• The move to the new server is done. There are some software and database maintenance updates in process. This has us passing the hat around to help out. We appreciate any donations. Seriously, even a dollar helps. The payment page may be found here - https://www.audiokarma.org/support.html

Testing and matching of transistors

oldvinyldude

oldvinyldude

Moderator
Staff member
Moderator
Subscriber
I am wondering how many transistors does one typically need to purchase, to get a 'matched' pair.
With some amp designs relying on careful matching to achieve proper and balanced performance, I am looking for members experience.

If I need a pair, how many transistors should I need to buy, in order to test and have a closely "matched" set?

A related question-I wondered what transistor tester do folks rely upon for testing? I see several in the $30-$40 range-are these good enough? I am not a shop owner-just a hobbyist.
 
Register to hide this ad
It depends on what you are matching? Small signal or power? I found if you buy a ksa992 on tape the unit side by side Vbe is within mV’s
Matching npn/pnp good luck to that, processes are so fine tuned these days that they do not even overlap to find matches.
I design power Amps and do not even own a transistor tester. For one thing the current that they put out is probably not what you going to run them at in circuit so imo it’s a waste of $. If you want to match run them in the intended app and measure select there.
To match a pair of ksa992 I use the DMM in diode test mode and select Vbe <2mV since the DMM usually runs at 1mA constant current
 
rothwellaudio said:
Depends on how closely you want them to match. Within 1%? Within 5%? Within 10%? :dunno:
Click to expand...

Say you are servicing a receiver, and there is no adjustment pot for idle current, or for Bias. If I understand it, one may need to replace transistors to achieve a suitable value. Being stereo, these transistors should be well matched.
I don't have a % in mind. That's the reason for my question. In the experience of others, how close do they need to be (%), for a good repair. Secondly, how many transistors, in their experience, are necessary to have a pair that meet whatever percentage might be prudent-4, 6, 8, 10?
In this example, I would think they would be considered "driver" transistors (which places them between small signal and power transistor, but I am unsure-trying to learn these things.
 
Transistor circuits are, or should be, designed so that the manufacturing spread of parameters makes no difference. There should be enough feedback to compensate for transistor Hfe variations.
I doubt you'll find any mainstream receiver that requires matched transistors.
 
rothwellaudio said:
I doubt you'll find any mainstream receiver that requires matched transistors.
Click to expand...

Interesting! Thanks

I could swear I have read threads where members have gone to the trouble to choose matching transistors in these applications. Maybe the OCD tendency many of us have-lol
 
It seems that many members go overboard and make assumptions that matching is necessary, when in reality they are not as @rothwellaudio mentions.
If the equipment mfg needed matched or selected devices in the ckt, then they would state so in the service manuals and/or schematics.
Like I said earlier, modern semi processes these days almost guarantee very low spread vs the devices used in the old days, so much so that they do not even offer or only have one hfe grouping.
ksa992 is an example, look at the new datasheet,, "P" and "E" grades are not sold anymore, but the other grades are available.
 

Attachments

Great info gentlemen!

I would still like to have a transistor tester-any thoughts on that item?

Thanks
 
The diff pair is probably the only spot it sort-of matters. For a while there were diff pair packages, two matched transistors in a single package so they were also thermally coupled. When those go, you're basically stuck subbing it with two transistors, ideally stuck together with heat shrink so they are still thermally coupled to some degree.

not a pro by any stretch but I have one of those cheap "tests everything" jobs. Push the button and you get the junction voltage and HFe. Between those two you can find a couple that are pretty close without much fuss. A curve tracer is a better option, there is apparently one of those that isn't dumb money which plugs into the computer to give you the curves. Can't remember the name of the thing though.
 
rcs16 said:
It seems that many members go overboard and make assumptions that matching is necessary, when in reality they are not as @rothwellaudio mentions.
If the equipment mfg needed matched or selected devices in the ckt, then they would state so in the service manuals and/or schematics.
Click to expand...

Everyone knows that some service manuals are more complete than others. In many service manuals there are warnings to readjust the bias when the driver and/or power transistors are replaced. If there is no bias adjustment in the OP's receiver it is a good reason to find replacement transistors that match as much as possible the originals.


rcs16 said:
Like I said earlier, modern semi processes these days almost guarantee very low spread vs the devices used in the old days, so much so that they do not even offer or only have one hfe grouping. ksa992 is an example, look at the new datasheet,, "P" and "E" grades are not sold anymore, but the other grades are available.
Click to expand...

