Common mode feedback

[dirkwright]

So, back in the 1930's, researchers had a problem. They wanted to measure skin voltages on humans but because of the very large hum fields and the very small voltages involved, ordinary amplifiers could not work. So, through experimentation they devised the differential amplifier and invented common mode feedback.

I thought I'd play around with this idea using tubes, pretending I was back in the 1930's and trying to measure microvolts in a high voltage environment. This circuit has a gigantic common mode rejection ratio, as well as high gain of almost 70dB and distortion of less than 0.003%. The common mode rejection is so high I'm not sure it's correct in my calculations. With a common mode input of some 7 volts rms, output voltage is 340pico volts rms (340X10^-12). I think that's 200dB, but I'm probably wrong.

The 100% common mode feedback is the wire between the joined cathodes of the two differential amplifiers. This would be a great gain stage for something like a moving coil phono cartridge, if the noise isn't a problem. The CCS are set for an internal impedance of 1M ohm. Disconnecting the common mode feedback increases the distortion by about 10 times. Bandwidth is not great, with the -3dB point at about 50kHz. Different tubes may give different results.

In the attached figure, the voltage generator source is wired up for common mode input. The output can be seen as 343pV rms.

Anyway, just thought I'd share.

[Osvaldo de Banfield]

Very interesting circuit, but 200 dB seems too much.

Thanks sharing these stuff.

[tomchr]

Neat... I haven't mucked with common mode feedback with tubes myself, but in the semiconductor world, CM feedback is a known source of trouble. CM feedback loops are notorious for being unstable - or difficult to make stable. Just sayin... Beware.

Nice circuit, though.

~Tom

[DF96]

You did make sure that each triode has different characteristics, didn't you? Otherwise all you are seeing may be floating point rounding errors.

[dirkwright]

So, I added a couple of followers so this little preamp can actually drive something. With 0.7mV rms input, output is over 2 volts rms with distortion @ 20kHz of about 0.002%. I think that's pretty good for no global feedback. Bandwidth is still pretty good with 70dB of gain and -3dB @ about 40kHz. I changed the CCS a little also.

This is an ideal situation with matched tubes and parts. Real world performance would be less of course.

Someone could put a balanced passive RIAA filter between the second gain stage and the follower stage for a pretty nice phono preamp. I would hook up the phono cartridge in a balanced line arrangement instead of the usual single ended one. I wrote an article about that 20 years ago. You just take the + and - from the cartridge for each channel separately and floating from ground. Simple.

[dirkwright]

Originally Posted by DF96

You did make sure that each triode has different characteristics, didn't you? Otherwise all you are seeing may be floating point rounding errors.

No, this is idealized fantasy world simulation. I haven't built any tube gear in almost 20 years. I'm just enjoying "fantasy high fidelity", where I pretend I have the ideal system and I can actually hear out to 20kHz.

[dirkwright]

Originally Posted by tomchr

Neat... I haven't mucked with common mode feedback with tubes myself, but in the semiconductor world, CM feedback is a known source of trouble. CM feedback loops are notorious for being unstable - or difficult to make stable. Just sayin... Beware.

Nice circuit, though.

~Tom

Thanks! I'm impressed that the distortion is so low without global feedback.

[dirkwright]

Originally Posted by Osvaldo de Banfield

Very interesting circuit, but 200 dB seems too much.

Thanks sharing these stuff.

Yeah, it's all about ideal parts and such. I doubt a real world circuit would perform this well, but it's nice to imagine it would do that.

[smoking-amp]

Are the simulator tube models simple 3/2 power triodes? Very interesting that it reduces distortion. I would guess that the 2nd stage normally has more tail voltage variation with it's bigger signal, so that variation is increasing the tail V variation for the 1st stage here too.

edit:
What happens if you move the CCS over all the way to the right now so that all three stages share the same tail? Or would that cause common mode positive feedback?

[DF96]

I'm impressed that the distortion is so low without global feedback.

The first two stages have heavy feedback, and the CF has heavy local feedback. Hence low distortion.