TL494 based 82V LiFePO4 charger - Bad part or suspect substitute?

[Riotpack]

Working on a batch job of chargers and 9 down, 1 to go. Why does the last one have to be such a pain? Before powering these 82v lithium chargers up I have a checklist including cold tests and visual checks.

OK, the first problem is one of the 2SC4140 switching transistors. It has the front blown off and there are metal vapour marks on the PCB at the ends of the FR157 diodes across the transistors. No 2sc4140? OK, my closest is an FJA13009TU so in two go and the diodes are replaced as well as D1 & D2 (being substituted for UF4007).

Great all resistors, capacitors (except C9 and C12) and diodes in the section have been tested. Now before powering this up I continue down my checklist and test the FML36S rectifier for the 82v line - Shorted! OK find something which substitutes, insert a RURG3060C and carry on testing - OK a shorted 2SC9013 pulsing the base drive transformer, Q3 or Q4, ok a KSD471 goes in but why did this part go short circuit? better replace the TL494 and LM358N just to be sure, and D5, D6 with UF4007.

Time to power up. The main 82v line goes to around 8v and stays and I measure 0v on the 12V line so where to next? I have tested D10-D13 in-circuit but am looking at them sideways. This PSU has a self start winding, how far can we substitute and still start?

This schematic is close to the circuit but is not the actual diagram.

Powering the TL494 from a 13V supply shows good sawtooth but seems to be a low drive - I may have to find a 24V supply to be sure.

I am thinking measure the input of the 7812 (has been replaced) ? I only have a bitscope mini so an untrustworthy oscilloscope. Any advice? Is there a faulty component still hiding in plain sight or have part substitutions taken the magic self-start away? I am confident with my substitutions but does this circuit seem delicate with respect to component substitution?

[Riotpack]

Measuring the power pins of the TL494 shows 1.7V - Both the 82v rail and the 18V-20V Tl494 supply are around 10% of what they should be

[Riotpack]

Who can guess what I stuffed up with this part KSD471ACYBU to substitute the 2SC9013 - Ill give you a hint, its one of the letters. Both are NPN and match nicely except I ended up with one small variation of the KSD471.

[Riotpack]

OK powers up and runs just fine!! I was scared as it showed 0V first but I didn't have the leads touching properly! So whats the moral? Read the datasheet, not just to match substitutes but to see what variations there are?? Don't add 5 suffixes to your product "ACYTA", I thought the part would be labelled "part - C" not "partACYTA", by the way, that C type in 3 options was the only KSD471 that my supplier stocked, I just assumed it would be the standard variety! Now to reassemble and final test.

[T3sl4co1l]

Well, the U2 threshold stuff is just wrong, but it doesn't affect anything and you can ignore the LEDs.

VR2 should be a fixed resistor, and the variable part should instead be in series with R32, but at least with this type of converter (as opposed to a one-switch flyback, say), output voltage is limited by turns ratio, so it shouldn't self destruct when VR2 is set wrong.

I like that it actually has current sense, inductor current sense at that.  It's not a 100% average current mode controller, but it seems it was compensated to operate smoothly under current limit so that's good.

C14 and C15 will have a hard time, but I suppose that doesn't matter.  Don't load them heavily.

No polarities are shown on the transformers.

There could stand to be some R+C dampening on the output rectifiers, depending on transformer design.  Those are usually seen in ATX supplies.

Oh, F1 is way too fucking big, haha.

I don't get why all the junction diodes are drawn as upside down schottky (except D9, D17), and D15 isn't [ed: oh, but it's a PN junction diode, and not 30A; weird, nevermind], but whatever, no accounting for style...

Weird... SS9013 is the original, well known, small signal BJT of that number (don't remember who created it)... 2SC9013 is almost certainly not a proper Japanese (EIAJ?) designation, but a co-opting of the numbering scheme?  I see a datasheet by an Indian manufacturer... it's a transistor for "potable radios".  I'll, uh, stick with the bottled water, thanks.   (Yes, any general purpose NPN will work, I've seen plenty of 2SC1815s in that position.)

Anyway, without analyzing it too closely, it should be okay.

As for faults, yeah, both transistors need to be replaced at the same time, and probably a lot more, as you've discovered.  Likely mechanism, fault current flowed through both transformers, delivering enough surge current to burn out transistors.  I would guess cause to be accidental reverse polarity on the output, or maybe an input surge.

Tim