Vari Mu compressor without input transformer

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

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

Heikki

Well-known member
Joined
Jul 31, 2008
Messages
401
Location
Finland
Here's something I'm working on right now.
http://ghr.fi/varimu/varimu2.pdf (couple resistor values are wrong on the schematic. I'll upload corrected schematic later.)

No input transformer and the control voltage is brought to cathodes. Impedance for the control voltage is very low but it's not a problem for solid state control amp. Only transformer used is on the vari-mu stage plates. Input stage is cross-coupled NE5532  and it drives the vari-mu stage differentially. If one input is grounded the gain goes down 6dB. Other than the input stage and the CV to cathode there's nothing unusual going on.

I got the circuit complete enough today to do some testing and it seems to work very well. For these initial tests I used Edcor XSM 10k/10k as the plate transformer and I can't understand how anyone can use them in their builds, they pick up hum and buzz like crazy. I would need a huge chassis to get them far away enough from the power transformer. I'll replace the transformer with Lundahl 1540 later on.

I don't know if anyone has done a vari-mu with CV to the cathode. To me it seemed like the easiest way to get rid of the input transformers, keep fast attack times and make a cheaper vari-mu with better performance. It works great, even though some purists might object to the use of op-amps and transistors. I'll do some more testing and measurements and post some specs later.

varimu-2.png
 
Get rid of the Edcor. Use maybe 5K plate resistors and .47uF blocking caps into the (cold & hot)op amp circuit below.  You now have single ended output from your tubes. Hang two 10K pots in parallel off the output of A2. One pot wiper feeds a DRV134 and is your balanced output. The other pot feeds another DRV134 and is your threshold to your sidechain.

For the input I would again use a line receiver chip ---> Pot ----> line driver chip -----> tube grids. That way you can use a single pot instead of a dual for input attenuation.

Fully transformer-less Vari Mu!!!!

Cheers,

Ian
 

Attachments

  • balfig13.gif
    balfig13.gif
    12.1 KB · Views: 64
bluebird said:
Get rid of the Edcor. Use maybe 5K plate resistors and .47uF blocking caps into the (cold & hot)op amp circuit below.

There's a large common mode voltage shift on the tube plates when going to deep gain reduction and the next stage has to ignore it. The circuit you posted wont work because of the large common mode voltage.
 
The circuit I posted will work. The caps block the high voltage..  Look at the Gates Sta level, no interstage transformer. Trust me.... ;D
 
Caps block the DC voltage but caps won't block the big voltage swing on the plates when it goes into compression. I know some of the old compressors didn't use transformer on the vari-mu plates, but they tend to have fairly slow attack times and the stage after the variable gain amp can handle large common mode voltages. Some speech limiters even cut the low frequencies significantly before the next stage. Your circuit would probably work with very slow attack times.
 
I'm telling you, it works, faster even. The tubes are not loaded by the transformer. I've done it many times. Some of my best varimu compressor circuits are like this, it sounds amazing. Just try it, you will thank me.
 
If the vari-mu stage is RC coupled to a tube stage it works fine. It'll clip until the coupling cap drains enough to bring the tube back to normal operation. I wouldn't expect op amp with +/- 15V supply voltage to clip or recover nicely when several tens of volts hits its inputs.
 
The AC swing of each tube when referenced to ground is not that much. Test it, unhook the transformer and put a .47uf cap and then a 100K resistor to ground off of each plate. Just like my schematic but without the op amp. Input a sign wave at a relatively high level and read the AC voltage at the junction of the .047uf cap and the 100K resistor. Your typical 5534 can handle a common mode input swing of +/- 13 volts on +/- 15 volt rails. Thats 26volts. So live dangerously and power your op amps with +/-20V that gives you almost 40v of swing. That tube is not swinging that much. If your really scared about it, hang 2 10v zeners back to back in parallel with the 100k resistors. they will shunt anything over a 20v swing. but you don't need them.

Another hint, you don't need two 6BA6's either. A 6ES8 or 6BC8 will work very well in this situation with a plate voltage of 100v.

I can see you already made your PCB's so you might not want this to work in your head, but it does. how do you think Pendulum audio is doing it? Those are input transformers in the picture...
 

Attachments

  • pendulum el8.jpg
    pendulum el8.jpg
    42.7 KB · Views: 114
Probably part of the reason your circuit 'works' is that o.o47uF and 100K have a 3dB point at 33 Hz. As others have pointed out, it is not the signal level that is the problem. it is the dc shift in the plate voltages with a fast attack time. This can be several tens of volts, way beyond the common mode capability of an op amp unless you deliberately attenuate low frequencies as you have.

Cheers

Ian
 
ruffrecords said:
Probably part of the reason your circuit 'works' is that o.o47uF and 100K have a 3dB point at 33 Hz. As others have pointed out, it is not the signal level that is the problem. it is the dc shift in the plate voltages with a fast attack time. This can be several tens of volts.

Yes I understand, but this has never been a problem with this circuit in practice, I've even used 1uf caps and it was fine. with a 10K resistor on the plate and 100v, your only gonna get about 60V on the plate. So your not going to start a fire even if you momentary swing 40 or so volts. Use an opa2604 with +/-24v rails and your definitely good to go.

