JFET cascode compressor

[TejfolvonDanone]

I read about IF amplifiers and cascode amplifier and automatic gain control in them and I thought: "Why don't i make a compressor like this?". So here it is: a JFET cascode compressor. I aimed for low parts count and a managed to breadboard it with the parts i've found laying in my parts cache. It has some gain (around 10dB) so i ended up adding a volume knob to make it a 4 knob compressor: Attack, release and compression, volume. So it can be used as a booster.
The values written in blue are the substitutes i used for the breadboarded version. All worked out well. The only thing i couldn't fix was a slight distortion with really loud guitar and hammered chords.
The Schottky-diode isn't backwards. The JFET needs negative voltage to turn off or in this case to decrease the gain and act like a compressor.
Warning: this thing needs tweaking if you do it like me (with whatever parts you find)! Also this bad boy is able to reduce the gain so much that it actually mutes the sound. I could manage to tweak it to a point where i strummed a chord but only the attacks got through immediately the rest was muted and slowly faded in. This means that the signal labeled "envelope" is too big (for my build below -2.5 volts) and the upper JFET turns off completely.

The schematic:
http://www.aronnelson.com/gallery/main.php/v/diyuser/JFETcascomp/CASCODE+COMPRESSOR.JPG.html

Sound samples may come in the near future. (If i ever have to guts to go near that mess i call "breadboarded prototype"  )

[Kipper4]

So quick question.
Two inputs?
Which one is the real guitar input?
What is the other input?

50k attack? is that right?
820nf charge caps versus a 500k release?

These are unconventional values for an envelope detector.

I'm ready to breadboard this and see what it does.
Rich

[Kipper4]

Furthermore
Is it adapted from a tube compressor?

[midwayfair]

So quick question.
Two inputs?
Which one is the real guitar input?
What is the other input?

50k attack? is that right?
820nf charge caps versus a 500k release?

These are unconventional values for an envelope detector.

I'm ready to breadboard this and see what it does.
Rich

They're the same input. Probably needs a buffer to keep T1 from loading the guitar a bunch ... or op could use a JFET for T1. Or MOSFET. 2M input impendance on each stage would drop to 1M, which would be normal for guitars.

As far as whether it's adapted from a "tube compressor" -- if you mean something like a 660, this is nothing like that. The 660 is a push-pull amplifier using pentodes where the signal is used to alter their bias. You kind of can't do that properly without actual pentodes to my knowledge.

This is reasonably interesting but it needs a way to put an absolute ceiling on how far you can drive the signal down. You could do this with a diode clamp (would clip any envelope signal that's too big).

I guess I'm also surprised there isn't crackling since this essentially wiggles the bias of T1.

>These are unconventional values for an envelope detector.

So? How expensive are a couple unusual caps? He probably grabbed what was handy or calculated them based on the time constant he wanted. It works out to about half a second.

EDIT: Actually, if you want to get clever, you could avoid an input buffer by taking the source output from T2, running it through a fair sized resistor (say, 10K?) and run that over to T1. Now you have an in-phase guitar signal without loading the guitar from the other transistor input.

[Kipper4]

"They're the same input" I did wonder that.

"As far as whether it's adapted from a "tube compressor"
I saw this and spotted  some similaritys. I wouldnt know a triode from a pentode so I felt obliged to ask.

http://webpages.charter.net/dawill/tmoranwms/Circuits/Compressor.gif

"diode clamp" just remind me what that it please.
series diode? diode to Gnd?
or is it like in the Bear Hug? Great compressor by the way Jon.

Thanks for the help Jon.
I'll go reread and try to digest.

[robthequiet]

Ah, thanks for deconfusing -- at first look I thought the Vg1 was a sidechain and Input was input. Interesting circuit. The extreme pump could be a feature as opposed to a bug. But speaking of sidechain, would it be possible to tweak this to mix 2 signals and favor one over the other, for example, to make a dynamic mix between two effect chains?

