Help with DIY Tube Optical Compressor

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

figuringstuffout

Well-known member
Joined
Dec 10, 2011
Messages
45
So I'm again screwing around with tube compression and have run into an issue that requires someone way more intelligent and experienced than my novice self.

A while back I was dealing with a crosstalk issue that was sorted out with a better grounding scheme, but this time I am having a distortion issue than is driving me nuts.

See attached schematic.

NOTES
1.  The gain makeup stage sounds great
2.  The compression sounds awesome
3.  Yes I am using a 12ax7 for the LED driving...why?  Cause I have zero knowledge about transistors or opamps.
4.  The grounding is following a star scheme
5.  The 680pf is there because if it isn't more distortion bleeds through to the signal
6.  The 68k resistor to ground is as small as it is otherwise the distortion making its way to the signal is really loud
7.  B+ rail includes a choke, 100uf caps, resistors, and three separate B+ taps....one for each gain stage in the schematic


QUESTIONS
1.  Looking at the schematic, what are the obvious flaws that I should know if I knew what I was doing?
2.  The 12ax7 is distorting often, even with no gain reduction happening, and I can't figure out how it is seeping into signal coming out of the compressor.  If I drastically lower the gain of the 12ax7 then the distortion seeping through goes away, but then there isn't enough voltage to drive the LED, rendering this unit useless as a compressor.
3.  Is there a way I can maybe up the voltage gain on the LED circuit in such a way that it doesn't distort the signal?  If I can get enough voltage going to that LED without distortion then my issue will go away.  I can only get enough voltage to the LED currently when the 12ax7 is pushed to distortion.


Thanks for the help!



 

Attachments

  • comp.jpg
    comp.jpg
    96.1 KB · Views: 228
There are a number of things I notice about your schematic.

The 220nF feeding the Jensen transformer is rather on the low side. I would recommend increasing this to at least 2.2uF

The two halves of the 12AU7 provided rather a lot of make up gain - probably 40dB - possibly why you placed a 6dB pad between them.

The 68K is necessary to refernce the 12AX7 output to 0V before the rectifier

The 12AX7 really does not have enough grunt to drive the Vactrol. It is almost certainly clipping severely and this is what is bleeding through into the audio chain. I would suggest you think about using a 12AU7

I suspect the 500K pot just before the Vactrol has little effect. The Vactrol impedance is so low that there will be practically nothing reaching the Vactrol except when the pot is fully up. An SRPP side chain based on the 12AU7  or 6DJ7 might be more effective

Check out my schematic here:

http://www.ianbell.ukfsn.org/EzTubeMixer/docs/PMTCL/PMTCLsc.jpeg

Cheers

Ian
 
Why not a transistor side chain? other than you don't know about them... if you have an scheme would you use it?

Much cheaper but need a new PS... still cheaper than a tube, a 12V aprox unreg 50mA PS is just a tiny one or you could take it from your heaters with a doubler or something.
Here is an scheme ATT, REL, THR controls, 2 fet and 1 BJT. Just a couples of values of resistors to set LED current.

In this link you have an APP to know how to use a trim for the fet: http://www.hansenaudio.se/FET%20impedance%20converter.jpg
The BJT resistor you will put it to set LED current I=(Ufet-0.7V)/R. If you trim your fet to have 4V at the bjt resistor now you have I=4V/R (a trim to match your led and adjust the range of compression)

For a faster less math and measuring you could live fet net out and use a 10K from fet to ground and let's say 610Ω for BJT to have a max under 15mA for the led.

JS
 

Attachments

  • pict2105 3904, no06.jpg
    pict2105 3904, no06.jpg
    41.5 KB · Views: 123
All, thanks for the suggestions so far!
See responses below:

CJ said:
try higher cap values for the 680 pf,

I tried larger caps and it takes an 8n2 cap to get rid of most of the noise...of course it also decreases the amount of voltage for the LED so the compressor is barely working at that point.

ruffrecords said:
There are a number of things I notice about your schematic.

The 220nF feeding the Jensen transformer is rather on the low side. I would recommend increasing this to at least 2.2uF

The two halves of the 12AU7 provided rather a lot of make up gain - probably 40dB - possibly why you placed a 6dB pad between them.

The 68K is necessary to refernce the 12AX7 output to 0V before the rectifier

The 12AX7 really does not have enough grunt to drive the Vactrol. It is almost certainly clipping severely and this is what is bleeding through into the audio chain. I would suggest you think about using a 12AU7

I suspect the 500K pot just before the Vactrol has little effect. The Vactrol impedance is so low that there will be practically nothing reaching the Vactrol except when the pot is fully up. An SRPP side chain based on the 12AU7  or 6DJ7 might be more effective

Check out my schematic here:

http://www.ianbell.ukfsn.org/EzTubeMixer/docs/PMTCL/PMTCLsc.jpeg

Cheers

Ian

Very helpful response Ian, thank you!

