On Mar 16, 2:01*am, "Pete KE9OA" wrote:
Actually, you do need AGC when using an MC1350. Each stage provides about
60dB of gain at 455kHz. If you don't have some sort of gain control on the
stage, you will drive it into clipping. Been there, done that. My first
attempt at a sync detector was a quasi-sync detector, similar the the
implementation Drake uses with their "Synchro Phase" detector, and similar
to the design used for AM recovery in the Racal 6790.
It used a 1350 as a limiting amplifier that was driven into clipping by the
radio's I.F. strip. This chip squared up the signal for the LO input of a
Philips SA-637 digital FM radio chip. The point of this is that I
intentionally let the MC1350 run at full gain, so that I could run it into
clipping, thus, squaring up the signal.
Now, about that time constant...................true, that 6Hz is the -3dB
corner frequency, but this is fine for attack/decay characteristics. In the
real world, there is no problem with modulation of the AGC bus. Ten times
this time constant would be way to long of a decay time.........we are
talking about 1.66 seconds here. 500 milliseconds would be better for a
release time for AM signals.
A hang AGC is fine for SSB signals, but for AM full carrier signals, it
would be very inconvenient. As far as cascaded AGC, there can be oscillation
problems of the control system if you are not careful when designing it. You
could grab the AGC line from the radio that you would be using, and
scale/translate the levels with a couple of op-amps. The MC1350 requires a
5V knee point at Pin5 that your AGC voltage is superpositioned over. As I
stated before, the AGC system for a 1350 is fairly complex; that is the
reason that I suggested a TDA1572. The 1572 is a much better chip for this
purpose, since it has its own AGC/demodulator system on board. In addition
to that, a few external components can be used for independant attack/decay
parameters. If you don't need SSB demodulation, a TDA1072 has the same
functionallity, minus the buffered 50 Ohm I.F. output.
The 160 millisecond time constant was mentioned in Ulrich Rohde's book,
entitled "Communications Receivers, Principles and Design". I didn't pull
that figure out of a hat.
You do have some good suggestions, Tom.

Pete

"You don't need another AGC if the Redsun AGC is controlling its IF
output, do you? Cascaded AGC's might be a problem. Maybe the Redsun
AGC is poor and you will want to replace it with an outboard one. AGC
can cause audio distortion if it is fast enough to chase the
modulation. Pete's mention of 160mS AGC must be referring to the
attack time constant or speed. A release time constant of about 10
times that is advisable to keep modulation tracking very low. If it
was symmetrical 160ms attack/release, then it would do a very good job
tracking 6Hz modulation (a rarity) but would still have a 10%
sensitivity to 60Hz and 1% at 600Hz. Then there is the concept of
'hang' AGC, where the release is delayed, maintaining constant gain
through the delay unless a greater attack is enountered that requires
a gain reduction and a new 'hang' delay. This is particularly
effective at avoiding modulation tracking if the hang time is equal or
greater than the attack time and allows a faster release without
jeopardising modulation tracking distortion.

I don't know whether there is any big advantage between audio derived
and IF derived AGC as long as there is modulation. You mention talk
radio so I assume that is DSBAM with carrier. The IF derived AGC will
output an AGC voltage affected by carrier strength and to some extent
by modulation so it will maintain an appropriate gain even during long
pauses in modulation. An audio derived AGC would increase IF gain
during silence and thus might permit peak distortion or clipping of
the next audio, especially if it begins with a plosive. Were this
suppressed carrier AM, then both would increase gain during pauses.

You would probably get more informed opinion than I can offer if you
were to post to rec.radio.amateur.homebrew.

Good luck with your project!"

Tom

I agree with Telamon about the two release speeds for speech
programming: the long release for fairly stable signals and the
shorter for rapid fading. The longer one gives lower distortion due to
modulation tracking. Under really stable conditions (local
groundwave), lowest distortion would be obtained by turning AGC off
and setting the RF Gain manually.

Pete, given that the Redsun already has AGC, why couldn't the 1350 be
manually set to an appropriate gain to match the controlled level
coming out of the tuner?

I have done a fair bit of investigation and modification of AGC
circuits in the Radio Shack DX-394 and the Yaesu FT-817 and been
surprised by the experience. Did a lot of research on published specs
or test reports for high-end radios and other articles. There seemed
to be a rough consensus about release speeds corresponding roughly to
what Telamon cites for speech/music; for interrupted CW and for data
modes, below 25ms seemed to be the desired speed. What was more
confusing or contentious is the attack - some radios citing sub 1ms!
Moreover, good audio compressor/limiters also have very fast attack to
prevent peak clipping and broadcast Peak Program Meters integrate over
10ms, effectively ignoring peaks very much shorter in duration as
these are psychoacoustically inaudible. I modified the AGC for the
DX-394 with sub 10ms attack for all release speeds and then was
dismayed with the effect of impulse noise - long holes in audible
modulation or even continuous suppression if the repetition rate was
high enough. Unless there is an IF noise blanker ahead of the AGC
detector, an attack that is proportional to the release is necessary,
something on the order of 5-10%.

With the FT-817, I found that the effective AGC speed is inversely
related to the RG Gain control setting and is very much faster than
the RC time constant would suggest. Pete, you mention a 'knee' with
the 1350 around 5V - I think the FT-817 has a knee around 1V and
follows a square law characteristic such that most gain variation
occurs over maybe a 100mV slice out of the 3V range over which the AGC
and RF Gain control swing. Hence, I have been unable to eliminate what
I think is quite objectionable distortion due to modulation tracking,
even with a release RC time constant on the order of 2.5s and an
attack RC time constant pushed up from a few ms to over 100ms.
Modulation wiggling the AGC by a millivolt distorts the IF envelope
delivered to the AM detector. Listening to DSBAM in ECSS mode (SSB)
significantly reduces distortion because the filter skirt suppresses
bass energy.

Tom