Another SMPS making a fool out of me

[PA0PBZ]

Trying to fix another SMPS for fun and learning. This is a 12V/2A one with a TL431. Replaced the output caps cause they were dead as a dodo, but now the output looks like a sawtooth. The oscillator starts, the output voltage ramps up and then the oscillator stops and the output 'slowly' drops until the oscillator starts again. All this with a period of about 20ms if I remember correctly. If I load the output the bottom of the sawtooth just gets lower.
So I had a look at how the voltage is controlled and it is a basic circuit with a TL431 feeding the opto. Not wanting to electrocute myself  I hooked up the bench PSU to the output of the SMPS and sure, whenever the voltage is over 12 volts the cathode of the TL431 goes low. Then I noticed something I can't explain. The cathode never goes below 2 volts, while the anode is at ground. There must be some kind of feedback from the cathode to the ref in, but I fail to measure or simulate it. When I take the TL out of the circuit and hook it up with a 1K resistor from the ref and cathode to the + and the anode grounded the voltage at the cathode rises to 2.5 volts and stays there when I rise the input voltage, so I thing the  TL is fine. So what's going on here?

[4cx10000]

Sounds like the feedback does not work properly. What kind of controller do you have on the primary side?

[KhronX]

That schematic shows that the TL431 is entirely not compensated.

http://www.onsemi.com/pub_link/Collateral/TND381-D.PDF
http://switchingpowermagazine.com/downloads/15%20Designing%20with%20the%20TL431.pdf
http://powerelectronics.com/site-files/powerelectronics.com/files/archive/powerelectronics.com/mag/Kollman%20and%20Betten%20September%202003.pdf

[xavier60]

What does the sawtooth look like? Does it momentarily flatten at the top at the the regulation voltage?

[T3sl4co1l]

That will make a sawtooth, yes, by design apparently.

Try an R+C from cathode to REF.  Start with R similar to the divider resistor values, and C such that the R*C time constant is similar to the sawtooth rate. 

BTW, TL431 isn't an "adjustable zener", its AC properties are extremely different from such a thing!

Indeed, it's a much more natural fit to call it an op-amp, with an open collector output (it makes an output voltage, but it can only pull down), an inverting input with a stupidly large offset voltage (yet that happens to be very consistent, i.e. 2.50V give or take ), and noninverting input strapped to "anode".

It may also serve to drop the recommended naming convention... there's nothing important about "anode" or "cathode".  Rather, they're better seen as "ground" and "output" respectively.  Or "emitter" and "collector" (with "ref" == "base", and Vbe being a very accurate 2.50V) -- a very high hFE, very slow (fT ~ low MHz) transistor.

Tim

[PA0PBZ]

Thanks all for playing along, it did make me realize that there should be something like compensation and indeed, after cleaning my glasses I found it



Of course that still doesn't make the sawtooth problem go away... I attached a scope picture to make it clearer:

Yellow - 2V/div, output, sawtooth 60ms 12.5 - 12V. If I load the output the period doesn't change but the lower part of the sawtooth drops to almost 4V
Green - 5V/div, D1 input from the transformer
Blue   - 2V/div, cathode of the TL, still wondering why it doesn't drop to ground

[SeanB]

Look at the primary side, your startup capacitor hes high ESR, which is causing the hiccups. the fact that it is occuring at full wave rectified line frequency is a very good indication, and if you put the scope on line triggering you will see it keeps lock with the mains frequency.

[PA0PBZ]

Look at the primary side, your startup capacitor hes high ESR, which is causing the hiccups. the fact that it is occuring at full wave rectified line frequency is a very good indication, and if you put the scope on line triggering you will see it keeps lock with the mains frequency.

Hi Sean,

Unfortunately that's not it, I already replaced that cap (learned my lesson) and full wave rectified line frequency would be 10 ms, not 60.

[Andy Watson]

Another vote for knowing what the controller is on the primary side. Some controllers will not go into full-power mode if they cannot sense some feedback - i.e. your opto may be broken or its transfer ration has deteriorated too much. Do you have facility to safely monitor the waveform on the primary side of the opto?

