Hi folks,
Was thinking about this, with no particular circuit in mind. Some audio compressors, use variable mu tubes. We all kmow the principal: As the grid voltage lowers with respect to the cathode, the grid contours get closer together, so the gain changes.
The characteristics looks a lot like the cold-clipper bias used in guitar amps to create intentional harmonic distortion! Now this would sound terrible in a studio compressor. So this cannot be the full story!? There must be a way to get the compression effect without distortion. Or no? Any idea what else is going on is this type of audio circuit?
Was thinking about this, with no particular circuit in mind. Some audio compressors, use variable mu tubes. We all kmow the principal: As the grid voltage lowers with respect to the cathode, the grid contours get closer together, so the gain changes.
The characteristics looks a lot like the cold-clipper bias used in guitar amps to create intentional harmonic distortion! Now this would sound terrible in a studio compressor. So this cannot be the full story!? There must be a way to get the compression effect without distortion. Or no? Any idea what else is going on is this type of audio circuit?
Isn't dynamics compression itself a type of distortion of the audio signal?
Perhaps it doesn't generate additional harmonics per se, but it does distort the waveform (which in itself can be thought of as changing the harmonic content of the amplified signal, I think).
Perhaps it doesn't generate additional harmonics per se, but it does distort the waveform (which in itself can be thought of as changing the harmonic content of the amplified signal, I think).
No.
You are describing a VERY non linear stge, to round or clip signal tops , and which of course will add tons of distortion.
Similar to clipping with Si diodes: tops will be rounded, not sharply clipped, but distortion appears.
In fact Guitar pedals as Tube Screamer or MXR Distortion + work that way; not here.
Not at all and you are missing half the picture.
And it´s not.
As I said above, *clipping* or getting on a very nonlinear area, are instant effects, they only depend on reaching a certain voltage threshold, period.
No time constants involved and waveforms get distorted, big way.
Real *compressors* are very different, you have TWO elemens interacting there.
1) a variable gain stage, which is voltage controlled.
As in controlled by an external voltage.
2) a control voltage generator.
Typically they tap Audio signal at some point, some even build a separate dedicated gain stage, rectify it so as to get a DC voltage representing signal, this voltage is processed in various ways, sometimes EQ´d , and applied to time constant filters which determine how and when it will affect the main stage gain.
It does NOT follow instantaneous signal voltage in real time, so it does not clip individual waves, just their "envelope".
And in fact there is a lot of Science/Mojo/Magic/"ear" involved in creating and processing that control voltage.
IF you had posted a compressor schematic 🙄 , it would be much easier to explain this and show you the different elements. 😀
Last edited:
Soft limiters reduce amplitude as signal gets bigger.
This causes distortion.
I use one in my guitar amplifier design to deliberately sound like over driven valve distortion.
To limit without distortion means reducing the whole of the signal and not just the tops.
I use audacity to amplify quiet mp3 files.
This causes distortion.
I use one in my guitar amplifier design to deliberately sound like over driven valve distortion.
To limit without distortion means reducing the whole of the signal and not just the tops.
I use audacity to amplify quiet mp3 files.
So the answer is: They don't. The assumption they don't introduce distortion or a lot less of it is not at all true. In fact, because of the way they work they add a lot of distortion in terms of spectral balance and phase because their slow response compared to silicon based compressors causes different parts of signal (frequency wise) to affect the compressor drive differently.
I suppose it just so happens that the way it works a lot like an optical compressor in terms of curves and the use of tubes creates a spectrum that masks a lot of the distortion because the harmonics can mask the spectral balance distortion.
Spectral balance distortion being relative to unaffected signal here.
I suppose it just so happens that the way it works a lot like an optical compressor in terms of curves and the use of tubes creates a spectrum that masks a lot of the distortion because the harmonics can mask the spectral balance distortion.
Spectral balance distortion being relative to unaffected signal here.
