SS Amplifier with automatic gain control

[tca]

I've been playing again with SS amplifiers and was thinking in including some kind of automatic gain control to prevent the typical harsh clipping. Something like this (from Mullard's book: Transistor audio and radio circuits)



I've read here that the simple audio compressors, for instance for speech, will not work as expected for a guitar player, but, nevertheless, what are your thoughts about this approach. I've simulated and it seems to work nicely (with some different values of the components). I'll breadboard it tomorrow to see how it sounds.

Any experience with automatic gain control circuits? Any schematic that I should read?

Thanks.

[PRR]

> Any experience with automatic gain control circuits?

"Sophisticated— God, I'm sophisticated!"

Find a Schematics Dungeon and study all the Peavey transistor guitar amps. PV has done OK with overload limiters that don't suck.

-------------------------

An issue with the plan you show (aside from being more complicated than it needs to be, yet not enough parts to be good) is that the signal at "Signal V2" can't be much over 20mV. Since guitar is 20mV to over 200mV, that means you have to make a weak signal weaker then build it up again. Like going through a mud-hole on the way to the car-wash.

A detail is that R5 C4 define a 380mS (really longer) attack time, which is a heck of a long time. When you start adjusting times and flatness, you find you probably need a lot of amplifier gain, stages, or beef.

There's multiple musician-approved limiters. The Ross. The Merlin Thumb. The RNC is a heck of a box for the bucks.

[tca]

"Sophisticated— God, I'm sophisticated!"

You know me!

BTW Mullard's circuit  is taken from a tape recorder amplifier and I was thinking in using it at the output of a 1W amplifier. Thanks for the ref on the Peavey.

[tca]

Just breadboard a version of this little AGC... it actually works. I'm using  it in a 1W amplifier (18v) and with this AGC it can take an input of 5V and maintain a good OD sound, not harsh at all.

My values are:

R1 = 22k
R3 = 100k
R4 = 1M
R5 = R6 = 1k
C1 = 220n
C2 = 1u
C4 = 1u

Time for some more testing.

[R.G.]

There's a whole range of these. Do a search on "soft clipping" in hifi amps. The idea is to introduce some kind of limiting ahead of the power amp with a clipping or limiting characteristic you can control, versus the power amp's clipping which you can't - very well at least.

One of the earlier examples I know of is in the Thomas Vox amplifiers. They used a variable-threshold clipper ahead of the power amp, using diodes with a variable offset and some resistance in series with the diodes. I suspect that they did this because the power amplifiers they used would self-destruct if driven into clipping for too long, so they clipped ahead of the power amps, not in the power amps, and kept the power amps linear. But it sounds much better than a typical high negative feedback power amp.

If you're doing a version of limiting that clips the signal instead of hard-compressing it after some threshold, you might look into a simple diode-resistor or transistor-resistor ladder that clips progressively harder and then cuts in more and more aggressive clipping, the last stage being a hard limit at less than the clipping threshold of the power amp. I've done a couple of these, and they work well. One possibility is to simply use a pair of diode-connected MOSFETs as a clipping pair ahead of the power amp, with suitable scaling for driving the MOSFETs to clipping and then making the power amp gain such that the full clipping level at the MOSFETs is just before the power amp would clip on its own, so it's always linear.

A real non-distorting limiter that goes into infinite compression at some point would work and not have distortion products, but would feel like the volume knob was just non-responsive after a certain point.

[tca]

> One of the earlier examples I know of is in the Thomas Vox amplifiers.
Yep, one of your favorites!

> If you're doing a version of limiting that clips the signal instead of hard-compressing it after some threshold...
I'm hard-compressing it after some threshold. It actually compresses the guitar signal and it sounds good with my strat. Although some more compression would be nice. It feels good playing with a amp that behaves like this, it compress and distort simultaneously It is different from other SS amps I've build and it is very dynamical, very responsive.

> One possibility is to simply use a pair of diode-connected MOSFETs as a clipping pair ahead of the power amp, with suitable scaling for driving the MOSFETs to clipping and then making the power amp gain such that the full clipping level at the MOSFETs is just before the power amp would clip on its own, so it's always linear.

That is also a good idea but the idea of  hard-compressing it after some threshold gives some inter modulation distortion that sounds very nice.

[R.G.]

KEwl. The only rules in this game are "If it sounds good, do it!"

