Replacing UC3835A PWM controller with UC3853A

[CG2]

As per the title, I've got an ATX PSU that's only giving 5v standby. I've checked the mosfets and diodes, and they all seem OK, so my next suspects are the output monitor chip (DM311) and the PWM controller (UC3845A). I'm too much of a coward to try to test these live (the PWM controller is buried between the two heat sinks), so now I just start swapping things out. I don't have any of either to hand, but what I do have is some UC3843AN, which are the same family as the latter, but a slight variant.

According to the datasheet, the first digit relates to operating temp. They're both 3, which is 0°C to 70°C, rather than 1 or 2 which are for more extreme environments.

The last digit describes the operating voltage, 3 and 5 are "TURNON AT 8.4 V, TURNOFF AT 7.6 V, SUITABLE FOR DC-DC APPLICATIONS" (as opposed to 2 or 4 which work at 16 and 10 volts), so again no problem there.

The last letter(s) just relate to design improvements over the years. They all seem to be compatible, but the later ones draw a little less current.

So the difference between the 3 and the 5 versions is related to the duty cycle. The 3843 can run from 0% to 100% duty cycle. The 3485, however has an internal flip-flop that pulls every second pulse low, so it can only drive the mosfet up to 50% (and presumably at half the frequency of the oscillator). The switching mosfets are a couple of 20N60C3, and the output diodes S30D45C and S30C100C which seem to be bog standard components in PSUs.

I can't see why it wouldn't work, but I'm just curious if anyone here has any thoughts on it.

[stj]

Re: Replacing UC3835A PWM controller with UC3853A

DM311 sounds familiar - i think thats your 5v standby chip

[momaka]

Re: Replacing UC3835A PWM controller with UC3853A

DM311 is not the monitoring chip. It is the 5VSB PWM-FET (offline switch.) If you have 5VSB, that one is definitely not your problem.

I also -don't- suggest to replace the UC3845 PWM just yet. Typically, the PWM controller won't fail unless the main switching MOSFET went bad and sent a high-voltage spike through it.

Perhaps it's better to describe what PSU you're troubleshooting, what you've checked / found out so far. Pictures would be helpful too.

[eccerr0r]

Re: Replacing UC3835A PWM controller with UC3853A

UC3842 - high lockout up to 100%, nfet driver
UC3843 - low lockout up to 100%, nfet driver
UC3844 - high lockout up to 50%, nfet driver
UC3845 - low lockout up to 50%, nfet driver
LM3485 for DC-DC converters using pfet

Though the UC384x are all very similar they are usually chosen to ensure you don't get into a problem when you have to deal with really bad situations that could cause things to burn... these are very resilient chips in general too. I'd also not replace a 50% chip with a 100% chip for testing, but may try the reverse for experiments only...

[CG2]

Re: Replacing UC3835A PWM controller with UC3853A

For the back story, 5 or 6 years ago I bought a job lot of 3 or 4 ATX PSUs at a salvage auction. Think Amazon returns, but they could have come from half a dozen or so different retailers. I needed one at the time, at least one worked, and a couple were dead out of the box, but otherwise brand new so they went into a cupboard and got shuffled around a bit over the years and eventually were forgotten. I stumbled across one a few months back, tested it a bit, tracked the problem down to the UC3843, bought some from Aliexpress, and when they eventually arrived fixed it.

I've just stumbled across this second one, and it's got the same symptoms. Caps are OK (it's never been used), no obvious burnt components, I've had both heatsinks off and tested that everything's OK (Large Double Diodes S30D45C, Small Double Diodes S30C100C, Mosfet 20N60C3). It now comes down to the PWM controller UC3845A, or the Power Supply Supervisor. As you say they're resilient chips, but stuff that arrives dead tends to have different failure modes from stuff that later dies.

I bought a tube of ten UC3843s to fix the previous PSU. That was a 850W model, but this one is a 650W (Both are branded Xanta, but that looks like some random word stuck on by a buyer who got a couple of pallets from the factory; they're cheap generic things). I thought that rather than wait a couple of months for the right chips to arrive from China, I could stick one of these in for a test. I'm just curious as to what difference the change in duty cycle could make. After all, it's just switching a mosfet on and off. That mosfet might get a bit hotter from being switched on and off twice as often but the 20N60C3 seems to be a pretty standard PSU component, and there's a temperature sensor on the heat sink that should stop it overheating. I suppose there's a danger of oversaturating the transformer and burning that out if I run it under a heavy load, but the diodes and caps should cope fine, and the supervisor chip won't have a clue.

Anyway, making this post, has given me a lead. Of course the DM311 is the standby chip. The Power Supply Supervisor is a WT751002S. And when I repaired the other one, I had the problem down as being either the UC3843 or the WT751002S, so I bought some of each. I changed the UC3843 first and that solved the problem, so the WT751002S were "carefully filed away". I'll just have to remember where.

[stj]

Re: Replacing UC3835A PWM controller with UC3853A

if you use ali saver shipping, stuff from china takes 2-3weeks not 2 months.
just saying - the shipping options are there for a reason

[CapLeaker]

Re: Replacing UC3835A PWM controller with UC3853A

I’ve also seen them fail dead and in one instance it quit working after 30 seconds being on. Replaced them and all ok.