Tube comp/limiter under development

[rafafredd]

Use two pentodes connected as triodes. There are plenty for ever and ever...

 

[analag]

[quote author="rafafredd"]Use two pentodes connected as triodes. There are plenty for ever and ever...[/quote]

Yes there are a few types of remote cutoff pentodes out there, not just the 6BA6. I like what a mu compressor does to a full mix and as a result I would build a stereo unit. I would probably wind the line out trannies myself to get better DCR balance and so forth. C core comes to mind.
No time soon though.

analag

 

[NewYorkDave]

Hmmm... Interesting. I'm getting the same nonlinear frequency response even with a 6AL5 when a fairly high (~25V) reverse bias is applied. I had blamed this on the 1N4148 but apparently they're not the culprits. In the case of both diode types, the response flattens right out if I keep the back-bias below 15V or so.

Practically, it's no big deal since 0-15V gives me a usable threshold control range, but it is a curious phenomenon.

 

[Larrchild]

Is the control amp output which drives the diodes flat?
I assume it is, but I had to ask.
What about matching impedances? could the bias be changing the load somehow?
Anyway, in the useable range it sounds ok=)

 

[analag]

I notice the same thing in mine and since the control amp is flat, then it must be a funtion of the diodes charging the TC network. Whatever the case may be, this compression curve makes the compressor become a subtle EQ which is propably inherent in all such compressors.

analag

 

[Larrchild]

I notice the same thing in mine and since the control amp is flat, then it must be a funtion of the diodes charging the TC network. Whatever the case may be, this compression curve makes the compressor become a subtle EQ which is propably inherent in all such compressors.

I agree and I'm sure the 1N507's or whatever they use in a 660 are not so flat.

However the 660 uses a minumum amount of prebias on the diodes (4-5v) and sets the ratio in the front of the sidechain amp by waveshaping. I think that is preferable to biasing the diodes to set the ratio.

 

[NewYorkDave]

Actually, the back-bias on the diodes is a threshold adjustment, and the gain of the sidechain amp determines the ratio (or "slope" as they called it in Ye Olden Days). Well, that's more or less the case, anyway: the two controls are more interactive than you'd tend to find in, say, a modern VCA compressor.

 

[Larrchild]

Thats interesting, cause Fairchild calls the sidchain amp gain "ac threshold" and the peak/rms control "dc threshold" but yeah, they interact cause when you only use the peaks to drive the sidechain, you need more gain, cause you have less waveform to work with.

 

[SSLtech]

[quote author="NewYorkDave"]Actually, the back-bias on the diodes is a threshold adjustment, and the gain of the sidechain amp determines the ratio (or "slope" as they called it in Ye Olden Days). Well, that's more or less the case, anyway: the two controls are more interactive than you'd tend to find in, say, a modern VCA compressor.[/quote]
I read it in a very similar way to Larry, with the back-bias basically progressively "cutting off" the lower part of the rectified waveform half-cycles, by "hiding" it behind non-conducting diodes.

Both will affect the "threshold" in terms of affecting the onset relative to a varying input signal level, but once the signal has crossed the onset of comrpession, it multiplies the effect more "suddenly" if the back-bias was the obstacle. For that reason, I see the back-bias as making the 'slope-after-onset' (which I equate to the effective ratio) more severe.

Certainly they'll interact. They do on the Fairchild and the Mercury, and I've no doubt on Larry's own product. I've had a hard time 'educating' users as to the precise effects, but a Neutrik A2, some sweeps folowed buy some nice, varying-dynamic guitar tracks have usually been enough to get the message across.

I personally would label them "AC threshold" (sidechain gain) and DC Threshold (back-bias) or something like "Sidechain gain" and "Peak threshold" for another option...

Labeling is a tough one... It affects user-perception, and that can be 95% of the game.

Keith

 

[Larrchild]

Also, I realize when I say "prebias" and you say "back bias" on the diodes, we are not saying the same thing.

The 660 puts the diodes into conduction at all times so that it doesn't take .6v of sidechain signal to wake them up. The back-bias technique on them you show forces conduction to occur at a higher level determined by settings.

I should think that is throwing away more sidechain amp energy, as it is taking place at higher levels, post-amp.

Try putting a variable (pos) voltage on the 1st sidechain tube's cathode, and move it's operating point up and down to change "DC Threshold".
Remove it's cathode resistor and put a pot there with the wiper connected to a pos voltage. Force the stage into class-B for peak limiting.

 

[SSLtech]

Ah yes. -Isn't that how they do it in the 'Prettyboy' 660... similar end result, less discarded power... I geddit now!

Keef

 

[Larrchild]

Yeah, I'm just the messenger. Thank Rein for that one.

 

[SSLtech]

...Anthough...

