How necessary is it to match, or rather to use a true complementary pair in a push/pull output stage? I cannot seem to convince myself that this is necessary for good performance.
My reasoning: if complementary pair A-npn/A-pnp work well together, and complementary pair B-npn/B-pnp work equally well (i.e., performance wise, e.g., distortion, noise, etc.), then why can't I use A-npn/B-pnp combination?
Would I be comprimising some linearity criterion? What are your practical experiences?
Someone please help me understand.
Regards
Shaun
My reasoning: if complementary pair A-npn/A-pnp work well together, and complementary pair B-npn/B-pnp work equally well (i.e., performance wise, e.g., distortion, noise, etc.), then why can't I use A-npn/B-pnp combination?
Would I be comprimising some linearity criterion? What are your practical experiences?
Someone please help me understand.
Regards
Shaun
I'm sure someone else wil chime in, but generally true complementary pairs don't exist. However, manufacturers have pairs that, for the most part, have similar specs and (more importantly) act in similar ways when thermally heated. A lot of things change when an amp heats up, and it is advantagous to have output pairs responding in similar ways to the changing environment.
I'm sure there are other reasons, but it's late and my brain hurts from an evening of amp troubleshooting...
I'm sure there are other reasons, but it's late and my brain hurts from an evening of amp troubleshooting...
Your logic is suspect.
2SA1302/2SC3281 is a nice pair, so is
MJ15022/MJ15023
But it would be foolish to mix and match, and why would you want to?
Anybody remember the early days of silicon?
There were actually amplifiers made with one PNP germanium output and one NPN silicon output.
Sounds like a pregnant mule to me.
2SA1302/2SC3281 is a nice pair, so is
MJ15022/MJ15023
But it would be foolish to mix and match, and why would you want to?
Anybody remember the early days of silicon?
There were actually amplifiers made with one PNP germanium output and one NPN silicon output.
Sounds like a pregnant mule to me.
Still wondering...
Hi
Thanks for your replies, but it is still not clear why mix 'n matching is a bad idea. For example: quasi-complementary's work- well I know that mixed complements will work - but I need to know why mixing them is generally frowned upon, and what the practical implications are.
EchoWars - your point seems to make some sense, so point taken. Any other reasons? I would like to learn from people's practical observations.
BTW, let's assume just one transistor per rail.
Regards
Shaun
Hi
Thanks for your replies, but it is still not clear why mix 'n matching is a bad idea. For example: quasi-complementary's work- well I know that mixed complements will work - but I need to know why mixing them is generally frowned upon, and what the practical implications are.
EchoWars - your point seems to make some sense, so point taken. Any other reasons? I would like to learn from people's practical observations.
BTW, let's assume just one transistor per rail.
Regards
Shaun
Re: Still wondering...
Hi Shaun,
I wouldn't go back to the days of the Crown D-150 amplifier with quasi-complementary design.
What I understand from this link fig.104 is that quasicomplementary design is only necessary if the high ft PNP powertransistor required is not available.
http://www.linear.com/pub/document.html?pub_type=app&document=50
Shaun said:
Hi Shaun,
I wouldn't go back to the days of the Crown D-150 amplifier with quasi-complementary design.
What I understand from this link fig.104 is that quasicomplementary design is only necessary if the high ft PNP powertransistor required is not available.
http://www.linear.com/pub/document.html?pub_type=app&document=50
With enough feedback and frequency compensation about anything can be made to work.....
But in my example of above, while both are 200V 15A parts, one has a gain of 100 and a Ft of 30Mhz, the other a gain of 15 and an Ft of 4Mhz. I think I would rather make an all PNP or all NPN amplifier rather than use one of each of these.
Of course if one were to use the slower transistor as a constant DC current source.....
http://www.passdiy.com/pdf/classa_amp.pdf
And there is always the circlotron needing only one polarity output device, and can easily be adaped to the X Super-Symmetric amplifier scheme
http://www.delphion.com/cgi-bin/viewpat.cmd/US04229706__?MODE=fstv
Or maybe I just don't understand your question?
But in my example of above, while both are 200V 15A parts, one has a gain of 100 and a Ft of 30Mhz, the other a gain of 15 and an Ft of 4Mhz. I think I would rather make an all PNP or all NPN amplifier rather than use one of each of these.
Of course if one were to use the slower transistor as a constant DC current source.....
http://www.passdiy.com/pdf/classa_amp.pdf
And there is always the circlotron needing only one polarity output device, and can easily be adaped to the X Super-Symmetric amplifier scheme
http://www.delphion.com/cgi-bin/viewpat.cmd/US04229706__?MODE=fstv
Or maybe I just don't understand your question?
Thanks for expanding on that. My situation is not as bad as the scenario you describe, though. I have a couple of 2SC5200's, and some 2SA1302's, and was wondering if I could use them as output pairs.
I asked myself that if two different complementary pairs - when used in the same circuit - gave similarly good performance, what would happen if the pairs were mixed. I could not come up with a satisfactory answer.
I think, now, that if the two are significantly different (as in your example) it would be a bad idea. That much is apparent. But if the transistors are similar (in spite of them being non-complements), that there should be no problem.
What do you think?
I asked myself that if two different complementary pairs - when used in the same circuit - gave similarly good performance, what would happen if the pairs were mixed. I could not come up with a satisfactory answer.
I think, now, that if the two are significantly different (as in your example) it would be a bad idea. That much is apparent. But if the transistors are similar (in spite of them being non-complements), that there should be no problem.
What do you think?
"I could not come up with a satisfactory answer. "
Shaun, I didn't quite understand your question until your last post. I think you are just asking whether the numbers on the devices really matter. Well, no. If you have two pairs that are very similar , ie four similar devices, then sure you can mix them up. It just depends how similar they are. The more similar the better sound you'll get (in general). Global NFB is a little sensitive to these differences so use less of this rather than more if you have the choice.
Shaun, I didn't quite understand your question until your last post. I think you are just asking whether the numbers on the devices really matter. Well, no. If you have two pairs that are very similar , ie four similar devices, then sure you can mix them up. It just depends how similar they are. The more similar the better sound you'll get (in general). Global NFB is a little sensitive to these differences so use less of this rather than more if you have the choice.
matching?
Shaun,
It is often helpful to consider the open loop case, without feedback. Assuming you have a class (A)B amp, one can see that each half of the wave form is reproduced by one of the two output devices. They will both be distorting. If they are matched, it is easier to find a bias point where they connect and overlap best (matched means similar characteristics). All this means that the feedback will have an easier job to reduce the distortions to a low value. If you use wildly dissimilar devices, it will be much more difficult to find a good takeover point, giving the feedback a much harger time with not so good final results. In class A you really have a similar problem, because the decrease of one will be dissimilar to the increase of the other. So, matching improves the end result wrt linearity and thermal equilibrium (as noted above).
Jan Didden
Shaun,
It is often helpful to consider the open loop case, without feedback. Assuming you have a class (A)B amp, one can see that each half of the wave form is reproduced by one of the two output devices. They will both be distorting. If they are matched, it is easier to find a bias point where they connect and overlap best (matched means similar characteristics). All this means that the feedback will have an easier job to reduce the distortions to a low value. If you use wildly dissimilar devices, it will be much more difficult to find a good takeover point, giving the feedback a much harger time with not so good final results. In class A you really have a similar problem, because the decrease of one will be dissimilar to the increase of the other. So, matching improves the end result wrt linearity and thermal equilibrium (as noted above).
Jan Didden
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.