UC3843B

[2X]

Hi,
I have a power supply from an old laser cutter and after some time stops to work. I noticed that when I have connected the cooler fan it takes more time to stop working.
The Current-Mode PWM Controller UC3843B when I open the power supply starts to heat. I leave open the power supply until reaches 64celcious and then I turn it off.
This temp for a chip like this isn't big?

I checked for short at the chip pins but nothing I found. In all pins in diode mode I measured around 0.55V except olny on pin 7 where I measure 0,2V. I replaced the chip with a new one but again is heated the same and I measured the same voltages in diode mode in all pins.

I measured also the frequency output (oscilloscope probe at pin 6) and is 100KHz... I have read on the internet that if the switching frequency is high it causes heating.

What could cause the heating of it - what could I check more?


Datasheet UC3843B - UCx84xA Current-Mode PWM Controller:
https://www.ti.com/lit/ds/symlink/uc3842.pdf

[Whales]

> This temp for a chip like this isn't big?

Yeah.

Does it suddenly and completely turn off?  The UC3843B doesn't really have features inside it to do that in response to overload or overheating.

Thoughts:

(1) Maybe the power supply is fine and instead the laser is tripping off?  Have you measured the power supply outputs during the fault occurring?

(2) The input voltage to the power supply could be sagging or be marginal.  The UC3843B has an UVLO that might be tripping, then resetting once things cool down.

(2) The PCB might have some temperature sensing circuitry that is forcing the chip to disable.  (Can you post more pictures of the board, both sides?)

[wraper]

Test ESR/replace electrolytic capacitors on the primary side, especially small one on UC3843B's VCC

I leave open the power supply until reaches 64celcious

It's an ambiguous claim, What exactly and how do you measure it?

[2X]

Test ESR/replace electrolytic capacitors on the primary side, especially small one on UC3843B's VCC

I leave open the power supply until reaches 64celcious

It's an ambiguous claim, What exactly and how do you measure it?

I measured the temperature with the UTi730E Thermal Camera.

[magic]

Is the chip the hottest part inside? What about the switching transistor or current sense resistor?

Is that output waveform taken when the PSU still worked, or when it didn't work properly?
Does the waveform change when it stops working? Probably, yes...

[Jakedevine]

Just a thought.

If the unit initially power up fine from cold and its subcoming to a overheating condition that causes the device to turn off and cooling fan reduces this, i would immediately be looking at solder joints. Had issues with PSU's that would work and then turn off after 3 or 4 hours.

Reflowing the tiny little SMD components fixed the issue in my case. Maybe worth looking at how dirty / old the solder looks. Given its a PWM I would expect it get hot.

They are notoriously crappy in my experience any circuit that they're in a psu they tend to fail. They're relatively inexpensive so replace it and see if it sorts your issue out. If not I would be looking at reflowing areas of the board that have got hot with power.

[2X]

I replace again the UC3843B chip and now the one from the two tranzistors where drive the SMPS transformer is getting hot... it reached 90celcius (I measured also with the heating probe with 121GW multimeter) and if I will not disconnect the PCB from the mains will keep rising (maybe the heat was from the tranzistor where is at the top side of the PCB under the UC3843B and the thermal camera was confused and actually the heating of the UC3843B was coming from the Q58).

This is going to crazy me. All day and today and I couldn't find something. I can't understand what causes the Q58 to reach such a temperature in less than a minute. I remove some chips and regulators from the output side of the SMPS trasformer to find if something will cause this issue, but the same again.. is getting hot).

[2X]

Is the chip the hottest part inside? What about the switching transistor or current sense resistor?

Is that output waveform taken when the PSU still worked, or when it didn't work properly?
Does the waveform change when it stops working? Probably, yes...

I am afraid now to leave it to turn off alone with such a temperature on Q59.

[Whales]

Thankyou for the photos and waveforms! 

The "choke" is probably a gate drive transformer.  This makes driving of the transistor gates very easy, regardless of how the transistors are wired to the rest of the circuit.  According to your drawings these two transistors are wired as a push/pull pair, which makes driving the top transistor's gate difficult (for the same reasons it was hard for you to probe).

This probably also means both transistors are the same model.  You attached an IRF540 datasheet, so I assume they are both that.

N.B. the whole circuit is called an SMPS, the bit you have labelled as "SMPS" is just the transformer in the SMPS circuit.


Toplogy

To understand your waveforms and why Q58 might be overheating I first need to understand what topology of power supply you have here.  They all look quite similar from the outside (transformers, capacitors, transistors) but vary a lot in how they are driven and can go wrong.  These slides by Brian King are a good start.

