Older Macintosh Magnetek power supply ressurection

[mitchw]

Hi, I have an old PowerMac 8100 power supply from a company MagneTek.

I need some advice repairing it. I am not sure if I got the right SMPS controller IC.

When I got it, the primary side was completely blown, both MOSFETs blown, control IC blown its cap off, etc.

I slowly got it to work, first I disconnected all the visually blown and shorted components, until all there was left was a standby power supply. There was a major darkening in the standby area, but it was caused by an overheating resistor (which still tests good - 68 ohms). This was caused by dried up capacitors, at first I decided not to recap it, but when I tried it first, the power supply drew 20W thorugh the lamp limiter. Recapping made it much happier (but still a bit high) at 9W. The resistor gets only warm now.

Next, I tackled the primary switcher. I ordered two new MOSFETs, new diodes, few resistors, two new optocouplers, etc. But there was a problem. The main SMPS control chip was completely blown, so I have not been able to identify it. I thought it would be most likely an UC3843 (Vcc, GND, Output pins were all in right places), so I put this one in.
Also I was not able to get the same MOSFETs, the original ones were STP5NA80FI which aren't made anymore, so I put in STP8N80K5 ones, main difference is that the Rds is lower (should only make it more efficient).

Now, the power supply kinda works, but something is still very wrong.
1. It starts automatically when you apply power (it has soft power, so it should wait for PS_ON signal)
2. The voltages on the secondary are too high, I get 6.4V on 5V rail, and 12.5V on 12V rail.

I checked the schematics of some of the UC3843 power supplies and none of them uses two optocouplers as this one does. So I was thinking, perhaps the main issue is that I just happened to get a wrong SMPS IC that just coincidentally kinda works.

I attached some photos of the circuit.
Thanks

[R_J]

Re: Older Macintosh Magnetek power supply ressurection

Did you look inside the case to see if the ic's top was inside somewhere?

[momaka]

Re: Older Macintosh Magnetek power supply ressurection

I checked the schematics of some of the UC3843 power supplies and none of them uses two optocouplers as this one does.

That's not quite true.
It really depends on the power supply.

Generally, yes, only one optocoupler is needed with UC384x ICs to keep regulation. But if the PSU has "soft off" / standby power supply inside it and the "main" PSU section is off and turned on remotely... then yes, you will see two optocouplers - and of these two, one will be for feedback into the UC384x IC, which the other will be there simply to turn power on and off to the UC384x IC, as demanded by the device when in "soft off" mode.

And in fact, on computer PSUs that use UC384x chips, you will typically see >three< optocouplers: one for main PS feedback, one for main PS power to the PWM IC, and one for 5VSB/standby power rail regulation.

[mitchw]

Re: Older Macintosh Magnetek power supply ressurection

Sadly there was no trace of the blown IC's top inside the case.

I checked the UC384x datasheets and found out that the UC3842 would be more appropriate one as I get 12V on the IC's Vcc, but UC3843 has on voltage of 8.5V. UC3842 has a turn on voltage at 16V (max).

Standby power in this supply is being produced unregulated at around 15V (I must check again for exact voltage) and is then being regulated by an 7805 to produce 5V.

In the schematics, soft power optocoupler was usually connected to a base of a transistor that opened when triggered and powered on the UC384x IC.

[momaka]

Re: Older Macintosh Magnetek power supply ressurection

Standby power in this supply is being produced unregulated at around 15V (I must check again for exact voltage) and is then being regulated by an 7805 to produce 5V.

Sounds about right, given the pictures. The 5VSB uses a single-transistor self-oscillating circuit that produces anywhere from 10-20V that then gets regulated by a 7805 linear regulator. This is really old and inefficient technology... and dangerous when the circuit decides to go haywire due to bad caps (usually burned parts and magic smoke, as you saw.)

In the schematics, soft power optocoupler was usually connected to a base of a transistor that opened when triggered and powered on the UC384x IC.

Yup, that sounds right too.

I checked the UC384x datasheets and found out that the UC3842 would be more appropriate one as I get 12V on the IC's Vcc, but UC3843 has on voltage of 8.5V. UC3842 has a turn on voltage at 16V (max).

