UC3845 not turning on (EWM Welding machine)

[TheD312]

Hi there,

i have bought not working welding machine that has a broken auxiliary power supply with the known UC3845 IC. I have no idea what it had suffered from, because the previous owner did not tell me. So after opening it up I saw areas where the magic smoke had escaped. And as you could expect, it was the "power part" that went boom.

To be exact the main switching MOSFET and all resistors around it (R13, R14, R31,R32, R35... see "right side" in my schematic)- including the UC3845. So I replaced them and now there is a little bit of a problem.

First I just want to remind that I kinda drew a schematic of the cirquit without any values, it is more about understanding the cirquitry. Some single resistors on the schematic may be two paralled on the pcb. Two zener diodes are 4-5V, the rest are just normal melf diodes like 1N4148. I hope I`ve done it right...

So the problem is, that the IC does not want to turn on. The Vcc stayed pretty much constant at 8V (Lower than the IC UVLO of 8.4V). What I did is to measure every component in the schematic, all seem to be fine. I also tried replacing the TL431 reference- no success.

So then I removed the UC3845 and then the VCC voltage is now about 80V, voltage at caps C57 and C68 is 0V and after some random moment- the failiure is intermittant and really annoying- the voltage at the VCC pin goes down to 30V and C57 + C68 each have perfect 5V. But 30V is still too much.

This schematic has a very weird "startup" cirquitry and when it comes to diodes and transistors my head is a little bit confused. So I randomly started to desolder and resolder components to see what happens (BTW I connected the PCB to three seriesed power supplies, giving me a current limited, isolated 300V DC on the bulk caps) and when I desolder Resistor R150 (Gate- Pullup resistor of V53) there is perfect 12V on the VCC line. I havent soldered in the UC3845 tho, just incase the ciruit would fail because that one missing resistor.

Does anyone have an Idea what the solution could be? It is prette sure something that has together with these three transistors. I would apprechiate some tips for troubleshooting.

Oh and sorry for the mixed american and  german symbols

[boB]

Vcc of 8.4V sounds too close to the UVLO (under voltage lockout)  which may be slightly higher.

Some of these parts have UVLO's of 15 or so volts also.   You might want to check that.

[boB]

Also check to make sure the Vfb pin is not above 2.5 volts   and Current FB is not above 1V and pin 4 is producing a ramp voltage at your desired switching frequency.

I suppose the Vref is sitting at 5V as it should ?

[xavier60]

A mistake or 2 can cause a schematic to make little sense.
The TL431 will be there to sense a voltage, for protection likely. There should be a divider on its input.

[TheD312]

Thaks for the reviews.

When the IC was soldered in I scoped the RT/CT pins- they were fine. Also the 5V reference pin was perfectly here.

I then (like I wrote before) desoldered it to make investigate the VCC thing. And yes I have the chip with the lower turnon voltage threshhold.

I might made a small mistake, it was already late yesterday and I was sittign there tryign to figure it out.

Now i measured (without the UC3845) VFB which is 5,8V and COMP which was 5,7V although it was very slowly rising to about 5,85 an then i stopped measuring.

I fixed a mistake in the schematic with the TL431 reference input. There indeed is a resistor devider being formed by the 300k and R157, 158 and 29 (They are parallel).

BTW the startup resistor is the 68k, not the 300k.

I also attached a rough picture of the PCB, maybe that might help a bit.

I am not sure now if I should solder that UC3845 back in- I guess 70V at the VCC pin might be harmful for it...

Sorry also for the turned around scan. No idea how to do it with bloody windows 10.

[TheD312]

Forgot the rest lol...

[TheD312]

And here is the "more important" cirquit on the PCB

[xavier60]

Most of the circuitry on the left is capable of forcing the VFB pin high, disabling drive to the MOSFET.
It's difficult to give details, I can't read the reference numbers.
It is risky to power it with components removed. If you do something that breaks the regulation loop or disables the current limiting, the MOSFET will fail, destroying other parts also with loud bang.
The MOSFET can be removed as a precaution for now.

[TheD312]

I know I should not remove anything but just for fun I did it.

Now I made some more meassurements.

With everythign soldered back in and conected to 300V: Vcc stays at around 7,9V, Vref gives me about 20ms long pulses of 4V every 40ms.  Removing V52 didn`t do any difference, so I guess the cirquit on the very right with V48 and V50 is there only to do something when the CU3845 is enabled. So the problem has to be around V53. But what is it???

I now resoldered, measured and soldered back in every single component and all of them semm to be fine! THis thing is crazy. The PCB is from around 2001, maybe the UCs were a bit different back then but I am very sure that I bought the exact same one.

I also connected 15V to the Vcc via one of my lab power supplies. Current draw is about 10mA and a perfect 5V at the Vref pin and also the  sawtooth at the RT/CT pin. The Mosfet is not enabled becaus ethe 300V was missing I guess, but at least the IC does as it seems like.

At this point I regret not completely understanding this cirquit...

[xavier60]

Keep testing with the 15V for now. If the VFB pin is externally forced above 2.5V, causing the comp pin to be low, drive duty will be zero.
The circuity on the far left doesnt make sense. Check for mistakes.
I need to be able to clearly read the references to be able to help.

[TheD312]

I dont even know what to try and where to poke around, I have tried everything that my knowledge allows. BTW at 300V and the VCC being 7,9V the Vref pin gives pulses of 5V and the COMP pin pulses of 3V.

[xavier60]

I dont even know what to try and where to poke around, I have tried everything that my knowledge allows. BTW at 300V and the VCC being 7,9V the Vref pin gives pulses of 5V and the COMP pin pulses of 3V.

The IC should be outputting Gate drive pulses while the COMP pin is at 3V.

[TheD312]

But it does not do exactly that. And the output is not shorted to ground, there is only 10k to ground (Mosfet pulldown).

[xavier60]

Look for "Figure 7: Current Sense Input Threshold vs. Error Amp Output Voltage." in the data PDF,  https://www.st.com/en/power-management/uc3845b.html
With 3V on the Comp pin, the output should go high every oscillator cycle and stay high until the Isense pin goes up to 0.6V.
It will be safer and more forgiving to mistakes with only the IC powered by a 15V supply.

Edit: I should have said "Comp" pin.

[TheD312]

Seems like I was a silly goose 

I noticed, that my current sense resistors are two 20 Ohm in parallel, same as the two resistors between the transformer and the positive voltage rail. Seems a little bit high, huh?

99% sure it was them. I am going to order some and I think this cirquit should work fine again.

[xavier60]

Seems like I was a silly goose 

I noticed, that my current sense resistors are two 20 Ohm in parallel, same as the two resistors between the transformer and the positive voltage rail. Seems a little bit high, huh?

99% sure it was them. I am going to order some and I think this cirquit should work fine again.

That would cause the Gate drive pulse width to be very short.