How important is complementary transistor pair matching in class B amp?

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Is it true that in class B amplifiers the NPN and PNP complimentary pair output transistor should well matched to minimize amplitude differences between the positive and negative going signal? Difference in transistor gain due to difference in hfe will result in amplitude differences between the positive and negative going signal and will also generate high cross-over distortion. I would think that a class A amplifier would be immune to these errors because both transistors are conducting.
 
more info

Hi,

Just finished repairing/refurbishing some adcom amps, not super hi-end but pretty well made none the less...

The rack of npn outputs were all matched to each other very well, the same for the pnp's, but they weren't even close to each other, the pnp parts had twice the beta of the npn's. If this had been true for just one amp I'd have put it down to abuse or something similar, but I worked on a pair of adcom gfa1's and trio of gfa565's, collectively using 38 pairs between them...in every case the pnp's were between 50 and 100% higher hfe.

I guess it depends on your topology, the little JLH 10w class A amp sounds best with matched parts, adcom quite obviously didn't used to care.

Stuart
 
I have come across matched MOSFET pair but not matched complimentary pair BJTs. A work around to this is to use NFB and servo control for DC elimination. It is not a perfect solution but should help to some extend. There are amps out there that claim zero NFB. Wonder how they do that? Hmmm.....
 
could be...

I hadn't considered that there was a selection process going on to ensure the relative gains of the devices, a good point. It seemed intuitive in a fully complementary design that if gain mattered at all that closer to matched pnp to npn would be better.

All the replacement devices I had access to (40-50 pairs) manifested the same relationship, the pnp parts had gains higher then their npn equivalents, at least in the mj15003/4 and 2sd424/b554 sets I was looking at.

I seem to remember than JLH had specific recommendations for his amp if one device didnt match the other...

Stuart
 
My limited personal experience has been not to bother trying to match because you nearly always have to buy more pieces than you use. However, the resulting amplifiers don't appear to me to suffer audibly. This may be something more aesily measured than heard (by me at least).

There seems to be more than one school of thought on this.
A. Don't worry negative feedback will take care of it.

B. When you use parallel output devices you need to match all the NPNs and match all the PNPs in order to avoid "current hogging". This will let you use smaller RE resistors.

C. Match NPN with PNP pairs to minimize XO distortion and amplitude differences. (This seems to be the basis on which a few still cling to quasi-complementay outputs.)

D. Match both ways. Combination of B and C.

My own thinking is that I'll match if it doesn't cost too much in regard to left-over components but I don't believe failing to do so will limit my listening enjoyment. Matching benefits me mostly from a "pride of construction technique" viewpoint. Even, if matching achieves nothing sonicly, it certainly does no harm.

I wonder that there seems to be no discusion about matching driver and pre-driver pairs. If "C" is important, wouldn't this apply to the "upstream" components as well?
 
sam9 said:
A. Don't worry negative feedback will take care of it.

B. When you use parallel output devices you need to match all the NPNs and match all the PNPs in order to avoid "current hogging". This will let you use smaller RE resistors.

C. Match NPN with PNP pairs to minimize XO distortion and amplitude differences. (This seems to be the basis on which a few still cling to quasi-complementay outputs.)

D. Match both ways. Combination of B and C.
My opinion on each of these:

  • A) True, but feedback is not a cure-all. A circuit should be as linear as possible before feedback.
  • B) Worth doing to improve power output and reliability.
  • C) Usually pointless because, as I said before, you can't really match NPN with PNP. Look at the datasheets for complementary transistors and you will see that too many things differ, such as variation in hfe with IC, ft, Vbe, temperature coefficients etc. Thus you can never achieve good symmetry, and even if you did it would only hold for a particular set of operating conditions.

If you really wanted to match pairs in low-power stages, you can buy NPN/PNP pairs fabricated on a single die. That would be the closest you could get to symmetry.
 
Hi Sam9,
driver & predriver are usually biased into class A. Only at very high current output might one of the drivers cut off & its complement runs at hi I to drive the heavily loaded output tr, but some designers might even try to avoid this as well.
If this is the case then could this be the reason for no need to match?
As an aside, Leach talks about some topologies running the drivers in class AB & his circuit avoids this. I cannot get my head around his reasoning?
regards
Andrew T.
 
"C" Is is also out on a practically basis. I've tried it on batched of 2sa1943/2sc5200 pairs and MJE340/350s pairs. hfe's were hopelessly far apart - forget anything else. "B" is the only matching that makes practical sense.

BTW- In the case of L-MOSFETS, www.tech-diy.com offers matched pairs of p-types and matched pairs of n-types for a price comparable to unmatched pieces from catalog sources. In this case it would be suilly not use matched pairs. I've not found a comparrable "deal" on BJT's.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.