pros & cons of current mirror to balance a push-pull ?

[Bigun]

I was noticing the offers on eBay to sell me 'quad matched' tubes - which I assume are for push-pull amplifiers.

And I guess the need for matched tubes is to ensure balanced average dc current through the two halves of the OPT and a desire for less distortion by having tubes that produce the two halves of the audio waveform do so as closely as possible.

And I've seen references to use of a current mirror in the cathodes of the output tubes to enforce balanced operation. But I haven't been able to learn much about the tradeoffs from this approach.

It seems like a nice idea, being able to use un-matched junk box tubes and not have to worry about tube ageing etc. But I don't see it being used much. So why wouldn't I use a current mirror for force balance ?

[Sch3mat1c]

It works as a servo, but you can't just plop in a current mirror because it'll ensure that all cathodes carry the same current as the first, resulting in no power output!

A mirror won't match transconductance, so you might get static balance but not dynamic (some tubes "turn on" harder than others).

A better approach is to use circuit topologies which naturally correct errors in the active devices. This is very common in solid state circuits, because, except for a few key features, transistors match very poorly and can only be operated safely with these techniques. I think these types of circuits are uncommon in tubes for two reasons: culturally, everyone copies the same boring old circuits, and circuits with heavy feedback tend to mask "that tube sound". There are minor limitations of having no "PNP tubes", lower gain per device, the requirement of output transformers, and the expense of all that heater power and those extra sockets. It's certainly possible to get comparable performance (distortion and bandwidth) to an SS amp, but the efficiency and size will suck.

Tim

P.S. Goddamnit, I typed a wonderful edit, hit ALT+S and it sent the empty quick reply, not the edit I now have no record of. Why would this designed this way???

[Bigun]

Yes you're right, if the mirror is only operated at d.c. it will balance the current through the trafo but not the dynamic misbalance. I can live with dynamic misbalance as it'll be primarily low order even harmonics which i would likely find acceptable and they'd exist no doubt even with well matched tubes to some extent.

I think you're also right that the current mirror is best implemented using solid state.

But having the dc balance would seem to be a no brainer.

P.S. I've lost long posts before, it's better to write them in something else and then copy-paste them into the forum.

[DF96]

Current mirrors typically have a low impedance side (for input) and a high impedance side (for output). This is so the input side can accept any current without fighting back through voltage, and the output side can produce that current while ignoring the load fighting back. A balanced current mirror (which is what you need for a P-P output) would be a poor current mirror. Unless I have missed something it seems a daft idea.

[Bigun]

I don't know that you've missed anything as there's always plenty of daft stuff going on with some amplifiers.

Here's one reference I have found that shows the idea: Looking Glass


Edit: also some thoughts in here: http://www.audiodesignguide.com/my/sepp.html

[Shoog]

I have used a current mirror under a differential 5687 preamp driving an output transformer. It is of the 3 transistor type to ensure high impedance in both legs. The only major issue with it is that it requires cathode bypassing caps.
So far its been running for six months and is keeping perfect current balance. Might be a few more years before a significant current imbalance would have naturally developed so until that time I cannot really say if it will be a good long term solution. Still I cannot see any reason why it will significantly degrade over time.

Sonically, to my ears, there are no down sides beyond the cathode bypass caps which would have been present anyway. I initially ran it with cathode resistors for a few weeks - but can't remember any noticable change in sound when the current mirror went in.

My implementation looks very like the "looking glass".

Shoog

[DF96]

OK, I see what you mean. It won't reduce the need for matched valves, because you are simply swapping the 'define voltage, get a current' of fixed bias for 'define current, get a voltage' of CCS bias. To use unmatched valves you either need separate cathode resistor bias, or individual fixed bias.

CCS bias of output stages always carries the disadvantage that it stabilises the wrong thing: average current instead of quiescent current. Not necessarily a major problem, but worse than resistor bias.

[Shoog]

Originally Posted by DF96

OK, I see what you mean. It won't reduce the need for matched valves, because you are simply swapping the 'define voltage, get a current' of fixed bias for 'define current, get a voltage' of CCS bias. To use unmatched valves you either need separate cathode resistor bias, or individual fixed bias.

CCS bias of output stages always carries the disadvantage that it stabilises the wrong thing: average current instead of quiescent current. Not necessarily a major problem, but worse than resistor bias.

It depends on your objective. Perfect DC current balance is the objective I was seeking to achieve. Its works very well - which is more than good enough for me. I don't necessarily share your concern about average current vs quiescent current - unless there is a strong set of empirical evidence to show that it has a significant impact on performance. Life is about choosing your compromises.
The end of life situation where the valves fall grossly out of balance is something we all have to face with whatever design we choose, but I don't think it should be the primary deciding factor here.

Shoog

[DF96]

OK, if DC current balance is what you are after then this will do it. Does not necessarily mean AC balance or distortion cancellation.

[Shoog]

Originally Posted by DF96

OK, if DC current balance is what you are after then this will do it. Does not necessarily mean AC balance or distortion cancellation.

The advantage of been able to use a none air gapped OT has to be considered as well as basic output tube distortion. The Piltron transformers are so amazingly superior in their frequency response etc that are likely to lower overall distortion in many situations - but they need perfect DC balance to achieve this. Many people avoid them because they are not prepared to enforce perfect DC balance.

Something worth considering if you have some voltage to spare is Garter Bias - my experience is that it works almost as well as current mirrors but has better AC performance.

Shoog