All professional limiter/compressors had push-pull gain reduction stages to avoid audible "thumps" at the compression knee. Push-pull cancelled the thumps.
I think variable gain stages work with low signal levels. At low signal level the transfer function is approximately linear. The generated harmonics due to nonlinearity are negligible.
Think about the AGC circuit in AM/FM receivers.
Think about the AGC circuit in AM/FM receivers.
I think this is really important. Can you explain the principles a bit more? I'm familiar with push-pull but don't understand why this would be.
1) They run VERY small signals
-or-
2) They run push-pull, which gives a first-order cancellation of the bentness.
The first was used in some consumer tape recorders. They even wound a special tube with less change of Ip for less thump. But it quickly get jammed between THD and hiss.
All the classics used a push-pull deal. See #8 above. Use your SEARCH to find more examples.
Oh, and 3) AM modulate the audio onto a MHz carrier and use standard IF AVC systems to change gain, then demodulate. The AM carrier rejects 2nd order nonlinearity. Even when half the wave is cut-off, the tuned circuits regenerate a sine carrier and whatever is modulated on it. Because this used to be THE core technology of most electronics, it seemed simple. Apparently it was not fabulous, because it was only sold a couple times. Perhaps because simple detectors distort, and good detectors cost a lot.
There were two other dead-end techniques, working the variable-Gm idea but backward, as NFB or as a null bridge. Both were common because one was G.E and the other was Gates, and if you bought the transmitter they usually sold you the rest of the station.
BTW, "variable µ" is mostly marketing. The µ can not be made to "vary" as much as we need for audio level control. We are unlikely to get µ to vary more than 3:1 before the tube craps-out for lack of current, but we want 10:1 or more of gain loss. Mostly we overwhelm µ with impedances and just go with the natural (and inevitable) drop of Gm with current.
Last edited:
Thanks directdriver, an excellent example, very well chosen because here "simpler" is good to explain better.
1) the Audio chain is 6BC8 > 6CG7 , the first one being the variable gain stage, the second one a miniature power amp because in Recording/Radio-TV studio use Compressors were built as Line stages to make them easier/versatile to use in that environment.
2) the control voltage generator uses the 6AL5 dual diode as a peak rectifier.
3) and then they add extra elements to make best use of that rectified voltage, provide a "Compression meter" calibrated in dB , etc.
4) being a (special) line stage, input and output are balanced 600 ohm, the old (and I mean **OLD**) school standard since the 30´s when Radio and "Talkies" movies started and only Engineers specialized in some kind of Audio were *Telephone* Engineers, so they brought many of their preferences with them.
In this case, hidden in plain sight:
* 600 ohm output can be wired as 150 ohm ... so it can drive *four* (count´em 😉 ) 600 ohm lines without problem
* 15K "bridging" input, which actully means it´s such a high impedance (15k 😱 😀) that you can "bridge" it across any 600 ohm line (which is already driving a 600 ohm load) and not load it down.
Now you see even if offering different "numbers", everything revolves around the Telephone standard 600 ohm convention. 😎
I clear this out because many think "numbers were pulled out of the Blue" .... they were not.
5) how do they change gain (attenuate) as a function of output signal voltage?
The gain stage uses 6BC8.
http://www.ralphselectronics.com/productimages/RCA-6BC8.PDF
Well, it depends.
What you say applies to regular tubes (99% of them): transconductance or mu varies with bias or current in all tubes, that´s why I always say a great part (and surprisingly little discussed) of Tube Sound is that they all compress, at least to a certain degree.
It´s also true that if going too far along that path, you *strangle* the tube, you also cut its current OFF.
BUT: there are tubes specially designed to avoid that, they are called Remote Cutoff types.
Typically designed for Radio or TV use , a market of tens of millions yearly, and not Studio Compressors (100 a year?) so the Audio guys must kneel and use available Radio/TV types.