1.  I have tested 220n all the way up to 22uf and have not noticed really much of a difference, and since I like the sound of polyester caps and not electrolytics, then I kept with a big Mallory cap.
2.  Yep...I like my tons of gain makeup...=)
3.  Agreed
4.  Yep, the 12ax7 is distorting like crazy and that is indeed somehow bleeding through to the signal on the first 12au7 makeup gain stage.  When I tried the 12au7, it fixed a lot of the distortion problem but now the compressor barely engages.  Thus, I need something to provide the voltage the 12ax7 provides, but without the clipping that is happening with the 12ax7.
5.  The 500K pot does act as a threshold control, but you are mostly right in that it only starts really working when about halfway up.  Unfortunately, when using a 12au7, there is simply not enough voltage for the LED even when the 500k pot is all the way up.
6.  Your schematic looks great!  I have never used a 6922 but it looks like you are using a modified cascode style structure for the LED.  Theoretically this would mean you are entering the realm of pentode design...and that for the extra gain i am assuming?  I do have an EF86 lying around...do you think I could use that to drive the LED without the distortion I'm getting from the 12ax7?


Thanks again for the help!
 
joaquins said:
Why not a transistor side chain? other than you don't know about them... if you have an scheme would you use it?

Much cheaper but need a new PS... still cheaper than a tube, a 12V aprox unreg 50mA PS is just a tiny one or you could take it from your heaters with a doubler or something.
Here is an scheme ATT, REL, THR controls, 2 fet and 1 BJT. Just a couples of values of resistors to set LED current.

In this link you have an APP to know how to use a trim for the fet: http://www.hansenaudio.se/FET%20impedance%20converter.jpg
The BJT resistor you will put it to set LED current I=(Ufet-0.7V)/R. If you trim your fet to have 4V at the bjt resistor now you have I=4V/R (a trim to match your led and adjust the range of compression)

For a faster less math and measuring you could live fet net out and use a 10K from fet to ground and let's say 610Ω for BJT to have a max under 15mA for the led.

JS

joaquins,
Thanks for the reply!  I have thought about the transistor route, but you nailed it in that I have zero experience with them and don't know the first thing about them.
I greatly appreciate the schematic and will probably give it a try!

Questions:
1.  The resistor below the 2n3906 to set the current for the LED....what ballpark number would I be looking for on that resistor?
2.  So this circuit would essentially be affecting the resistance from the LED to ground from a constant 12v supply right?  How would this react differently than the LED responding to a tube amplifying the signal for the voltage to the LED, or would there be basically no difference?


 
What I want to know is why you aren't using the second triode in your side chain. Using an au7 you could be getting the gain needed with two stages, or using an srpp to drive the led. Maybe take the side chain input off the first gain stage output.
 
gemini86 said:
What I want to know is why you aren't using the second triode in your side chain. Using an au7 you could be getting the gain needed with two stages, or using an srpp to drive the led. Maybe take the side chain input off the first gain stage output.

1.  I was wanting to make two channels and double up on that 12ax7.  I might end up having to just add another tube after all.
2.  With the two 12au7 gain stages, yes I'd be getting plenty of gain, but would it be distorted?  I get even a tiny bit of distortion from one 12au7 stage and it isn't enough to really drive the LED...with more gain from two 12au7 stages I have a feeling I won't be able to keep it clean and have the voltages.  Maybe I'm completely wrong (probably am).
3.  I haven't tried the srpp route yet...that might get me enough voltage without distortion...kind of like ruffrecords's circuit
4.  If I take the sidechain off the output of the first gain stage then the gain makeup will control the LED voltage as well
 
figuringstuffout said:
joaquins said:
Why not a transistor side chain? other than you don't know about them... if you have an scheme would you use it?

Much cheaper but need a new PS... still cheaper than a tube, a 12V aprox unreg 50mA PS is just a tiny one or you could take it from your heaters with a doubler or something.
Here is an scheme ATT, REL, THR controls, 2 fet and 1 BJT. Just a couples of values of resistors to set LED current.

In this link you have an APP to know how to use a trim for the fet: http://www.hansenaudio.se/FET%20impedance%20converter.jpg
The BJT resistor you will put it to set LED current I=(Ufet-0.7V)/R. If you trim your fet to have 4V at the bjt resistor now you have I=4V/R (a trim to match your led and adjust the range of compression)

For a faster less math and measuring you could live fet net out and use a 10K from fet to ground and let's say 610Ω for BJT to have a max under 15mA for the led.

