Transistor matching

Just about to build Jig for testing BC550/BC560 as used in API opamps - it's the test jigs on Bauman's site that uses 741 opamp and 2N3904/2N3906. got the opamps but can any suitable transistors be used in the jig - ie BC550/BC560 - just want to make the jig ASAP

Cheers

Do you mind posting a link to the schamtic you are referring to? I don't have the whole net in my mind... :wink:

Anyway, I suggest that you don't match the input transistors. For almost all audio applications 1 mV offset is not required, and trimming offset current is pointless because we don't have equal DC resistances at both terminals. Matching does not influence the sound of an opamp, just it's DC precision.

Get a free sunday morning instead.

Samuel

Not sure how you link sites or post pictures which I need to do.
Maybe there's a BIG clue in me saying Baumans site and API and transistor matching - I'm sure all web sites wouldn't contain all these criteria! :!:

[quote author="Samuel Groner"]Do you mind posting a link to the schamtic you are referring to? I don't have the whole net in my mind... :wink:

Anyway, I suggest that you don't match the input transistors. For almost all audio applications 1 mV offset is not required, and trimming offset current is pointless because we don't have equal DC resistances at both terminals. Matching does not influence the sound of an opamp, just it's DC precision.

Get a free sunday morning instead.

Samuel[/quote]
Hmm, would life indeed be that easy ? The DC-specs are indeed less interesting for audio, but I've always lived in the assumption that a mismatched input-pair would in principle give a different distortion-signature.

and trimming offset current is pointless because we don't have equal DC resistances at both terminals.

Click to expand...

Could be made to, but indeed it can't always be done. But then it's handy to have high beta devices for the diffpair. And also for noise.

Can't imagine the various 'DOA's' use that expensive LM394 only for fun.

Get a free sunday morning instead.

Click to expand...

Fully agreed, we shouldn't overdo things. Maybe we could spend the time saved on what started it all, misic :wink: But I feel some selection for the discretes for the DOA-input-pair would still be required. (Although I haven't compared a 'selected' versus a 'just throw it in'-version)

Note this is not meant as criticism, but I'm joining the discussion for a better understanding of the topic of how relevant those matching efforts are.

Bye,

Peter

I've always lived in the assumption that a mismatched input-pair would in principle give a different distortion-signature.

Click to expand...

Well if you use a tiny SMD transistor in one leg and a fat TO-3 in the other there will surely be a rise in 2nd harmonic distortion. However the distortion-generating exponential V/I law is very predictable and basically the same for all bipolars (as far as I know...).

What needs to be equal is collector current and--far less important--collector voltage. But this is mostly given by topology and not by carefull matching.

Can't imagine the various DOAs use that expensive LM394 only for fun.

Click to expand...

Surely not. Find me a discrete NPN with similar low noise and high beta? Yeah, there are some from Japan nowadays, but not when the 990 was developed.

I don't want to say that matching is pointless in any case. If you use DC servos it is very helpfull to have good DC precision to start with. And I'm myself the perfectionist who rather matches his transistors instead of getting the free sunday morning. :grin:

But I suspect that many people matching transistors are not aware what matching does and what it doesn't. And I doubt that many people have the patience to do tests that are worth the effort. Collector current and temperature are critical. The former needs a thorough understanding of the circuit to be built and the later is not easily controlled. And there is still drift and ageing that sets a firm upper limit to how much matching helps.

Samuel

Good points there, Samuel..!

[quote author="Samuel Groner"]

I've always lived in the assumption that a mismatched input-pair would in principle give a different distortion-signature.

Click to expand...

Well if you use a tiny SMD transistor in one leg and a fat TO-3 in the other there will surely be a rise in 2nd harmonic distortion. However the distortion-generating exponential V/I law is very predictable and basically the same for all bipolars (as far as I know...).[/quote]
Hmm, it'd surpise me that the difference needs to be that extreme to see at least some influence. :roll:
But as said, I haven't compared yet with deliberately unmatched inputs to see what's actually happening.

What needs to be equal is collector current and--far less important--collector voltage. But this is mostly given by topology and not by carefull matching.

Click to expand...

The 'need' for giving both BJTs of a diffpair the same collector-voltage will of course depend on Vearly (which will be such that it's no direct issue) and then still, there's feedback around this all.
But let's hope that when people do some matching it's then only for Vbe & hFE, and not for even more parameters :wink:

I don't want to say that matching is pointless in any case. If you use DC servos it is very helpfull to have good DC precision to start with. And I'm myself the perfectionist who rather matches his transistors instead of getting the free sunday morning. :grin:

Click to expand...

I thought you had become another person but from this it's clear you're still the same - all good. :wink:

But I suspect that many people matching transistors are not aware what matching does and what it doesn't.

Click to expand...

Fully agreed - I'm sure there'll be many hourse spent (& waisted !) worldwide because of 'better safe than sorry'.

And I doubt that many people have the patience to do tests that are worth the effort.

Click to expand...

True, just doing 'some' matching could very well give a false impression.

Regards,

Peter

Hm, it'd surprise me that the difference needs to be that extreme to see at least some influence.

Click to expand...

You're probably right that a very linear opamp topology with well balanced input stage will show a tendency for minor deviations. However most (actually all I have seen, I'd say) topologies used for DOAs seem to have other linearity problems that come first.

For my defence I have to add that a simple Vbe match will not provide very good distortion cancelation in the input pair as it tests the transistors at one current only. You'd need to select them on a curve tracer. A swiss company (FM Acoustics) claims to do that--oh sure, you pay for it!

Samuel