Per datasheet, 6BC8 is called a "medium Mu twin triode with Semiremote Cutoff characteristics", go figure.
"Intended for VHF TV tuner applications" (why doesn´t that surprise me?) 🙄
So 6BC8, contrary to, say, 12AX7, can stand important negative bias which will murder its Gain (what we want) but without chopping off the signal.
6) now to how-it-works
Input transformer gets input signal.
Very flexible, it allows for balanced/floating input, also "no load" (by 1930 standards) 15k or 600 ohm by simply strapping a 600 ohm resistor across input terminals.
Dual P1 pot supplies balanced variable signal to following gain stages, which are also balanced end to end.
Probably labelled "Compression on the front panel, because in practice the higher the signal, the more compressed it gets.
6BC8 provides gain, but specially it can be *voltage* controlled.
Control voltage feeds T1 secondary and both halves of P1, notice these 3 elements are floating, not ground referenced by themselves, and will take any DC voltage fed them.
Shown but not explained (in schematic) is that voltage there can go from -0.5V (idle/no signal) to scary -30V (full compression).
Yes,-30V could block any regular signal triode ... but remote cutoff tubes can handle that.
6BC8 plate signal feeds 6CG7 which further amplifies it, drives output transformer, but most important, feeds signal rectifier diodes 6AL5, through C5 and C6.
Since diodes have a significant threshold which makes them "deaf" to low level signals, they are pre-biased by the R15 - P3 - R7 network so they are already on the edge of rectification.
This can also set a "threshold" level beyond which compression appears, according to Engineer´s wishes.
Pre-bias varies from 2.3V to about 20V
I bet there MUST be detailed setup and calibration instructions in Compressor Manual, here I have to guess them.
In any case, an educated guess, I have worked with Studio equipment for decades.
Rectified negative voltage charges C4.
I talked earlier of "time constants" to make compression "natural" to our taste.
R12 and C4 set the "charging" time constant, so "how fast" will compressor react, it´s called "Attack Time".
Not way too short, or it will distort waveforms like its cousin the diode clipper (which is "instantaneous") nor too slow so it will let, say, a drum hit pass uncompressed and mess everything.
Here time constant is 0.033 seconds (33 milliseconds) which is fine, quite fast to our ear but its associated frequency is 4.8Hz, so it won´t distort , say, 50 Hz which is a decade away.
C4 and R9+Pd set how long it takes for C4 to discharge (to get out of compression) , typically called "Release Time", variable between 0.27 seconds and 1.27 seconds.
So now you have the full picture: the variable gain stage, the control voltage generator and how they interact.
ALL time constants are way longer than any Audio signal, so they do not affect individual waveforms (so they do not clip/distort) and only distortion present is very acceptable, low order "smooth" type created by natural tube nonlinearities, like on every "normal" gain stage.
In fact Vari-mu or Remote-cutoff tubes are not very linear but using them at lowish levels keeps everything polite.
Almost forgot: measuring actual compression would require a microprocessor/DSP and software , not exactly available way back then.
The practical-minimalistic solution was to measure V1 current at zero signal, call it "Zero Gain reduction", then at full compression, call it Full Compression/Gain Reduction or even write the actual dB attenuation, and label the meter scale accordingly.
State of the Art and valid even today 😉
There seems to be a confusion between limiter and compressor.
A limiter acts only when the max level is reached and creates lots of distortion. It is used to protect the amp from being over-driven.
The compressor starts to work at much lower level like -12 or -6dB (threshold) and reduces the overall gain at the ratio rate. 2:1 rate means a signal 6dB above the threshold is attenuated by 3dB.
The control voltage has attack time(short) and release time(long).
Of course a bit of distortion is created the moment when reducing the gain. but then gain remains there as long as high levels are coming in so it is like somebody turning the volume knob up and down.
All bold items are adjustable over a wide range.
The track seems to be louder, the dynamics are reduced, properly adjusted it is hardly recognizable.