[Quackzed]

 

you might look into a simple diode-resistor or transistor-resistor ladder that clips progressively harder and then cuts in more and more aggressive clipping, the last stage being a hard limit at less than the clipping threshold of the power amp. I've done a couple of these, and they work well.

agreed! i've played with this idea quite a bit and with a decent voltage swing 18v say, you can get a nice soft clipping just before the rails.
generally you want to avoid any clipping for smallish signals and then ,as rg says a progressively harder clipping at each rung of the ladder with a last 'hard'ish clip just before the rails to prevent rail clipping. you also can treat the positive and negative swings separately as long as you stay within the rails at the last respective rung. can be a bit grainy if overdone -intermodulation dist- but retains dynamics while producing greater harmonic content at higher signal levels. sounds as well as feels great! also diode thresholds should optimally be progressively lower and lower from first rung to last as each rung is compressing the signal a bit and makes the next threshold (rung) that much harder to reach... you can really tailor the compression/distortion curves to taste after a bit of experimentation. not to mention cutting high end hash progressively!   

[teemuk]

Thomas Vox amplifiers. ... They used a variable-threshold clipper ahead of the power amp, using diodes with a variable offset and some resistance in series with the diodes. I suspect that they did this because the power amplifiers they used would self-destruct if driven into clipping for too long, so they clipped ahead of the power amps, not in the power amps, and kept the power amps linear.

Well, their patent is kinda interesting reading. It does not address any "self-destruction" issues but it does mention the prior art of using -hard- clippers before the power amp section. (Probably because at the time it was already known that many "old school" and simple power amps may not behave all too well when driven to clipping and may introduce oscillations, hysteresis, rail latching, etc. And yes, in most extreme cases they might even self-descruct due to those issues).

The patent actually implies that power amp's clipping is unharmful (like it is in many amps but probably not in Thomas Organ amps) but makes a point about resultant tone output being "badly distorted". The patent specifically addresses using -soft- clipping limiter before the power amp instead of hard clipping one, which is implied to produce "sharp-cornered, fiat-topped wave into the amplifier, which contains the same undesirable high harmonic distortion content that results when the amplifier itself does the clipping."

So, basically the point and the invention covered by the patent is specifically employing a -soft- clipping limiter before the power amp which "minimizes the generation of these deleterious high-frequency harmonics by rounding off the corners of the wave as it is limited or clipped."


---

Anyway, since most clipping limiters and limiters in general have "fixed" threshold of operation they do not address the issue that when power amp gets overdriven the current draw is usually high enough to sag the power supply voltage, which further on reduces clipping threshold of the power amp. At fixed level the limiter circuit may not track this and the soft clipped output from the limiter still gets hard clipped by the power amp due to sagged rail voltage.

Instead of using a fixed reference for clipping diodes I have experimented with variable DC reference that can be provided, for example, by an envelope follower circuit. Basically, I rectify the signal voltage and extract that from fixed DC voltage. When I do this, reference voltage to clipping circuit is reduced whenever signal amplitude increases and thus basically higher signal amplitudes drive limiter to operate earlier. It's also a tad more "touch sensitive" and "dynamic" solution (if you like such catchphrases) than just using a fixed value of clipping threshold.

[tca]

> Well, their patent is kinda interesting reading.

Have to check it, thanks for the ref.

> Anyway, since most clipping limiters and limiters in general have "fixed" threshold of operation they do not address the issue that when power amp gets overdriven the current draw is usually high enough to sag the power supply voltage, which further on reduces clipping threshold of the power amp. At fixed level the limiter circuit may not track this and the soft clipped output from the limiter still gets hard clipped by the power amp due to sagged rail voltage.

My amp is a .5W (rms) class A amplifier so no sag from the power supply: 18V, .6A quiescent current and output impedance of 20Ohm. It makes a lot of noise!


> ...have experimented with variable DC reference...

Thought of that too!

[R.G.]

Didn't find the patent number in the posting, so I went looking. It's 3509373.

Interestingly, it's assigned to Warwick Electronics Inc.; I don't know (yet  ) how Warwick relates to Thomas Organ Vox. Whirlpool Corp is mentioned, and they were the last holders of Thomas Vox stuff, including the IP, it seems. The award in 1967 would have expired in 87.

It does work the way I thought it did - which always makes me feel better!

The circuit looks to be the same, so that's where Thomas got it. I think there was some degree of separation between the patent originators/holders and Thomas. The patent doesn't mention preventing melt-down in the power amplifiers, sure enough. However, I know for a fact that some Thomas Vox amplifiers will die stone cold dead if overdriven and not limited by the limiter. It may have been good fortune that a limiter that made the amps sound better also helped prevent destruction. It may also be that the circuits were used for prevention of destruction and also happened to make them sound better. I picked up on the latter, as my focus has been on trying to keep the blasted things running, but it could have been the other way round, too.

