heap more time
delay.
Probably there really needs to be at least a couple independent AGC
control circuits, not necessarily all fedback loops. There needs to be
an IF derived (envelope) AGC to control the peak level to the product
detector. Then there needs to be an AF derived AGC controlling audio
gain to remove the zits of the IF derived AGC. Most of the time AF
derived AGC tends to be slow because of the fact that the loop filter
has to not pass audio, only envelope information, or it can easily
oscillate. Perhaps the audio control should not be a feedback loop but
something accomplished in DSP where the audio gain is controlled by the
input level to the audio DSP. There the control bandwidth can be
independent of the audio bandwidth and their spectra can overlap though
an excessively aggressive control can lead to artifacts in the audio.
The audio derived hang AGC typical of the Corsairs is what had the most
publicity in amateur radio technical journals as SSB equipment was
developed. The simple IF AGC circuits were just that, simple, and with a
tendency to be driven by unwanted signals, not the desired signal. The
audio derived hang AGC circuits are more complex and do hold the gain
constant (and low) after being hit but a static pulse. Perhaps the time
delay is sometimes too long for such and event, but more rapid recovery
from audio peaks of the desired voice can be considered to be excessive
distortion of the desired audio. With simple feedback and simple RC
timing, its difficult to satisfy all radio users.
AGC for AM and for a single CW signal in the pass band (where hang is
nice to not bring up noise during key up periods) is easy and IF derived
works super. Audio derived AGC for AM would not be good because it would
remove the nuances of the modulation signal, but there's that nice
carrier at the IF to create a super AGC.
73, Jerry, K0CQ
--
Entire content copyright Dr. Gerald N. Johnson. Reproduction by
permission only.
|