Matching MJ15024/25 and other output transistors...

[Lyra]

Hi.

I have recently get a new job involving service and maintanance of a lot of PA equipment. A lot of really NICE ! equipment I may add. d&b, Crown, QSC, LABGruppen, Chevin, and a lot more

We have a lot of Crown amps, and these sometime get broken despite the good quality, and their reputation of beening able to handle "anything".

Right now there are five amp's (two Crown's) waiting for new output-transistors, and the original Crown-type is rather expensive. Thats why I've been curious in using MJ15024/25 transistors instead.
The question is: How many of each (NPN/PNP) transistors would I need too bee able too match them properly if I use the method described here : http://sound.westhost.com/transistor-matching.htm , and how many categories could I expect ?
Total match is probably out of the question so how big difference would You consider as a good match ?
My thought was to get the best matches, and use the transistors that I can't match in amplifiers that don't need matched outputs (like Studiomaster ++)

Would really appreciate some advice in this matter.
Lyra

[Centauri]

I usually just make sure all transistors are same batch number - I don't really think close matching is going to make much audible difference..

Cheers

[pinkmouse]

Hi Lyra

Whilst most Crown amps are very good, don't go near the Powerbase series. Several people I know just gave up on them and threw them away they were so unreliable. It was just not economical to fix them.

[Lyra]

Originally posted by Centauri
I usually just make sure all transistors are same batch number - I don't really think close matching is going to make much audible difference..

Cheers

The sound-quality wasn't my main consern....it is after all PA-equipment.... but more the stability issue.
.... it would bee nice to split up all the power equally between the transistors, and not in a way that one transistor is operating close to it's max while the others have a "comfortable time".........with the amount of power theese amps deliver this would be BAD !?

This might be one of the reasons that a Crown amp is a lot more long-term stable than for instance Studiomaster..... and niiizzzer sound too

Lyra

[Zero Cool]

I would bet the reason crown charges so much is because the devices are matched...


In another post somewhere i read that 50 is generally a good number to start matching with and that 10% is close enough....but again im quoting what i rememeberd from another post....

Some one else have any suggestions???


Zero

[Centauri]

Originally posted by Lyra
it would bee nice to split up all the power equally between the transistors, and not in a way that one transistor is operating close to it's max while the others have a "comfortable time"

Usually these amps have low value emitter resistors for the purpose of evening out the current sharing. If the transistors have the same batch number, then it is fairly safe to assume their parameters will be very close.

Cheers

PS: I service such pro amps for a living, and have never had any problems with the above approach (not including fake transistors though....)

[jackinnj]

i do a lot of matching up MOSFETs for people -- and some wonder why it costs so much more to get a "quad" or "quint" than a "couple" -- the reason is the "precision" (number of significant places) and the distribution -- in the case of MOSFETs if you want to match to +/- 0.01V you limit the number of candidates in your "bins" -- we will see how it goes with Renesas Lateral MOSFET's -- you could pretty well bet on the Hitachi's to be normal, or at worst bimodal. For the MOSFETs, at least to my experience, there is a strong correlation between Vgs and mu.

The ambient temperature will affect the measurements -- so if you get in a shipment it's a good idea to do all the matching at the same time.

for small signal BJT's there are a couple tests that should be done, i.e. HFE and VBE matching.

You find that in a group of smal signal BJT's the variation of DC beta will range from 97 to 140 in a group of a couple hundred, and the distribution is far from "normal". Doesn't matter if they are ON-Semi, Fairchild, PMC etc.

Here's the MOOG VBE matching routine:


the MCUTracer (it was originally designed for tubes, since tube curve tracers are so expensive) could be quite easily pressed into service to plot the characteristic curves of two semiconductor devices at the same time and calculate the differences. It would just be a matter of rewiring a couple resistors.

[gmphadte]

Read Bob Pease's article to know more about how to...

[Lyra]

OK thanks for the replies. Think I have figured this out !?

BUT....some other questions:

....could someone please describe the difference between MJ21193/94 and MJ15024/25.
When do I want to choose wich transistor ?

...and any idea witch one is the closest match to Crown original ??????


Lyra

[jackinnj]

the difference is in the HFE with the 21193/4 being a bit higher.