[TejfolvonDanone]

It started as a more complex circuit and i didn't want to route from on end to the other with a long wire in the schematic and make it confusing. The two "input" marks are just jumpers so the schem looks more polished. I'll post an updated version with all the changes and try them out maybe tomorrow.
http://www.aronnelson.com/gallery/main.php?g2_view=core.ItemAdmin&g2_subView=core.ItemAddConfirmation&g2_itemId=52812

EDIT: Actually, if you want to get clever, you could avoid an input buffer by taking the source output from T2, running it through a fair sized resistor (say, 10K?) and run that over to T1. Now you have an in-phase guitar signal without loading the guitar from the other transistor input.

That's clever. Point taken, I'll do it as you said.
IMHO the 10k resistor is not necessary.

50k attack? is that right?
820nf charge caps versus a 500k release?

820nF is from the E12 series but it looks like it's a rather unconventional value. 1uF NP can work too (i breadboarded it with 1uFs as you can see it in blue).
The 50k attack and 500k release is like that so the attack knob doesn't really interact with the release or compression.

Is it adapted from a tube compressor?

I didn't really take inspiration from a tube compressor i just had fun with the idea of a cascode amp AGC. It can be adapted to a tube version. You just have to swap the JFETs for tubes, alter the B+ going to them, and swap the resistors.

This is reasonably interesting but it needs a way to put an absolute ceiling on how far you can drive the signal down. You could do this with a diode clamp (would clip any envelope signal that's too big).

Point taken again.

The extreme pump could be a feature as opposed to a bug.

The envelope signal is smooth and there is no pumping because the output cap is really low.

But speaking of sidechain, would it be possible to tweak this to mix 2 signals and favor one over the other, for example, to make a dynamic mix between two effect chains?

If i understand you then you'd like to have a signal input which feeds the envelope detector beside the input signal. You'd need to add a buffer to that input and add a mixing pot going to the input of T1 and you have it. Please correct me if i misunderstood your question.

I'm ready to breadboard this and see what it does.

Good luck with it! Make sure the upper JFET's Idss is higher that the lower's. The 2n5484 and 2n3819 worked well for me. Maybe the 2n3819 can be substituted to 2n5485 but i couldn't try that. I got a BF245C but the result was really poor.

[PRR]

> IF amplifiers ... "Why don't i make a compressor like this?"

Because what works for modulated carrier (RF/IF) is often horrible for straight audio. You can cut-off half of an AM wave and do no damage, the next tuned circuit rounds it out again. FM is deliberately whacked top and bottom, AGC used only to keep pre-limiter stages out of cross-modulation.

Is a different world.

> i couldn't fix was a slight distortion with really loud guitar

Yeah, in guitar-cord work (typical signals not-large compared to typical JFET Vto) the "problems" may be minor or acceptable.
_____________

> taking the source output from T2

As the compressor clamps-down, T2 Gm drops, gain falls from near-unity to very small. Input to the level detector falls, just when we need more.

I do wonder why it is feed-forward. Simple limiters usually work best feed-BACK. Take level detector input from AGC amp output. (And FWIW, that is how IF strips work.)

> in-phase guitar signal

Phase is not critical for low-gain limiters. High-gain limiters usually have to be full-wave (detector, VCA, or both) to avoid LF instability.
_____________

> I wouldnt know a triode from a pentode

That example would be superficially more-like Tejfolvon's cascode if G2 were the control port and G1 the signal port. A pentode can be studied as two triodes in cascode, which could be JFETs. The JFET stack does not need a G3 (no secondary emission in a crystal), and Tejfolvon has approximated the high G2 voltage by picking the upper JFET different from the lower.

But when you actually build stuff, the differences mask the similarities.

And as tube limiters go, that one is crude.

[TejfolvonDanone]

Because what works for modulated carrier (RF/IF) is often horrible for straight audio. You can cut-off half of an AM wave and do no damage, the next tuned circuit rounds it out again. FM is deliberately whacked top and bottom, AGC used only to keep pre-limiter stages out of cross-modulation.

