Another diode bridge compressor

[Heikki]

Here is a diode bridge compressor I’m working on right now. The idea is pretty much same as in my Vari-mu compressor except diode bridge is the gain reduction element. Here is the schematic and sidechain is the same as in my vari-mu. Wall wart and a DC-DC converter are used for power supply.

I’ve done some measurements and the distortion spectrum is fairly similar to vari-mu distortion (mostly third harmonic). With the sidechain ratio is adjustable between 2:1 and 20:1. Next week I’ll probably get the thing inside an chassis and get some more measurements and listening done.

I found this old thread about diode matching but in the end matching diodes was unnecessary. Just picking 1N4148 diodes randomly resulted at worst little bit over 1 mV mismatch in the circuit. With 1mV mismatch the signal to thump ratio is similar to well matched vari-mu compressor.

Here is another old thread about diode bridge compressors with some good info.

 

[living sounds]

Very interesting! How does it sound?

 

[Script]

How is noise floor ?
I recognize those VUs

 

[Heikki]

Very interesting! How does it sound?

I haven’t done much listening yet.

How is noise floor ?
I recognize those VUs

Noise is somewhere around -60 dBu at 11dB gain. Now that it is just open in a table it is picking up some hum and buzz. I’ll make better measurements after I get it inside a metal box.

 

[abbey road d enfer]

It seems there is a renewal of interest for diode bridge compressors these days.
One of the big issues is noise. Did anybody look at the possibility of using a string of several diodes in series? Seems to me it would result in less gain needed from the gain make-up amp, and probably solve the issue of diode matching.

 

[dogears]

It seems there is a renewal of interest for diode bridge compressors these days.
One of the big issues is noise. Did anybody look at the possibility of using a string of several diodes in series? Seems to me it would result in less gain needed from the gain make-up amp, and probably solve the issue of diode matching.

going from 1 to 3 diodes in series on each part of the bridge reduces the loss by ~6 dB. You think that matters so much?

Isn't part of the desire of the loss to arrive to a useful area of the diode's turn-on characteristic in term so signal amplitude?

 

[abbey road d enfer]

going from 1 to 3 diodes in series on each part of the bridge reduces the loss by ~6 dB.

It should be 9.5dB.

You think that matters so much?

Gaining 9dB on S/N ratio doesn't seem negligible to me.

Isn't part of the desire of the loss to arrive to a useful area of the diode's turn-on characteristic in term so signal amplitude?

Yes, but it results in a very low audio signal, which requires much gain for recovery.

 

[dogears]

I suppose the loss depends on where you bias the diode string.

 

[abbey road d enfer]

PRR gave a good explanation in one of the links in the OP. The allowable signal across the diodes is nearly constant, whatever the bias point.

 

[Bo Deadly]

If you can work out how to put the diodes on the inverting input of the op amps, performance might be drastically improved because it would be current based and not voltage based.

 

[dogears]

I feel like I'm missing something. As you increase the bias voltage the diodes turn on one at at time. So depending on the bias you may get more or less signal loss with additional diodes in series (some on, some off).

 

[abbey road d enfer]

I feel like I'm missing something. As you increase the bias voltage the diodes turn on one at at time. So depending on the bias you may get more or less signal loss with additional diodes in series (some on, some off).

This is not how it works. There is no turn-on. Diode bridge attenuation uses the logarithmic I/V characteristics of diodes.

 

[gyraf]

..could we do something like the 1176's "-LN"-mode - subtracting a bit of the output AC signal from the bias voltage to linearize it for a higher working level? (just thinking out loud..)

 

[abbey road d enfer]

I made a simulation with the standard 4-diode bridge, an 8-diode and a 16-diode.
The signal-to-noise ratio for constant THD follows exactly the number of diodes, i.e. 6dB better for the 8 and 12dB for the 16.
There is no significant penalty regarding the side-chain, since the control voltage does not exceed a few volts (much less than 15) even with the largest number.
BTW, thumps could be reduced by using balanced control voltage.

 

[Heikki]

6dB increase in signal-to-noise ratio costs about 20 cents. I think we can justify that in DIY since we can't buy that kind of performance even for 3500€. I think thumps are really a non-issue since it seems like modern diode manufacturing makes near identical diodes. Looks like richiyobs didn't match the diodes in his compressor and there were no audible thumps. I don't remeber the exact level of worst case thumps I measured but it was low enought for me not to worry about it.

 

[Script]

Nice ratio. If it held up on the bench, it would justify trying 4*3. But I guess there's a limit at some point ?

 

[abbey road d enfer]

Nice ratio. If it held up on the bench, it would justify trying 4*3. But I guess there's a limit at some point ?

The limit is the next 6 dB takes a total of 32 diodes, and the next one 64. You quickly end up with a whole board populated with diodes.
Also there is a time where +/-15V rails cannot deliver the control voltage that's needed.

 

[Heikki]

I made a mistake when measuring the output noise. I had not limited the bandwidth to 22 kHz. With 22 kHz bandwidth the output noise is -68 dBu at 11 dB gain.

 

[richiyobs]

Woooow finally something serious.... Very happy....
I also used matched 4148... And the performance was better... The breath at fast setting smoothed.... Hope it helps
I was busy with an old otari beast...
Best
Richi

 

[Script]

Would 1N4448 offer any advantage here ?