If you put new output transsistors, class A/B, will tou match them?

[ostie01]

What I mean is, how important it is to match (hfe) output transistors on a class A/B power amplifier, if they have emitter resistors.

[KISS]

Very. Suppose 2 transistors are in parallel. One has an Hfe of 10. The other 110. One will carry most of the current, heat up and you won't be able to compensate for thermal runnaway.

i.e. Repeat
If it heats up, it conducts more then it heats up and conducts more...
Until failure

[john_ellis]

hi ostie

two issues here.

first is if you have parallel output transistors.
KISS meant these. I think. Emitter resistors should allow for mis-matched transistors, but to do this theyhave to be big enough to even up the current even if mismatched.

But often a low gain transistor may require a larger Vbe to turn on to the same current as a high gain transistor. The voltage across the emitter resistor has to allow for Vbe as well as gain and so on.

If you purchase several devices at once, you may find they are from the same lot, from the lot number on the package. If so, you may well have die from the same silicon diffusion, and have closely matched gains that modest emitter resistors are adequate.

I think you meant whether the NPN/PNP have to be matched. The answer to this is yes. Self has said that manufacturers often do not bother. This is, he pointed out, because a high open loop gain can handle non-linearities like gain variation in the output stage.

I disagree, and always try to match transistors. Even a "blameless" amp will give lower distortion if the NPN/PNP driver and outputs are matched. It is not always possible to get a good match because NPN and PNP's never come from the same wafer and the next best thing is to try to buy different lots and search. Or maybe, diy-er's might be willing to list gains of their devices and be willing to swap ...

cheers
John

[anatech]

Hi John,
I completely agree with you with regard to matching.

Also consider the crossover point in the waveform. Your emitter resistors will not generate enough voltage drop to reinforce current sharing. The outputs will turn off at different times.

Some transistor compliments match closer than others. This is where the Japanese parts were superior to the US ones. They were faster too (as everyone knows) with a flatter gain vs current curve.

-Chris

[AKSA]

Strongly agree with John and Chris.

Only the parallel devices need be matched; you are not advised to match pnp with npn - that's very difficult, as John remarks, it will ensure each half of the output stage has the same transconductance, but in my experience it's difficult to find NPN/PNP beta matches from standard complementary pairs.

Think on what's happening here - parallel transistors share the same bias voltage (usually from one driver per side), and since in an AB amp we might only have a bias around 60mA using a 0.22R emitter resistor, a typical value, the voltage dropped at idle is only 13.2mV. If the base emitter voltage is 600mV, as an example (and they all vary slightly), then 97.8% of the bias output voltage is dropped across the transistor input junction.

This tells us that if another transistor sharing the same input bias has a base emitter voltage of 605mV, only 8.2mV will be dropped across its 0.22R emitter resistor, and this corresponds to a bias level of only 37.3mA, a very substantial difference compared to 60mA passing through the '600mV' transistor.

The solution to this problem is to match beta within about 5%, and match Vbe AT THE SPECIFIED BIAS CURRENT within 1mV. This is my solution, and in the above example though using 0.47R emitter resistors all parallel transistors normally check out at 60mA +/-3mA, which is acceptable.

You will need ten transistors from the same batch for two pairs of matched transistor for both beta and Vbe. You need around 15 for one triple, and around 25 for a quad.

I take up a batch of say 100 devices, and then at 60mA and a fixed collector emitter voltage (I use 6V) I measure the base emitter voltage with an accurate DMM. I then grade devices into Vbe bands. Once all the devices are graded, I pick up one pile, say at 600mV Vbe, and then measure their betas at constant current and voltage. If within 5%, I match them up. I go through all my piles, and then match each pile. It takes about 40 minutes to match up 40 pairs.

You can buy matched pairs and triples for some output devices, as I understand, but you pay a premium. I use 2SC5200/2SA1943 from Toshiba, and buy by the thousand. These are very good, tight tolerance transistors.

Anecdotal evidence suggests that open loop distortion of devices so matched is around ten times less than unmatched. This has profound effect on the final, closed loop THD.

Cheers,

Hugh

[anatech]

Hi Hugh,
On Semi has mentioned that in their datasheets for audio output transistors. It seems everyone agrees.

-Chris

[ostie01]

Thanks guy to take some of your precious time to answer.

This is because I had a discussion with another member on another forum and he said that this is not important to match output pair.

For me, I think this is important and I always try to match them, at least for the Hfe. But would like to know if I'm doing this for nothing,
matching NPN and PNP

Thanks again.

[anatech]

Hi ostie01,

But would like to know if I'm doing this for nothing,
matching NPN and PNP

No. I do this as well, and have for years. The closer everything can match up, the better the amp operates. As Hugh mentioned, this is sometimes impossible (well, most of the time) so just do your best.

-Chris

[AndrewT]

Hi,

The solution to this problem is to match beta within about 5%, and match Vbe AT THE SPECIFIED BIAS CURRENT within 1mV.

I absolutely agree with AKSA.
The more important matching is Vbe.

[john_ellis]

Hi all

seems everyone here agrees matching is a good thing.




cheers
John