Difficulty repairing UC3842 power supply...

[dave_k]

Was recently asked to look at a small switchmode power supply to repair.

Initial diagnosis was blown mains fuse (2A), shorted FET and failed TL3842 controller. Dead semiconductors were replaced, along with 2 small electros on the primary side.

Applied power (via isolation transformer and 100W series bulb). 100W bulb glowed briefly confirming main filter cap charged up, however no output voltage was measured.

After some probing and measuring it appears the 3842 is not starting up, pin 7 not reaching the 17.5v needed to begin operation (measured a peak of ~14v).

Feeding 18v from an external supply into pin 7 causes the 3842 starts up and the supply runs. During operation, pin 7 has between 10.5-11.5 volts present (coming from the AUX winding supply).

The start-up supply for the 3842 is a little unusual in this design. Instead of a resistor from the main DC rail, a small capacitor is connected to one leg of the incoming mains (used as a voltage dropper) then rectified by a pair of diodes. The value of the cap is 0.22uF which has a reactance of about 15k ohms at 50Hz.

All other components on the primary side have been individually checked and tested OK.

My question is what would be stopping this supply from starting up normally? 


Reverse-engineered schematic:
(secondary side has been omitted, it's a bog standard TL431 driving the opto).

[Armadillo]

Have you replaced the 0.22uf?

[dave_k]

Have you replaced the 0.22uf?

Yes ..

[Armadillo]

Would you be able to probe the aux supply?

[dave_k]

At one stage D1 and D2 were removed. Starting the 3842 using a pulse of 18v external supply, pin 7 measured 10.5v after external supply was disconnected.

It appears the AUX winding is able to power the 3842 without difficulty (once it starts)..

[Armadillo]

Maybe the secondary needs to be loaded?

[dave_k]

I thought this as well. Loading the output with 1A doesn't change the result .. still refused to start on it's own.

Once started, it can happily run and supply up to 3A. Voltage regulation is quite good, output dipping by only a fraction of a volt at this load.

Fundamentally the supply works, it just doesn't start. This is where I am stuck, there is something I am overlooking. 

[kripton2035]

I would replace C2

[dave_k]

Dead semiconductors were replaced, along with 2 small electros on the primary side.

Already did that .. C2 and C4 both replaced.

[carl_lab]

Are you sure your wiring of the optocoupler is correct (emitter is open)?

OK, you have corrected now...

[Armadillo]

I am sure you have dig through the power supply and replaced all that can be replaced, except the aux windings.

The aux could be a bit nicked.

Maybe the short cut, would be to replaced the 0.22uF to something a bit higher, say try 0.47uf.... 0.68uf.... if you have the appropriate capacitors.

Just a suggestion.

[CJay]

I reckon D1 should be a Zener and C2 should be a higher value, they're usually around 22uF-100uF, has someone else been in there first?

[xavier60]

I notice there is no Gate to Source bleed resistor shown. With some versions of the UC382, the output pin stays in a Hi Z state while the IC is in UVLO state. If there is no Gate bleed resistor, the Gate eventually charges up, blowing the MOSFET again.
http://www.st.com/content/ccc/resource/technical/document/application_note/3b/42/e6/d9/78/47/4d/df/CD00003930.pdf/files/CD00003930.pdf/jcr:content/translations/en.CD00003930.pdf

[xavier60]

Leaky diodes often don't test bad with a DMM. I notice that when small glass diodes have corroded leads, they tend to be leaky.
The value of C2 seems unusually small.

[T3sl4co1l]

Dang, full schematic capture?  Very nice!

The 3842 is supposed to be off (<1mA supply consumption) until it kicks on.  If it can be powered by the startup circuit, it should more than start up normally.  That's your mystery to solve: what is sucking down all the current, or keeping the startup from delivering enough current?

I wonder if the 3842 has been damaged in such a way that it's very leaky, but still otherwise functional?

Tim

[Kevin.D]

Are you sure your schematic of the startup circuit is correct , ? The way it is shown it  only supplies a single startup current pulse on the first mains  cycle as C3 is initially charged  (so c3/c2 form a capacative voltage divider with C2 charging up-to ~ Vmains(peak)* C3/C3+C2).
Once C3 has charged it has no means to discharge hence no further pulses are available on following mains cycles.
If this is what is actually intended then the ratio of c3/c2 determines Vstart (and after a power down you would have to wait for C3 to discharge (shown here only possible via D2 leakage) before another startup pulse was possible).
The only other thing to check is if it's being dragged down like others suggested.

regards

[T3sl4co1l]

Oh, yeah, that one diode is on the wrong side.  I hope? 

