Cancelling Harmonics

[dsavitsk]

Perusing old threads, I came across a discussion of 300b drivers. One of the assertions (my paraphrasing which may well be incorrect, so apologies to the discussants) was that high distortion in the driver was actually not all bad because it cancelled some of the distortion in the output valve -- being out of phase with each other.

This raises two questions to me -- the first is that, won't the second valve (the driven valve as it were) produce harmonics of the harmonics making this all a bit for naught, or are those of such low level that they really aren't audible?

And second, assuming that the first is not important, does it make sense to drive a valve with itself so that the spectrum aligns and cancels more totally? This suggests to me that a medium mu valve that can put out a couple of watts might be ideal to drive itself rather than driving a low mu with a high mu. So, does such a tube exist?

[Tweeker]

Distortion in a good 300B shouldnt be all that high in the first place, so even if they did cancel, you wouldnt want driver distortion to be high.

300B distortion vs load and current from WE datasheet.
All curves at 350V anode, with grid bias:
1: 30ma
2: 40ma
3: 60ma
4: 80ma

[rdf]

Authors usually mean second order harmonic cancellation in the articles I've read. If I'm right, under ideal conditions even orders cancel and odds add. My sense is the best that can be achieved in the real world is a driver with sufficient 2nd harmonic content to cause cancellation with the output but with few to no harmonics from 3rd and up. I've played with the technique using a 6CN7 front end driving a beam triode cathode follower into a transmitter tube and was able to get 0.6% at 13 watts SE without feedback. However 3rd harmonic comes up fast at higher powers.

[Sheldon]

Some relevant reading: http://www.tubecad.com/articles_2001...ion/index.html

[Wavebourn]

It would be great to have "A tube" and "Anti-tube" so they compensate each other.

Unfortunately, it is impossible. What is possible, to use the shortest linear part of a transfer characteristic. But it is problematic when big power amplificastion is needed: both in microphone and power amplifiers, that's why it is so challenging, so people still continue developing both power and microphone amplifiers, repeating the same mistakes, reinventing wheels...

[rdf]

Originally posted by Wavebourn
Unfortunately, it is impossible..

I can understand 'impractical' or 'ineffective in practice' but why impossible?

[Wavebourn]

Originally posted by rdf


I can understand 'impractical' or 'ineffective in practice' but why impossible?

Say, to compensate the 2'nd order you need one device with a member x^2 in a transfer function, and another one with x^-2



No antiworld exist to compensate world.

However, you can always find an equilibrium and trade off something less significant for more significant. This is the art.

[pmillett]

Great subject.

As was said, for the most part even harmonics cancel and odd add. Normally 2nd harmonics dominate in a SE stage, so you can somewhat cancel the 2nd and to some degree the higher orders. My experience has been that it's mostly the 2nd that you can cancel significantly.

I guess theoretically the second stage would multiply up the harmonics (i.e., and 2nd order present will generate some 4th order in the subsequent stage), but unless the amount is very large, it drops into the noise fast. 1% of 1% is .01%, below the noise for many/most tube amps.

The problem with cascading like stages is that to get the same distortion characteristics they'd both have to run at the same bias point and same input/output voltages. So it's isn't very useful. Of course you could bridge the load between two like stages, and then you have a push-pull amp.

What I've tried to do with some sucess is just find a driver that has similar harmonic characteristics when operating at it's normal point as the output stage. That increases the cancellation. But you need to be careful that the harmonic content doesn't shift around as the output power goes up; I've done some experiments where 3rd is dominant at low power and then 2nd, then 3rd again as you approach clipping. Doesn't sound good to me.

In my opinion, you don't want to cancel so much of the 2nd that the 3rd becomes dominant in any case until you start to get to clipping.

Pete

[rdf]

Hmmm, I've done it on the bench, once accidentally, and somewhere at home have a Spice sim showing complete 2nd harmonic cancellation with an admitedly rather useless and artifical circuit arrangement built around two 12ax7s (as I recall.) I can post it after work.
My understanding is it works because tubes generate 2nd harmonic in anti-phase with the input. The 2nd harmonic generated in the second tube cancels with the 2nd harmonic 'received' from the 1st, so obviously it's related to the 2nd stage amplification factor as well. I agree with pmillett though that for the most part it's not tremendously useful, or at best done in moderation. To be truly effective the injection of 2nd harmonic distortion of both tubes have to track closely with level across a wide portion of the operating range. The odds of finding the right combination of devices and circuit seem to me pretty low.

[ilimzn]

True, unless you are using the same tubes for both stages, and the same working points. Of course, this is rather useless in a power amp - there you are condemned to attempts of finding the right combo of driver and driven tube, plus the right combo of working points. In a preamp, you can achieve cancelation relatively easy, by using a common cathode amp followed by a common anode (follower) amp, both the same tubes, DC coupled, with working point chosen to get the same DC voltage across the tubes and the same DC current through the tubes. Of course, the second you put a load to it the cancelation is gone unless you know the load in advance. Even so, the cancellation is dubious - it is really only the 2nd harmonic that can cancel completely.