Output stage linearization.

[JMFahey]

Has it been tryed to compensate the non linearity of a class B output stage adding an opposite non linearity made of passive components.

Yes, I have, in fact still do in one amplifier type I produce commercially for some 20 years now.
It is a 40/60W battery powered (12.6V gel cell battery) using push/pull MosFets driving a center tapped output transformer.

Since I use "switching" MosFets, I had a terrible crossover distortion problem:

1) they have an abrupt turn on area, quite high and very non linear

2) I am using them as common source gain stages, not the common drain follower configuration which has 100% internal feedback which minimizes this problem

3) it IS a gain stage, but gain is low, thus not having "disposable extra gain" which can be spent on NFB.
I do use end to end NFB, but it´s not that high, doubly so because using an output transformer complicates things.

4) so I had to "precompensate" by adding antiparallel diodes in an Op Amp driving stage so it roughly nulls output stage threshold distortion.

I was sucessful, result is not Hi Fi by any means, but this allows my Voice/Keyboard/Bass/Guitar portable amplifiers deliver unheard of power when fed from a puny 12.6V single rail supply ... no voltage step up/converter involved of course, high efficiency, relatively low idle current (longer battery life).

 

[jan.didden]

Oups,
Forget about gain seen over 1.
This was from an error of my own about a DC offset I had overseen.
With a correct calculation, taking care of the offset, I have a gain in the A region lower than 1 and higher than the gain in the B regions.
So, gains are as expected in a Class AB output stage.
Now, how could I compensate these two gains for a constant gain.
I am thinking of resistor switching.

Actually this is what Quad did in the 405 with current dumping. What they did was to increase the feedback factor as soon as extra gain was produced by the dumpers.
This is a fundamental difference with 'normal' feedback that is a constant factor throughout the cycle.

But do not expect wonders from it. It suffers from the same issues as all feedforward systems suffer from: to do (much) better than NFB your correction circuit (it is a bridge in the 405) must be extremely stable and accurate in ratio.

Jan

 

[mchambin]

I do have a very stable circuit in gain and bias where I tried gain switching acting on the feed back resistors.
No success so far.
I can achieve accurate gain ratio, the issue is at the gain switching where the gain law must fit.
I had a go with a crude switcher, no luck so far.

 

[jan.didden]

I do have a very stable circuit in gain and bias where I tried gain switching acting on the feed back resistors.
No success so far.
I can achieve accurate gain ratio, the issue is at the gain switching where the gain law must fit.
I had a go with a crude switcher, no luck so far.

I think the problem is to decide the moment when t switch the gain. The transfer into and out of gain doubling is not a clean, nice, sharp moment you can 'see'. It is a continuous change from one state into another. There exists not a specific moment where you can say: now there is no gain doubling, and a very short time later, now there is. It is like a sliding scale.

The nice thing about the current dumping concept is that it reacts according to the actual gain at any moment; it also has the sliding scale. I can only recommend to study it.
There are online articles by Peter Walker the inventor, and also some very good articles by John Vanderkooy.

There are some threads here at diyaudio about 'current dumping' but most completely miss the point of the variable feedback factor and think that current dumping is just slapping a class B stage on an amp; it is very much more subtle than that.

Jan

 

[soekris]

I have over time been investigating kinda the same thing, to get acceptable distortion in a power amplifier without using negative feedback.

So far my conclusion is that the best way is to move the output devices operating point up into their more linear operating range, by running them in high bias class AB, which will generate heat and waste power....

All the other different fancy schemes seem to have undesirable side effects.

 

[mchambin]

Thanks I appreciate yours experiences and pointers.
I am aware this not easy.
I am in a projet where I get right away thd 0.0017% at 20Khz 30v peak in 8 ohm using massive negative feedback.
I am trying to get better than that with a more linear output stage.
The thd at low level is superb simply because it runs in class A. The not so good thd at high level is not from Beta drooping; Despite it is at high level it does come from gm doubling.
I proved it to myself running with a DC offset to stay in one of the two B regions.
Results at Vpeak = 15volt 20Khz.
DC offset +21v gives thd 0.00028%
DC offset 0 gives thd 0.0019%
DC offset -21v gives thd 0.00025%

This shows to me that eliminating the effect of gm doubling has the potential of a near 20dB thd inprovement.