Tim

[dave_k]

Dang, full schematic capture?  Very nice!

Altium is good for something, right? 

The 3842 is supposed to be off (<1mA supply consumption) until it kicks on.  If it can be powered by the startup circuit, it should more than start up normally.  That's your mystery to solve: what is sucking down all the current, or keeping the startup from delivering enough current?

Yes, and this is where I am getting stuck.

The schematic I posted probably doesn't tell the full story, the part numbers are of the original unit.
The SSP6N60A MOSFET (made by Samsung) was replaced with a IRFBC40PBF (Vishay). The original PWM IC was a Texas Instruments TL3842P, it has been replaced by a Texas Instruments UC3842N

One avenue of thought was the replacement MOSFET was unsuitable (thinking Q_g was too high) and the 3842 was having difficulty initially turning it on. Comparing data sheets shows original MOSFET Q_g is 77nC max, replacement is 60nC.

I wonder if the 3842 has been damaged in such a way that it's very leaky, but still otherwise functional?

Original 3842 was replaced, plus I have tried 2 other 3842's from junked equipment - same result with each chip.

[dave_k]

Oh, yeah, that one diode is on the wrong side.  I hope? 

Tim

Oops, yes - D1 is indeed connected to the anode of D2 (drawing error).
Original post has been modified with the correct schematic.

[dave_k]

UPDATE: problem found and fixed. Anybody want to have a guess at the solution?

[carl_lab]

Anybody want to have a guess at the solution?

Dead electrolytic cap?
Open soldering joint?
Some loose connection? 

[oldway]

What's the voltage on C2 ?

[xavier60]

Anything to do with the isolation transformer?

[mzzj]

Leaky diode

[dave_k]

The value of C2 seems unusually small.

C2 should be a higher value, they're usually around 22uF-100uF

Turns xavier and CJay were on the right path. After reading through a few data sheets and application circuits for the 3842 it was became obvious that C2 was too small. Increasing the value to 100uF allows the supply to reliably start.

Now another problem needs attention. With the supply running (even with minimal load) the MOSFET heatsink becomes quite warm after a few minutes.

Thoughts are either: 1) low gate drive and MOSFET not turning fully on, or 2) problem with snubber circuit.

Will try and capture some scope traces tomorrow..

[dave_k]

What's the voltage on C2 ?

voltage is about 11.5V after increasing value of C2

[mzzj]

What's the voltage on C2 ?

voltage is about 11.5V after increasing value of C2

I'm guessing you are talking about C3 value, not C2?

[xavier60]

With C2 being too small, I would expect that the UC3842 would have been repeatedly making start attempts. There should have been some output voltage.
The 22 ohm gate drive resistor might have gone high causing sluggish switching of the MOSFET and higher dissipation.

[dave_k]

Quite strange that there was originally only a 1uF cap there to begin with!
The 22 ohm resistor is okay, and substituting it for a 15 ohm resistor doesn't improve anything.

[xavier60]

There is soft start circuitry on the Comp pin,1. So it is possible with  C2 being 1uf, the UC3842 might not have had a chance to produce any MOSFET drive before pin 7 drops to the UVLO voltage. Pin 7 might have been cycling between 16v and 10v and the multimeter reading some between average.

[xavier60]

Take note of the operating frequency. It doesn't happen very often, I have seen the oscillator timing capacitor go low in value causing the switching frequency to go too high resulting in poor efficiency.

[xavier60]

A shorted snubber might also explain all of the trouble.

[dave_k]

Given the values shown in the schematic the operating freq works out to be about 115kHz. Again, typical application circuits show a normal value for Ct of 4700pF .. hmm.

I did check the snubber diode and resistor, they are okay.

[CJay]

The value of C2 seems unusually small.

C2 should be a higher value, they're usually around 22uF-100uF

Turns xavier and CJay were on the right path. After reading through a few data sheets and application circuits for the 3842 it was became obvious that C2 was too small. Increasing the value to 100uF allows the supply to reliably start.

Now another problem needs attention. With the supply running (even with minimal load) the MOSFET heatsink becomes quite warm after a few minutes.