There is no inductor on the secondary side, so it's not a forward converter.

Based on your drawings: the two transistors are wired as a push/pull pair.  This means one is connected to GND and one to VCC, both share a middle wire/node.

This could be an active-clamp flyback.  That would explain why one transistor is doing all the work and getting hot, whilst the other is doing basically nothing (just eating a bit of oscillations).  Your waveforms show 50% duty cycle, which means this power supply would be trying to output a decent amount of power.  This could potentially mean that the power supply is perfectly fine and just overloaded.

The other option is an LLC converter.  I note the big 100nF film capacitor on the primary side of the power supply.  But I don't think I've ever seen someone use a UC3843 to do variable frequency control (I could be wrong).  Unless the UC3843 is a red herring and is actually for a different power supply on this board? 

EDIT: The gate drives are switching at tens of MHz!  If it's a flyback then it's very likely the timing parts near the UC3843 are bad, most offline flybacks run at 60Khz to 100KHz.  If it's an LLC then this frequency makes much more sense, but then I don't know why one fet is hotter than the other.


Question 1: Were these scope shots taken whilst the power supply had a load connected, or whilst it was unloaded?
Question 2: Does the power supply still overheat when no load is connected?
Question 3: Have you checked to see if it is your load that is the problem?  Your laser (diodes?) might be aging and going out of spec. It would be worth measuring the output voltage & current to see if it matches what your laser should be pulling (or the limits of your PSU).
EDIT Question 4: Are the voltage scales on your plots for the differential voltage probe correct?  Your plots show the node between the two transistors is only swinging by 18V up and down, but it should be something more like 170V or 340V if this is an offline power supply.  Perhaps a x10 or x20 probe?  Green trace in this graph:

[2X]

Question 1: Were these scope shots taken whilst the power supply had a load connected, or whilst it was unloaded?
These waveforms are form the power supply when is on the bench with no load connected.

Question 2: Does the power supply still overheat when no load is connected?
It is overheated when it is on the bench and no load is connected.

Question 4: Are the voltage scales on your plots for the differential voltage probe correct?  Your plots show the node between the two transistors is only swinging by 18V up and down, but it should be something more like 170V or 340V if this is an offline power supply.  Perhaps a x10 or x20 probe?  Green trace in this graph:
The probe is a TESTEC differential probe at x20 scale. The green trace is around 90Vp-p.

Also, one thing it seemed strange to me is that I didn't find an optocoupler chip between input and output.

If it is a push-pull configuration when Q58 mosfet is ON the Q61 must be OFF and when Q61 mosfet is ON the Q58 must be OFF. At the "SCR02.PNG" the two gate driving signals are in the same phase and don't have 180 degree phase different or I made a wrong measure?

[Whales]

Ahah, that's key information.  Overheating without a load connected narrows the possible causes.


If this is a flyback power supply then it is behaving very wrongly.  With no-load the duty cycle should not be anywhere near 50%, it should be more like 1% or even skipping cycles (0%).  The transistors also should not be switching this fast (60-150KHz is more common) and the excessive speed might be causing the transistor to overheat.   This suggests bad parts (perhaps even passives like resistors or capacitors) or bad solder joints are causing the controller chip (UC3843?) to misbehave.

If this is an LLC converter then that might explain why the freq is so high (LLC's raise their frequency dramatically under no load).  But then your UC3843 is a distraction?  Not sure, I would need to be able to read the numbers on your parts in your photos.


> Also, one thing it seemed strange to me is that I didn't find an optocoupler chip between input and output.

That is strange.  Perhaps the circuit is using primary-side sensing as an approximation?  Or maybe you're missing wires/parts that add sensing somewhere else nearer the laser diode?

Reverse engineering and drawing a schematic will tell you where it is expecting to get feedback from.


> If it is a push-pull configuration when Q58 mosfet is ON the Q61 must be OFF and when Q61 mosfet is ON the Q58 must be OFF. At the "SCR02.PNG" the two gate driving signals are in the same phase and don't have 180 degree phase different or I made a wrong measure?

If this were true then your transistors would instantly detonate with fire and explosion.  Turning both on at the same time would short VCC (hundreds of volts) straight to ground.

Your comment about the signal being below 0V makes me think you just had the differential probe pins the wrong way around. 

[2X]

If this is an LLC converter then that might explain why the freq is so high (LLC's raise their frequency dramatically under no load).  But then your UC3843 is a distraction?  Not sure, I would need to be able to read the numbers on your parts in your photos.

What do you mean with "But then your UC3843 is a distraction?"?
Also, some more better pictures where show the values.