I have engineer friends who have been using boxes I made with this very topology in them (minus the cathode CV) for the last 5 or so years and they love it. And I never had to fix a bad op amp in them either.  Promise... I used a more traditional input stage .10K:10K input transformer with CV to the center tap. But I like the idea of a transformerless input as well. You do have to have those nasty lytic blocking caps in there though.
 
bluebird said:
Yes I understand, but this has never been a problem with this circuit in practice, I've even used 1uf caps and it was fine. with a 10K resistor on the plate and 100v, your only gonna get about 60V on the plate. So your not going to start a fire even if you momentary swing 40 or so volts. Use an opa2604 with +/-24v rails and your definitely good to go.

This does not stack up. With a 100V supply, a 10K resistor and 60V on the plate you have 4mA plate current which is not the zero bias point for a 6BC8 or the 6ES8. A 10K load line with 100V plate on a 6BC6 is about 40V plate. On a 6ES8 it is about 25V so in both cases the swings are going to be much bigger than you think.

Cheers

Ian
 
In practice you can get away with a lot of unnoticed offset if there's no low frequency content passing or triggering.  You can get away with it further if threshold/ratio is low and the unit is practically always doing some amount of GR.  I had a single ended cap coupled 6386 compressor once, definitely not clean but not as bad as you would imagine either.  Depends on what you are after.
 
ruffrecords said:
This does not stack up. With a 100V supply, a 10K resistor and 60V on the plate you have 4mA plate current which is not the zero bias point for a 6BC8 or the 6ES8. A 10K load line with 100V plate on a 6BC6 is about 40V plate. On a 6ES8 it is about 25V so in both cases the swings are going to be much bigger than you think.

Thanks Ian, still learning everyday!
I was figuring for a 6BA6 with screen connected to the cathode like in Heikki's schematic.  With maybe a 200 ohm cathode resistor? I could be wrong.
I remember shifting around the operating point of the 6ES8 I was using because I had a weird sidechain schem where I dropped the cathodes down to -12v

Regardless, like I said I never had trouble with a big offset screwing up the opamp. If I did, I would not recommend doing this.

Ian

PS sorry I feel like I derailed this topic which was about cathode control voltage...I'll shut up about it now.

 
bluebird said:
Thanks Ian, still learning everyday!
I was figuring for a 6BA6 with screen connected to the cathode like in Heikki's schematic.  With maybe a 200 ohm cathode resistor? I could be wrong.

Which is also puzzling because the 6BA6, unlike the other two tubes, is a pentode. Presumably it is wired as a triode in a vari-mu. Unfortunately I cannot find and dat for it wired as a triode.

Cheers

Ian
 
ruffrecords said:
Presumably it is wired as a triode in a vari-mu. Unfortunately I cannot find and dat for it wired as a triode.

Yes, wired in triode the 6BA6 is very similar to the 6386 tube.
wired like this:
 

Attachments

  • 6BA6 Triode.jpg
    6BA6 Triode.jpg
    28.6 KB · Views: 88
Very nice Heikki.

I have discussed  transformerless  Vari Mu with friends many times. I would definitively do it like this:

Input: CV mixed to signal, to grids. Not cathodes. ECC189 as tube, -15 volts total Vg is enough.
Output: 10:1 resistive pad to bring anode voltage changes at all times into OP amp voltages. Carefully chosen OP amps to do the "conversion". Input capacitance modulation has to be small, because source impedance will be fairly high. Common mode linearity has to be good too.

How ever...I will never do it.

 
bluebird said:
The AC swing of each tube when referenced to ground is not that much. Test it, unhook the transformer and put a .47uf cap and then a 100K resistor to ground off of each plate. Just like my schematic but without the op amp. Input a sign wave at a relatively high level and read the AC voltage at the junction of the .047uf cap and the 100K resistor. Your typical 5534 can handle a common mode input swing of +/- 13 volts on +/- 15 volt rails. Thats 26volts. So live dangerously and power your op amps with +/-20V that gives you almost 40v of swing.
Cancellation of the transient is dependant not only on the CMRR of the opamp, but also on the dynamic balance between the vari-mu tubes, which is a big uncertainty. The CT xfmr inherently provides additional reduction of DC shift because of its low DC impedance. It should be possible to replace the CT xfmr with simulated inductors, which can even use tubes.
 

Attachments

  • simulated inductor 1.jpg
    simulated inductor 1.jpg
    144.9 KB · Views: 68
Finally got back to working on this one. With LL1540 transformers the hum and buzz issues are gone. From a box of 100 6K4P tubes it wasn't hard to find bunch of pairs that matched very well for gain and low thump. It was pretty hard to find matches for both channels to have same distortion figures at same amount of gain reduction. I'm not sure if most people go through the trouble of matching both channels for distortion.

Some measurements
Frequency response 20 - 20kHz ± 0.5dB, stays pretty much same even at deep gain reduction.
At +10dBu out and 10dB gain reduction, THD @ 1kHz < 0.25% (mostly 3rd harmonic)
At +10dBu out and 10dB gain reduction, CCIF IMD < 0.15%

I still need to measure CMRR of the input stage. I used 1% resistors so worst case scenario might be 30dB CMRR.

Distortion and frequency response are similar to my 6L6 compressor and to my ears they even sound very similar. I'm not sure I can hear distortion <1% when it's mostly 2nd and 3rd harmonic, so the sidechain probably has the biggest influence on "sound". My 6L6 compressor has almost the same sidechain, except control voltage goes to grid and is negative.

The negative side of cathode CV obviously is the large current needed for CV but I think it's worth it. Good input transformers are expensive and with cathode CV it can be replaced with cheap fairly simple circuit.

13.png
 
Back
Top