The F flavor has a range of hFE from 300 to 600. That's a very wide range.

In general if the resistors in a circuit have 5% tolerance (typical tolerance in many amplifiers) it is not needed to match transistors with an accuracy better than 5%, even in a differential pair. The most important parameter to match in a differential pair is the hFE, not the Vbe.
 
Those inexpensive "component" testers work well for this. Probably not super accurate, but all you are looking for is very similar values for hFE and Vbe for differential pairs. Transistors mounted on tape for machine assembly tend to be pretty close in value since they are typically from a single large die. Handle them with pliers while checking for match though, hFE in particular is very sensitive to temperature and mounting them in the tester by picking them up will require a few minutes cool down time to get a stable reading. If you have a set of mini-grabbers, leave the transistors on the tape and attach the leads without touching the plastic body. You can write the numbers on the paper tape for later use too if you want/

Long tail pairs need to be close enough you can set the bias and for some amps the center (offset) voltage. If an amp does not have an offset pot, closer matching may be required to get the offset close to zero.
 
ecluser said:
The most important parameter to match in a differential pair is the hFE, not the Vbe.
Click to expand...
Do you not think the two parameters correlate?
Take a look at a old TI or Motorola 2n2920 data sheet as a form of education as how dual transistors are specified.
 
hFE and Vbe don't correlate. Look at any transistor datasheet and you will see that hFE can vary over a wide spread. The datasheet that you linked on post #7 shows variations of 300 to 600 for the KSA992F. Vbe is between 550mV and 650mV at 1mA but it doesn't mean that Vbe is low when hFE is low, and Vbe is high when hFE is high. Over a bunch of transistors, Vbe variations are much smaller than hFE variations. I think you should invest in a tester that can measure hFE and "educate" yourself rather than trying to "educate" others...
 
gadget73 said:
The diff pair is probably the only spot it sort-of matters. For a while there were diff pair packages, two matched transistors in a single package so they were also thermally coupled. When those go, you're basically stuck subbing it with two transistors, ideally stuck together with heat shrink so they are still thermally coupled to some degree.

not a pro by any stretch but I have one of those cheap "tests everything" jobs. Push the button and you get the junction voltage and HFe. Between those two you can find a couple that are pretty close without much fuss. A curve tracer is a better option, there is apparently one of those that isn't dumb money which plugs into the computer to give you the curves. Can't remember the name of the thing though.
Click to expand...
Diff pairs are the only transistors I really worry about matching. And I just buy off tape. They'll be very close.
 
I think there is an NPN one as well but yes that sort of device. Lovely idea, but I don't think anyone makes a direct sub for them.
 
I replaced many 2SA798 in amplifiers and receivers because they were mismatched. When they get old it is not rare that the hFE of the two transistors inside the 2SA798 differs by 30% or more. In some amplifiers, that will produce a lot of noise. The output voltage will swing at random from the positive supply rail to the negative supply rail, triggering sometimes the protection relay. This transistor figures on the list of the "Top Ten Worst Transistors - noisy, failure-prone, whatever... and replacements | Audiokarma Home Audio Stereo Discussion Forums…" for that reason. Dual transistors does exist today but they are so expensive that it is much cost effective to make our own pairs.
 
It matters for diff pairs and for paralleled outputs on big amps, and you might have to buy a dozen to get a good match. Or two. Depends on luck. Note that Crown has a reference on matching transistors for the old amps and they probably don't do it the way you'd think. In the past I've tried to match NPN and PNP output devices. If you can do it, there's a small benefit, but it's near impossible unless you have both luck and a large number of devices. It's not important for bias, as the current travels through both devices- it's never different between the two!
 
does Crown have you do a current sharing test on the outputs? Phase Linear does, basically just check voltage drop across each emitter resistor when its running at some power level into a dummy load and if it differs by more than some %, figure out why. Its either the transistor or the emitter resistor. If the resistor is good, you swap the transistors until its all happy.
 
You must log in or register to reply here.
Back
Top Bottom