Matched power transistors in large packages are rare, or don't even exist at all?

[treez]

Hello,
I have been seeking matched PNP BJTs in packages bigger than SOT23.
However, I cannot find any.
Do they exist at all?

Is it harder to match transistors that are in the bigger packages?

These matched PNPs by nxp.com are only in tiny packages..(page 3)....

**broken link removed**

 

[PA4TIM]

They made plenty of them in the past, but all uin throughole version. I think they had to much competition of the opams. I used a few in the project I'm working on,nmbut they came from my own stock.

 

[treez]

Yes, and unfortunately designs like the following are using transistors Q4 and Q5, which are "supposed" to be matched. (they are sot23).
These transistors are nowhere near enough rated for the power that they may see, but, presumably because bigger_package transistors are not sufficiently matched, they have had to use these SOT23's for Q4 and Q5?

So it looks like any BJT bigger than sot23 cannot be made in sufficiently closely matching batches?


**broken link removed**

 

[dick_freebird]

Matching depends on assumptions like temperature
being equal - a bad bet in a pair of power devices
if collector voltage or current aren't similar. Like
the reason you don't parallel bipolar power devices,
even identical ones, without adequate ballasting.

I think this concern is why you don't see much of
this kind of product, in the power space.

Maybe you just pick a high volume running part,
buy in batches, and use a bit of emitter degeneration
to enforce a tolerable degree of matching?

 

[crutschow]

Why do you need matched power transistors? Perhaps there's a work-around.

 

[treez]

Why do you need matched power transistors? Perhaps there's a work-around.

unfortunately, Q4 and Q5 in the following need to be matched...
**broken link removed**

Maybe you just pick a high volume running part,
buy in batches, and use a bit of emitter degeneration
to enforce a tolerable degree of matching?

...well I tend to agree, this appears to be the only choice, but I cannot understand why ti.com, in the above link, have just used tiny sot23's...these will be overpowered...does any reader know why they have used these?

 

[keith1200rs]

I don't see why Q4/Q5 are expected to be matched - the circuit has emitter degeneration resistors to help balance any mismatch. Matched transistors are pretty rare nowadays. Except for a few specialist applications there are other designs that can be used instead.

Keith

 

[treez]

if they are not matched, the following can happen.

 

[keith1200rs]

I know what the effect of mismatch is. Where do you get the assumptions for Vbe from? I would expect transistors from the same batch to be better matched than that, although as you are trying to use a parameter of the transistor that is not controlled for your purpose there will always be a risk. However, based on your assumption of the mismatch you can adjust the degradation resistors accordingly so in you case your can increase the value.

I know people who 'make' their own matched transistors by buying a load and measuring them. As already stated, there is usually a work-round. First you need to define your requirements specification then design to meet it. If you current tolerance is 30% then you design to meet that. If it is 2% then you design to meet that. Trying to massage a design based on one set of requirements to meet a tighter specification is not necessarily the best design approach.

Keith

 

[123jack]

Matched pairs used to be required in the very early switched mode power supplies. These days that isnt such a requirement of course.
If you cant find a suitable matched pair in the package you need - if you have space you may be able to cut and bend the
legs of a bigger package - I did this once to good effect years ago.

Looking through my TITS it says a BCX19 is a replacement for the one you list. That gives you extra to search for maybe.
Anyway I can suggest **broken link removed**
In their selection guide page 65 are matched transistors - You'd need to check the specs. for a suitable device.

 

[treez]

even if the mismatch is less than that, the sot23 soon gets above the 100mW dissipation level with mildly different vf strings, so I am very surprised to see such a circuit as this ti.com one.
That mismatch could well happen as you allude. If the RHS string is lower voltage than the LHS string, then the RHS BJT will stand off more voltage and get hotter, thus its vbe will be less for a given Ic.
So thermal runaway works against us, even though the emitter degeneration tends to work for us.

Thanks for the nxp link , though they are all small packages, -smaller than sot23

 

[iimagine]

Speaking of alternative, how about this:

 

[treez]

Thanks, yes, that's always an option....a little too dissipative for the goings-on here , but always nice to see that circuit.

Other options are in this thread at post #5

Post #5 of this thread
https://www.edaboard.com/threads/304497/