For testing purposes, you could just inject a voltage with a ungrouded power supply straight to your UC3842/3843 IC, just to see if the PSU will turn on. But note that you should use a series incandescent light bulb if you do this, since there will be nothing to disable the UC3842/3 chip from trying over and over due to a fault (as you are providing power to it and this means most protections may not be working.)

[mitchw]

Re: Older Macintosh Magnetek power supply ressurection

Hi, thanks for the reply.

For testing purposes, you could just inject a voltage with a ungrouded power supply straight to your UC3842/3843 IC, just to see if the PSU will turn on. But note that you should use a series incandescent light bulb if you do this, since there will be nothing to disable the UC3842/3 chip from trying over and over due to a fault (as you are providing power to it and this means most protections may not be working.)

The PSU actually turns on and it kinda works, but as I said, the output rails are not really regulated, there is a lot of squealing from the switching transformer and the PSU turns on when there is a power applied (it should wait for PS_ON signal). Also with light load, the power consumption goes way high, for example adding a 10W 12V bulb will increase the wattage by 25W (checked with wattmeter), and squealing will change a bit.

I ordered a pack of UC3842 chips and we will see if it will be any better.

I used 150W bulb in series to protect if perhaps something shorts

[stj]

Re: Older Macintosh Magnetek power supply ressurection

it probably needs a re-cap
specially in the area that got so hot the board turned black!!

[mitchw]

Re: Older Macintosh Magnetek power supply ressurection

it probably needs a re-cap
specially in the area that got so hot the board turned black!!

It was already recapped as I have written in the first post
All of the caps were dried up.

[stj]

Re: Older Macintosh Magnetek power supply ressurection

try resoldering everything in an area where the board went black

[momaka]

Re: Older Macintosh Magnetek power supply ressurection

The PSU actually turns on and it kinda works, but as I said, the output rails are not really regulated, there is a lot of squealing from the switching transformer and the PSU turns on when there is a power applied (it should wait for PS_ON signal).

In that case, I suggest replacing any 431 shunts you see on the secondary side and also maybe the supervisor chip. Chances are, with the 5VSB going bad originally the way it did, it could have gone over-voltage and killed the supervisor chip (in fact, this was quite common on cheapo ATX PSUs with this type of single-transistor self-oscillating 5VSB/standby circuits.)

[R_J]

Re: Older Macintosh Magnetek power supply ressurection

I was looking at the circuit and some pwm ic's and your circuit has pin2 connected to ground, I can't see a 3842/3 working as that is the FB line
BUT, a FAN7554 looks like it might be it. It uses pin1 for FB and pin2 is the softstart, wich maybe it does'nt use

[mitchw]

Re: Older Macintosh Magnetek power supply ressurection

I was looking at the circuit and some pwm ic's and your circuit has pin2 connected to ground, I can't see a 3842/3 working as that is the FB line
BUT, a FAN7554 looks like it might be it. It uses pin1 for FB and pin2 is the softstart, wich maybe it does'nt use

Hi,
I got a UC3842 and now the PSU won't start at all.

I was looking at the FAN7554, but I think it is too new to be correct.
This PSU is from 1994 after all...

[R_J]

Re: Older Macintosh Magnetek power supply ressurection

Thats likely due to the under voltage lockout of the UC3842.

[stj]

Re: Older Macintosh Magnetek power supply ressurection

i remember a device with the 3842 that had a common problem, it had a pair of high voltage resistors supplying the chip that went up in value.

it was a schneider sattelite reciever - before anybody asks.

[momaka]

Re: Older Macintosh Magnetek power supply ressurection

Hi,
I got a UC3842 and now the PSU won't start at all.

As per my previous post, perhaps try replacing the supervisor chip on the secondary side too. When that 5VSB circuit went kaboom, it could have taken it out. Also, double-check all power circuitry supplying the UC3842 chip. The 5VSB circuit has an auxiliary primary winding that supplies power to the UC3842 IC. Typically, this winding is connected to a rectifier, a cap, and then a transistor + some control circuitry on the optocoupler that enables the main PS (once PS-ON signal is asserted.) Check the rectified DC voltage on this primary auxiliary winding from the 5VSB with the PSU plugged in (but do NOT pull PS-ON to ground - i.e. keep the main PS off.) Post what you find, please.