Series pass transistors, gain varies a lot. Is this a problem?

[Chris Wilson]

I am here again, sorry! My ongoing restoration of a huge Farnell 60V at 50 Amps linear power supply led me strip out the tunnel heat sink of its ten 2N3055 transistors. 2 are used as a Darlington pair, feeding the other eight. I suspected one had failed as a circuit board had a blown resistor and diode and a fellow owner recognized the component failure from when his similar unit suffered a shorted pass transistor. The labour time to get at these things was quite hefty, so I thought it prudent to buy ten new ones. I used my Peak transistor tester, a DCA55, to check out old and new components. Sure enough one of the old pass transistors proper was shorted emitter to collector. I checked all ten new ones as a matter of course and saw the gain figures on them varied from 28 to 38. One of the old originals had a gain of 55! Does this variance matter in such an application as this? I will collate diagrams and ask later if a emitter / collector short was likely to be the cause of the blown control board diode and resistor, if I may? Thanks for reading. 

The schematic for the pass transistor section is at http://www.chriswilson.tv/H60005.jpg

[AG6QR]

It shouldn't matter.  The gain of a transistor is generally not controlled very tightly, and designers shouldn't rely on a particular specific value of gain.

According to the data sheet, http://www.st.com/web/en/resource/technical/document/datasheet/CD00000895.pdf , in table 4, it says that, at collector current of 4 amps, and Vce of 4V, the gain is somewhere between 20 and 70.  Your transistors all test within this range.  At a collector current of 10A, the gain is specified to be at least 5, with no maximum specified.

This design isn't very sensitive to variations in gain.  If one of the transistors has a lot more gain than its neighbor, its base-emitter current will be less, but who cares?

[calexanian]

As long as you have emitter balancing resistors and the gain is high enough to allow all of the emiters to come up to voltage you are good. Just go down and measure the voltages. If one is way too low then the gain may be too low. You will find in those kinds of circuits things will find their happy place and don't be shocked to see them vary a lot under a light load. Under a heaver load things tend to balance out.

[Chris Wilson]

OK, thanks gentlemen! My Audiophile (Audiofool?) pal is a bit anal over gain figures, but I thought his worries were OTT for this sort of thing. But if you don't ask, and all that.... Appreciate your time and your patience Cheers, great forum, and you are very kind to us beginners.

[T3sl4co1l]

hFE varies by -70/+300% in a given device, over manufacturing and temperature range.  Why is this even a thing..?

2N3055s are comically awful.  The original process probably doesn't exist anymore.  New devices can be orders of magnitude better.  You're likely to get an oscillator by using new production (which could very well be something like MJ15002 that failed to meet that spec).  Be careful and put a nice beefy ferrite bead in series with the base, if the transistors are chassis wired.

Tim

[Chris Wilson]

Hmm, you're worrying me now They were Ebay buys, not the cheapest, but not RCA as per the originals. Logo reads a graphical rendition of  "ST",  Malaysia. Have I created a powerful oscillator....?

Thanks for the replies.

[calexanian]

No worries...  Even if you get some really hot transistors in the darlington pair it makes little difference in those types of circuits. Just toss in the new transistors and you will be fine. Provided the rest of the supply is ok.

[David Hess]

The emitter ballast resistors reduce the variation do to gain and Vbe but it is possible that the manufacturer graded the transistors as well for better current sharing.

When I am rebuilding big power supplies which use multiple output transistors in parallel, I usually buy more than I need (2N3055s are cheap) and grade them by Vbe at high current to weed out the outliers which takes very little time.

[calexanian]

True. If you can find them from the same date code they will most likely be rather close. There is a company called STI that can sell you some of the most kick ass TO-3 transistors made today. They will match them up for you also. I think they are in Florida.