Thoughts are either: 1) low gate drive and MOSFET not turning fully on, or 2) problem with snubber circuit.

Will try and capture some scope traces tomorrow..

You may find that whatever is causing the MOSFET to heat up is also contributing to your starting problem because the aux winding on the transformer will also be affected by the same problem and of course that will affect startup/run.

But, the typical application of that chip sees C2 being powered via a high value series pair of resistors to a zener and capacitor, the capacitor is usually 22 or 47uF* and is almost always faulty.

(which is where my 'rule of thumb' law comes from, if it's got a 4 and a 7 in the value, it's faulty, yes, I've repaired a *LOT* of SMPSUs, hundreds a month for two or more years of my life)

[dave_k]

Couple of scope captures with no load on the PSU...

[dave_k]

.. and with load on the PSU

[tautech]

This is what worries me:



One can only assume the gate drive is sufficient to drive the gate to ON then apparently falls below the gate threshold. 

That needs further investigation.

[xavier60]

The gate and source are noisy likely because non ideal probe grounding which is difficult to do with SMPS circuits.
The MOSFET turn off seems sluggish. Does the replacement MOSFET have a lower gate threshold voltage than the original?

Extra:
Take note of the plateau duration on the falling edge of the gate waveform.

[xavier60]

On second thoughts it looks like there is a high capacitive load on the drain circuit. The drain voltage is still rising after the gate has gone down to zero.

[T3sl4co1l]

You have to take gate waveforms with a grain of salt.  It's not mentioned if that's at the 3842, or at the gate -- the gate resistor (if present), and trace length, makes a difference.  Probing technique is important.  Some transistors simply bounce, especially when ferrite beads are involved.

I measured a similar waveform on my LED lamp, which is the same basic circuit, when I recently fixed its EMI problem finally (it was buzzing in the shortwave band; I threw ferrite beads at the switching transistor and rectifier, and that did well enough for the improvement I was looking for).

Tim

[dave_k]

The waveform on pin 6 of the 3842 looks nice and square, so I'm happy the IC output is okay. Similarly the voltage on pin 7 is stable and free of ripple, so I'm fairly sure it's not being excessively loaded down either.

The datasheet for the original MOSFET states V_GS(th) is 2.0(min) to 4.0(max) - V_DS=V_GS, I_DS=250uA. Replacement MOSFET values are exactly the same.

Similarly for input capacitance, original is 1800pF (f=1.0MHz) replacement is 1300pF (f=1.0MHz).

I'm beginning to form the opinion that perhaps the AUX winding is damaged somehow, and can barely supply enough voltage to run the 3842 properly .. pin 7 is 11.4 volts, and factoring a 0.6v drop for the internal switching transistor doesn't leave much headroom to drive the MOSFET.

Datasheet for replacement MOSFET is attached...

[xavier60]

I'm starting to think that nothing it wrong. The 2 MOSFETs have comparable specs. The replacement has slightly higher Total Gate Charge which will increase switching losses by a bit.
The UC3842 does not do any Green tricks like burst mode, it drives the MOSFET at the same frequency regardless of load. So the switching losses stay high all of the time regardless of load. 

Check the waveform on the snubber capacitor anyway, it should have only DC on it.

[dave_k]

Oh, I thought the replacement had a slightly lower total gate charge than original?

Original = 77nC (max) V_GS=10V, I_D=6A, V_DS=480V
Replacement = 60nC (max) V_GS=10V, I_D=6A, V_DS=360V

[xavier60]

11.4v is low considering that the UC3842 stops working at 10v. The UC3843 works down to about 8v.
Maybe the AUX winding is corroded. A resistance reading might show the problem.

[xavier60]

Oh, I thought the replacement had a slightly lower total gate charge than original?

Original = 77nC (max) V_GS=10V, I_D=6A, V_DS=480V
Replacement = 60nC (max) V_GS=10V, I_D=6A, V_DS=360V

Yes,you are right, I was looking at the wrong part.

[tautech]

11.4v is low considering that the UC3842 stops working at 10v. The UC3843 works down to about 8v.
Maybe the AUX winding is corroded. A resistance reading might show the problem.

Or D5 is sick.

[IanMacdonald]

Usual reason would be R11 (820k) gone high. You did lift one leg to test this? It can't be measured in circuit.