Is a different world.

I meant to just borrow the idea from RF/IF solutions not mindlessly copy. As in this case i took out the idea of altering the bias of the upper JFET to alter the gain of the cascode stack. And it turned out rather well.

I do wonder why it is feed-forward.

In a feedback arrangement the envelope signal decreases due to the negative feedback nature of the circuit. I couldn't manage to get enough voltage swing on the envelope signal to get a highly compressed sound.

I think i will opt into the feedforward topology with an added input buffer to eliminate the loading effect. I'll return with an updated and tested schem later.

[Kipper4]

Looking forward to that. /\

I haven't had a lot of spare tinkering time this week. I'm on a week at home holiday.
I'm getting through the "honey do list" slowly.

[Rob Strand]

Because what works for modulated carrier (RF/IF) is often horrible for straight audio. You can cut-off half of an AM wave and do no damage, the next tuned circuit rounds it out again. FM is deliberately whacked top and bottom, AGC used only to keep pre-limiter stages out of cross-modulation.

Is a different world.

Indeed!

Many ideas for RF simply are not practical or useful in audio.
In RF the *** bandwidth is narrow  ***.   If agc in a 1MHz radio creates non-linear distortion the distortion is way outside of the band of interest.   If you use the same idea for an audio compressor for a 1kHz signal the crap at 2kHz, 3kHz ... etc is totally unacceptable.   It's only when you start using linear signal processing that the ideas overlap.   

Most simple compressors have a sh*t load of control feedthrough, usually as a result of DC shift at the output.   It can often sound worse than letting the thing clip in the first place!

[midwayfair]

Prr, I don't se how the source level for T1 would change wwith th fm. It's the bottom half of the pair, not the top -- were you looking at a different spot?

[PRR]

> T1 would change wwith th fm.

T1 does not have a source?

"wwith th fm"??

If gate of T3 is at zero, its source is a few V +. T2 works at very small Vds, in its triode zone.

If gate of T3 goes negative (to reduce gain), its source follows. T2 works at even smaller Vds, and lower current.

Gm is a function of current.

Gm in a "happy" circuit is often similar to (inverse of) the cathode/source bias resistor. So at idle the stage has gain near half.

When current is further reduced by lowering Vds, Gm falls while source resistor is of course fixed. Gain falls even more.

This happens to be the basis of a different, obscure, tube VCA.

[Rob Strand]

In a feedback arrangement the envelope signal decreases due to the negative feedback nature of the circuit. I couldn't manage to get enough voltage swing on the envelope signal to get a highly compressed sound.

With a linear vca feedback only gives you 2:1 compression but feedforward can you infinite compression.   This comes from mathematics, not a specific circuit.  If you want an example look at the Philips applications notes for the NE570/NE571/NE572 (IIRC there's a few) you will see the different gain element connections for 2:1 compression and an agc.  In order to get more than 2:1 with feedback you have to introduce non-linearity into the control voltage -  For example the dynacomp uses the two transistor rectifier for non-linear conversion.

[midwayfair]

> T1 would change wwith th fm.

T1 does not have a source?

Now that I'm looking on a computer screen instead of a phone ... I meant the source of T2

Sorry!

[rankot]

Is there any further development of this?

[TejfolvonDanone]

Is there any further development of this?

Beside a source follower on the input to reduce the loading of guitar and a 2.4V Zener from T3's gate to ground (anode to gate and cathode to ground) to clamp the control voltage none.

[rankot]

There is a link in this thread: http://www.aronnelson.com/gallery/main.php?g2_view=core.ItemAdmin&g2_subView=core.ItemAddConfirmation&g2_itemId=52812, but it is empty. Was that the improved schematic?

[TejfolvonDanone]

Was that the improved schematic?

I don't remember if I ever posted the improved version but I did t just now.
http://www.aronnelson.com/gallery/main.php/v/diyuser/JFETcascomp/CASCODE+COMPRESSOR+V2.JPG.html