In this particular version, the rectification is AFTER the amp (Isn't that the same in the Prettyboy?) and varying the class-B bias of the first stage is a long way before the rectifier... won't that make the unit a) progressively more 'pole-dominant', (i.e. ignoring the the half-cycle that gets chopped) whereas Daves version is rather better-balanced, even if it wastes a bit of power.

-Short of a push-pull sidechain, I don't know... Now i have to dig out the 660 schemo to see how it's done...

Keef

 

[Larrchild]

-Short of a push-pull sidechain, I don't know... Now i have to dig out the 660 schemo to see how it's done...

Ahh good point. Without a push-pull class b circuit, the polarity of the peak thru sidechain amp would be only one one side of waveform. Right?

 

[SSLtech]

Well, either that, or as the 'carrier' background DC level modulates (I'm thinking transistor topology here, I assume some of it carries over to hollow-state) with signal, it can 'sort-of' move the waveform up & down... that might help the negative-going peaks to 'lift' the positive level of the waveforem, but either way it certainly favours one pole...

Either way, I see it as rather less ideal in terms of end result, compared to Dave's iteration. Dave's version certainly favours neither pole, (or favours them bothe equally) and it's hard to do that any other way given this toplogy.

How much power is wasted here? -I'm not tube-savvy enough to have an instinctive grasp. -If it's not a signifigant issue, then I like Dave's version enough to choose that. -If there are compromise issues as a result however, then perhaps it'd be worth looking for improvements.

Keith

 

[Larrchild]

re: wasted power.
Since -15v is all thats needed for full GR, it's probably inconsequential. In a circuit that wants -70v, it's a factor.

I think they will just sound slightly different. I think the diode linearity is more critical in Dave's.

 

[SSLtech]

Okay then.

I just peeped my copy of the 670 schem, I see that it's a push-pull sidechain, so that now makes perfect sense to me.

Will diode linearity primarily affect the 'knee' at onset, or is there more to consider?

-I wonder if Schottky diodes could be made to work in this situation? ...although the back-bias trick may be asking too much of them...

Keith

 

[Larrchild]

Dunno, Keef. I used to take 1N4004's and make varactor diodes out of them for little fm transmitters. I know they have a few pf's of variable C when back-biased, but I think its something else going on too. Dave has it flat within the main GR range, so it mite not be an actual issue here. And yeah, I would think the knee would be different than the 660 setup.

Not sure if the knee is the prob or flat sidechain voltage output vs freq. Probably the latter.

 

[SSLtech]

I used to take 1N4004's and make varactor diodes out of them for little fm transmitters

Hehehe... yeah, my entry into studio madness began when I built a small FM pirate radio station back in the late 1970's... Enforcement in the UK was more lenient than here in the USA...

You know, I'm looking at the DC threshold adjustment and wondering if it isn't a little high-impedance... I know that instantaneously the cap will 'absorb' a lot, but once the cap has 'taken a few punches' in a short time period, it may start to 'drift' from its set position. The source impedance of the detector is definately in the sub-10K range, whereas the wiper on the DC threshold control varies with different settings, and can be in the 10K range... I think the detector may be capable of 'tugging' it a little... and I'de expect that to happen more at higher voltage settings because the impedance gets 'worse' the further away from zero volts it gets.

If this was an active semiconductor circuit, I'd try a few things, but I wonder if Dave could test my suspicions by applying a little low-source-impedance voltage there -from a bench power supply or similar-, and seeing if the results are any more steady at higher threshold voltage settings...

(I'd try it myself, but I'm just at the 'gathering parts' stage for an exploratory build)

Keith

 

[NewYorkDave]

Sounds like we three have something in common. I've designed and built a low-power frequency-agile "3-meter exciter" myself. (Discrete transistors and hand-wound coils). I used proper varactors in that one, but I did use 1N4007s as varactors in my tube theremin about ten years ago.

As you know, when you vary a reverse bias on a P-N junction, the size of the depletion region changes and essentially you're moving the plates of a capacitor back and forth. But I don't think that's what's at play here, since I noticed an identical nonlinearity from a 6AL5 with the same back-bias.

In my circuit, the best way (as I see it) to pre-bias the diodes into conduction would be to arrange for -1.2VDC at the xfmr centertap when the threshold control is at minimum. But providing that extra supply is a pain in the ass compared to the minimal benefit received; in its present form, my sidechain amp can drive the CV to -30V or so before running out of steam, and that's more than enough to drive the gain cell effectively into cutoff. The SC amp itself has nearly 30dB of gain from input to the primary of the rectifier transformer.

With the high PIV involved, Schottkys probably wouldn't work for the two diodes in the fullwave rectifier. However, you could use one for that third "steering" diode if you feel the need to reduce the forward drop a bit.