In either case, it is a nice effect. The diode/resistor ladder approach I did worked, too, but it needed about a 3-4V peak signal to get enough signal to do the variable thresholds where they needed to be with silicon diodes. I then adjusted the power amp to need that big a signal for full power. A divider would work for a normal-sensitivity power amp.

In re the sag of power supplies under heavy loading, one could do a variable threshold, etc., but I believe that it's simpler to just set the fixed threshold below where the power supply sags to, for two reasons:
(1) the amplifier is always going to be putting out a lot of power at those input levels, so it's going to be sagged to some extent anyway
(2) even with light loads, the difference in output voltage lost with a fixed threshold below the sagged power supplies loses you very little in perceptible output loudness.

We probably ought to quit talking about it, or Randall will go re-patent it. 

[R.G.]

Interestingly, it's assigned to Warwick Electronics Inc.; I don't know (yet  ) how Warwick relates to Thomas Organ Vox.

Never mind. Warwick apparently owned Thomas Organ at the time.

[PRR]

> The patent doesn't mention preventing melt-down in the power amplifiers, sure enough.

You would be wary of writing that up in a patent. Aside from suggesting your product is fragile, a patent is supposed to have _broad_ public use. If only Thomas' amps were frying, then who would else need it? (Yes, a lot of amps fried, but the industry was in de Nile.)

> It's 3509373.

Looks a lot like stuff widely published, for example in ham-radio mags as hash-clippers. Or sine function generators.

Only "novel" claim would be use in a musical instrument. Even that can only be a matter of days-months priority.

They _do_ give some possible circuit values. Cleverly they do NOT give the node impedance, so we can't really guess the harshness.



BTW: these are "instantaneous limiters". Distorters. tca's post was about "distortion-free" time-enveloped gain-control limiters.

[tca]

> ... was about "distortion-free" time-enveloped gain-control limiters.

In a certain way this AGC is more like a tape saturation phenomenon (VM1=input, VM2=output) than clipping.

[samhay]

^Just breadboard a version of this little AGC.
Any chance you could post the schematic you have been using?

[tca]

^ Here you go!



P.S. (edit)

The power amp is inspired by the Dead of Zen amplifier by Rod Elliot.

[samhay]

Thanks - will add it to my list of things to have a play with.

[teemuk]

BTW: these are "instantaneous limiters". Distorters. tca's post was about "distortion-free" time-enveloped gain-control limiters.

If "distortion-free" is the goal then there's very little point in mediocre late 1970's / early 1980's designs that are mainly aimed for applications like recording computer programs to tapes or broadcasting ... in which retaining constant peak value is more important than low distortion (which you will NEVER get from OP's circuit example).

Today we have many excellent integrated solutions that - distortion spec wise - will run circles over old school circuits like this. THAT Corp. comes to mind as one good option.


Rod Elliot's pages have a very through article about voltage controlled amplifiers, which is good reading.... in fact, I'd regard it as mandatory reading material for the topic. (sound.westhost.com/articles/vca-techniques.html)

[tca]

> If "distortion-free" is the goal then there's very little point in mediocre late 1970's / early 1980's designs ...
"Distortion-free" is not the goal, actually that particular design, from Mullard's book, is from 1969, probably earlier, but not as old as some tube amps! The idea is to distort the signal, but in particular taylor-made way.

[R.G.]

> The patent doesn't mention preventing melt-down in the power amplifiers, sure enough.
You would be wary of writing that up in a patent. Aside from suggesting your product is fragile, a patent is supposed to have _broad_ public use. If only Thomas' amps were frying, then who would else need it? (Yes, a lot of amps fried, but the industry was in de Nile.)

It's fascinating to read through the patents of musical effects stuff. For the life of me, I can't get over how trivial a lot of them are. In looking for this one, I stumbled on one patent for using diodes in the feedback path of an amplifier to cause distortion - filed in the mid 80s! Feedback clippers were standard items long before that. I can't fathom how this got through, other than perhaps the patent examiners were completely clueless about the music industry.

Peavey patented a variable resistor in series with clipping diodes to soften the clipping in the 90s, the uniqueness apparently being using a dual-section pot to mitigate the change in apparent volume that causes. I almost can't imagine something more "obvious to one skilled in the art". Well, OK, using two resistors as a voltage divider, or a resistor to limit or sense current, maybe. But still.

> It's 3509373.
Looks a lot like stuff widely published, for example in ham-radio mags as hash-clippers. Or sine function generators.
...
They _do_ give some possible circuit values. Cleverly they do NOT give the node impedance, so we can't really guess the harshness.

That resistor is ... drum roll, please ...    10k.  Ta-da! It's in all the Vox service lit.