JS

joaquins,
Thanks for the reply!  I have thought about the transistor route, but you nailed it in that I have zero experience with them and don't know the first thing about them.
I greatly appreciate the schematic and will probably give it a try!

Questions:
1.  The resistor below the 2n3906 to set the current for the LED....what ballpark number would I be looking for on that resistor?
2.  So this circuit would essentially be affecting the resistance from the LED to ground from a constant 12v supply right?  How would this react differently than the LED responding to a tube amplifying the signal for the voltage to the LED, or would there be basically no difference?

1- Something between 470Ω and 1K should be the way to go... about the same you use for that led to fully bright at 12V, It won't get that high but close and will be safe that margin. Transistor output will be a variable current source limited by that resistor. The fet net Try with a 10K just to ground, maybe a little lower but not much, don't go under 4k7.

2- the main difference is that is not full rectified input wave so pos peaks won't reach the led, not a problem I thing for a slow opto. If you wan't you can use a bridge as in the other but try this first and then you could add it with some mods. I don't think you have other sustancial difference, It only matters when non symetrical signal at input but not a big deal, this is not a precision compressor anyway, in a VCA will be a crime do half wave rectifier, in an opto I don't think so.

JS
 
a transistor to buffer the led drive signal from 12ax7--mashup of above ideas.could be one rated for high voltage to use existing psu...or just do Ians plan
 
I think at least some of your problems are due to gain structure. You have a lot of make up gain and a lot of side chain gain. I suspect this is because your input levels are generally quite low. My compressor, when set for maximum compression, starts to compress at about -20dBu input and reaches maximum compression when the input is +10dBu. At this point the gain is reduced by 20dB to -10dBu so you only need 10dB of gain make up to get back to  0dBu. 

If you actually do need more gain in both the gain make up and the side chain then you could and probably should add it before the opto. So I would suggest you have a 12AX7 stage at the input driving the opto pot divider. You then only need one half of the 12AU7 for gain make up and you could use the other half to drive the opto LED.

Cheers

Ian
 
Look at the LA2a.

Tube sidechain BUT there's a FAT tube driving the lamp, preceeded by 1 or 2 small tubes.

Not one skinny tube trying to do it all. Not using a bamboo fly-casting rod to land tuna.

Worse because LA2a's lamp is a hi-V low-I crystal, and your lamp is a 1.6V 16mA dingus.

And LA2a puts the sidechain gain where it belongs: in the front of the sidechain.

OTOH your *500K* pot feeds a *100 ohm* (1.6V/0.016A) LED. First the poor 'AX7 strains to force anything into the 500K, but is severely mismatched to the real load.

Start with a 10mA-20mA tube. 12AU7 is not ideal but a heap fatter than 12AX7.

Run the 12AU7 *HOT*. With 300V, 150V across the tube, you want plate resistor 150V/10mA or 15K. (Note that a 2W part is hardly ample here.) Cathode resistor may have to be a few hundred ohms; start with 470 but work down to 100r to get plate voltage near mid-supply.

Put another gain stage before that. 12AX7 can easily drive 12AU7.

Put the sidechain gain pot before that.

We still have massive mis-match between 12AU7's ~~5K output and 1K-100 ohm LED. It would be nice to have a 10K:150 transformer, but this violates your junkbox approach, and Ian can tell you that headphone OTs are a tough problem.
 
Again... go for transistors! BJT is a perfect match for a led and much cheaper! If you use 12.6V for your heaters (perfect for the 12AX7 you are using) you just need a rectifier and a couple of caps to get it work! if you already have a 6.3V secondary a doubler should do the trick.

JS
 
> The 68K is necessary to refernce the 12AX7 output to 0V before the rectifier

The rectifier is just wrong in many ways.

68K is out-of-sight compared to a 1.6V 16mA (100 ohms) load. Waste of a dime.

So is 22n. (Though maybe not so bad considering how little current the skinny AX7 can drive.)

No clue how the 680p landed there.

Connected the way I show, no "DC bleed" is needed. With "ideal" drive it is fully symmetrical. With 12AU7 biased as shown, it can swing ~~7mA negative peaks, which ought to be more than enough LDR clamping for a 68K divider leg. (Positive peaks can be much higher; but I didn't pick the tube-types or supply voltages; a little extra pos after the neg side bottoms does no harm.)

 
I'm not sure if anyone has mentioned, but do some searching on the 'daoc' unit - this is a very interesting 'analag' design of a tube compressor using vactrols (led+photoresistance) which used the la2a as a starting point.

daoc = d-aoc  -> dual analog optical compressor  from members analag and silent arts

Very popular and you will find lots to interest you.

Cheers
 

Latest posts

Back
Top