A limiter acts only when the max level is reached and creates lots of distortion. It is used to protect the amp from being over-driven.
The compressor starts to work at much lower level like -12 or -6dB (threshold) and reduces the overall gain at the ratio rate. 2:1 rate means a signal 6dB above the threshold is attenuated by 3dB.
The control voltage has attack time(short) and release time(long).
Of course a bit of distortion is created the moment when reducing the gain. but then gain remains there as long as high levels are coming in so it is like somebody turning the volume knob up and down.
All bold items are adjustable over a wide range.
The track seems to be louder, the dynamics are reduced, properly adjusted it is hardly recognizable.
Thanks all, that makes a ton of sense. I was really just interested in the core principles, and did not expect such detail, but that is really appreciated!
>> Quote: "variable µ" is mostly marketing.
> Well, it depends. What you say applies to regular tubes (99% of them)
I don't know why I type if you will not read.
Even in a "remote cutoff" the mu does NOT vary enough for audio purposes. "Vari-Mu" is an over-simplification and now a Trademark.
6BC8 works slightly better than 12AU7 which has a fairly constant mu.
> Input transformer gets input signal.
Aside from floating input, it allows bias/GR signal to mix with audio signal without load on either. First thing the young kids want to do is lose the transformer. Audio now becomes gain control, time-constants messed-up, it sucks. The function can be done with op-amps but anything less than 5 opamps is compromised somehow. (Yes, 5*$0.19=$0.95 is cheaper than a transformer.)
> diodes have a significant threshold which makes them "deaf" to low level signals
Hollow-state diodes in a 30V control system, the threshold is insignificant. The bias is to delay operation to give *either* compressor or limiter curve.
> ALL time constants are way longer than any Audio signal
Attack time is often adjustable into the bass range, making thump a major problem. Slow attack would avoid this but would allow momentary over-level, causing over-cut on disk or tickets from the broadcast authority.
> A limiter acts only when the max level is reached
> The compressor starts to work at much lower level
Either curve may be set by sidechain gain and bias. In the simple Altec 436 there is the one pot P3 which sorta does both. Another version has switchable level into the rectifier. The Fairchild has more controls and many time constants. When <$5 opamps arrived analog designers went crazy. Actually Barry Blesser was very sane but adjusting all the adjustments (even the few front-panel ones) for best result in live transmission was challenging.
> Well, it depends. What you say applies to regular tubes (99% of them)
I don't know why I type if you will not read.
Even in a "remote cutoff" the mu does NOT vary enough for audio purposes. "Vari-Mu" is an over-simplification and now a Trademark. 6BC8 works slightly better than 12AU7 which has a fairly constant mu.
> Input transformer gets input signal.
Aside from floating input, it allows bias/GR signal to mix with audio signal without load on either. First thing the young kids want to do is lose the transformer. Audio now becomes gain control, time-constants messed-up, it sucks. The function can be done with op-amps but anything less than 5 opamps is compromised somehow. (Yes, 5*$0.19=$0.95 is cheaper than a transformer.)
> diodes have a significant threshold which makes them "deaf" to low level signals
Hollow-state diodes in a 30V control system, the threshold is insignificant. The bias is to delay operation to give *either* compressor or limiter curve.
> ALL time constants are way longer than any Audio signal
Attack time is often adjustable into the bass range, making thump a major problem. Slow attack would avoid this but would allow momentary over-level, causing over-cut on disk or tickets from the broadcast authority.
> A limiter acts only when the max level is reached
> The compressor starts to work at much lower level
Either curve may be set by sidechain gain and bias. In the simple Altec 436 there is the one pot P3 which sorta does both. Another version has switchable level into the rectifier. The Fairchild has more controls and many time constants. When <$5 opamps arrived analog designers went crazy. Actually Barry Blesser was very sane but adjusting all the adjustments (even the few front-panel ones) for best result in live transmission was challenging.