[T3sl4co1l]

The little cap AND the big cap, on the primary side?

The TL431 is commanding full power (blue curve swings up and saturates!!!), but nothing is happening.  The controller is off, and rebooting.  Looks like a short circuit load condition, where it doesn't have enough power to keep itself going.  But the output is unloaded, as we know!

Maybe the aux supply winding, or diode, or something around there, is open.

The TL431 won't saturate below about 2V because it has an internal "Baker clamp" diode (i.e., from REF to K).  Saturation, in fact, regulates at 2.50V - Vbe.   (There's also a reverse protection / substrate / ESD diode from A to REF, and actually from A to K, which is the only sense in which "A" and "K" are meaningful for this device.  But it's not a useful sense. )

Tim

[xavier60]

Also, have you replaced the output capacitors with original values?

[PA0PBZ]

The little cap AND the big cap, on the primary side?

The big cap is ok, it gives a nice BANG a minute after disconnected from the mains. (yeah, I know, I shouldn't... it's just a practice toy, probably binned after repair)

The TL431 is commanding full power (blue curve swings up and saturates!!!), but nothing is happening.  The controller is off, and rebooting.  Looks like a short circuit load condition, where it doesn't have enough power to keep itself going.  But the output is unloaded, as we know!

Nothing is happening until after it has been saturated for 5 ms, I think it should start somewhere at the up ramp? Having a working one would simplify things, but yeah...

Maybe the aux supply winding, or diode, or something around there, is open.

Once the primary starts to work the output ramps up fast enough, maybe it's just lazy?
I will have a look at the aux supply circuit, it got me before, somehow I was assuming that the failure was (only) due to the output caps gone to heaven, but who knows.

The TL431 won't saturate below about 2V because it has an internal "Baker clamp" diode (i.e., from REF to K).  Saturation, in fact, regulates at 2.50V - Vbe.   (There's also a reverse protection / substrate / ESD diode from A to REF, and actually from A to K, which is the only sense in which "A" and "K" are meaningful for this device.  But it's not a useful sense. )

Ah, ok, that explains at least that point.

[PA0PBZ]

Also, have you replaced the output capacitors with original values?

Yes

[tautech]

Once the primary starts to work the output ramps up fast enough, maybe it's just lazy?
I will have a look at the aux supply circuit, it got me before, somehow I was assuming that the failure was (only) due to the output caps gone to heaven, but who knows.

Me too.
Replace the diode. 

[T3sl4co1l]

The big cap is ok, it gives a nice BANG a minute after disconnected from the mains. (yeah, I know, I shouldn't... it's just a practice toy, probably binned after repair)

Well, you're lucky that didn't stuff the controller... a 100A+ surge over 10us will do wonderful things to a circuit!

Tim

[xavier60]

The output rail voltage is overshooting causing the TL431 to stay on for too long causing a restart delay.
When the rail first rises from zero, current through R1, R2, C2 and R5 should cause the TL431 to start turning on just before regulation voltage is reached. Maybe these parts have been damaged. Is R5 really 15K?

Edit- A close look at the waveforms indicates that C2 and R5 are intact. I still have doubt about the value of R5. Maybe the rails needs more load.

[PA0PBZ]

Alright, let's have a look at the aux supply, time for channel 4 and the differential probe hooked up on the aux cap:



According to the datasheet of the controller (3843B) 8 volts is marginal and well within the region where the controller can shut down. Since I already replaced the cap twice it's time for the diode, but unfortunately that changed nothing. All that's left is the ic or the transformer and well... the aux supply winding is open! 
Took the transformer off the pcb and had a close look:



Yes, that is not going to work, but it looks fixable:



That's better. Reassembled the thing and...  It lives again! That was another interesting experience, again thanks to all that played along.

[4cx10000]

Well done PA0PBZ! 

Fault finding a circuit does not necessarily end up with the obvious