Switching Mode Power Supply half output problem

[jward]

I’m trying to fix a Switching Mode Power Supply for an audio component that is producing only half of the designed output voltage. It has 12V and 5V outputs that are now only 6V and 2.5V. I replaced an obviously bad capacitor on the output of the bridge rectifier and the unit initially worked fine for about a minute until it went off again with the current symptom. It is switched by a Power Integrations TOP245PN (datasheet here). The design of the SMPS is similar to the one in Figure 44 but stripped down to just the 12V/5V outputs with a few component value differences.

I don’t know what the output of the transformer should normally be but it appears that the duty cycle may only be half of what it should be according to what I’ve seen with an oscope.

The 5V output (blue) is overlaid on the 12V output (green) measured at the transformer secondary. The time between the yellow/magenta markers is ~25ms (40Hz).
Could this apparently half-duty cycle be normal or be the problem?

 

[Nigel Goodwin]

Well it's a bit bizarre, in that it seems to be trying to regulate too rails (5V and 3.3V) in one supply, which can't be done - so it can only try and 'average' between them.

However, by FAR the most common cause would be electrolytic capacitors (because they fit sub-standard ones), particularly C13, 14, 16 and 17.

Try placing the scope on the +ve's of C13 and C14, and see what the ripple is like.

 

[dr pepper]

R1 might have gone high, or possibly one of the rails is overloaded or shorted.

 

[ronsimpson]

Do you have a load on the supply?

 

[jward]

The example schematic includes a 3.3V output but the actual design has only 12V & 5V. Replacement capacitors have been ordered but I'd like to understand the cause of the problem based on the indications. The ripple at C13 (used for 5V output) is 4mV p-p. At C11 (for 12V output), ripple is 2mV p-p.

 

[jward]

Measurements have been with no load. In the brief period the PS worked after the cap replacement, it provided the proper voltages with and without the normal load.

 

[unclejed613]

first thing i would try is measure the DC voltage across the TL431. it should be about 2.5V. if the TL431 is shorted, the LED in the opto will light at too low a voltage, and scale your output voltages accordingly. i've repaired supplies where the TL431 is shorted, and now it's one of the first things i check if the supply seems to be running, but with low output voltage.

 

[ronsimpson]

first thing i would try is measure the DC voltage across the TL431

It is a 3 legged part. What is "across".
If you short out the TL431 with your probes, depending on which legs you short, the PWM will go full on or full off. Full on is not good.

I think the scope is showing that the supply is turning on and off at 40hz. (burst mode) It is hard to see but I think the PWM is turning off for 12mS our of 25mS. This often happens when the supply has no load, or some time when the supply see too big a load.

 

[ChrisP58]

It is a 3 legged part. What is "across".
If you short out the TL431 with your probes, depending on which legs you short, the PWM will go full on or full off. Full on is not good.

I think the scope is showing that the supply is turning on and off at 40hz. (burst mode) It is hard to see but I think the PWM is turning off for 12mS our of 25mS. This often happens when the supply has no load, or some time when the supply see too big a load.

I agree.

It looks to me like it's starting up, seeing that something is wrong so it shuts down, then restarts.

My best guess would be on something that would appear as an overload condition to the TOP245.

 

[ronsimpson]

If the supply is unloaded: It might take 50 cycles to get to full output. Then with out load the output is at or even above where it should be so the duty cycle goes to zero for a long time. The output voltages slowly drop back to where the IC starts switching again. (burst mode)
If there is a near short somewhere (maybe a damaged diode), or too high a load, bad cap, etc : The current limit is hit hard causing the IC to shutdown and delay then restart. This restarting is a second form of (burst mode).

 

[ChrisP58]

jward I'm assuming that the behavior that you show in the scope capture in your original post goes on continuously?

I'd be interested is seeing a couple of other scope images:
1) Ch1, 5V trans secondary. Ch2, 5V DC output
2) Ch1, 5V trans secondary. Ch2, the cathode of the TL431
As for the time scale. That would be long enough to show either a repetitive behavior, or reaching an asymptote.

 

[jward]

1)
Ch1 (green): 5V trans secondary. 1V/div AC coupled
Ch2 (blue): 5V DC output. 1V/div DC coupled
Timebase: 10mS/div
Markers: 20mS (50Hz). Note: input is 120VAC 60Hz

2)
Ch1 (green): 5V trans secondary. 1V/div AC coupled
Ch2 (blue): the cathode of the TL431. 1V/div DC coupled
Timebase: 10mS/div
Markers: 20mS (50Hz)

 

[ronsimpson]

It has 12V and 5V outputs that are now only 6V and 2.5V.

I do not know where the opto U2 is getting its power. If it is 2.5V then there is not enough voltage to turn on the LED. If it is 6V ... maybe the LED is not on.

If U1 sees too much current it will reset and do a softstare. That will look much like what you are seeing.

Please unplug the supply. Connect a bench supply to +12 or +5 and see if you can lift them. Each supply should only need mAs to lift to 12/5 volts. (this is a simple way to see if the diodes are OK. (not a good diode test but better than nothing)

While you are there test D6, and make certain that VR1 is open.
Connect across C3 and see if it will take 5 volts. (test D6) Should take 0mA,

 

[unclejed613]

the TL431 sets up a reference voltage of 2.5V, and another 2.5V is across R7 and the opto LED. if the TL431 is shorted A-K, the LED turns on when the 5V rail is 2.5V. all of the other output voltages will be reduced proportionally (50%). the pinout of the TL431 is 1) REF, 2) Anode, 3) Cathode. since the TL431 CAN fail "dynamically" (i.e. shows ok with an ohmmeter, but fails with voltage applied) the way to check it is measure across C20 when the supply is operating, and a normal reading will be 2.5V, and will read near zero if it's bad. the other thing that will cause the supply would be a short on one of the rails. that usually shows up as most of the rails at reduced voltage, and the supply in burst or chirp mode, AND one of the rails having a disproportionally low voltage (i.e. all the rail voltages at 50% and one at near 0V). the one near 0V will be the one with the short. a third possibility is one of the small electrolytics on the primary side (C5 or "the other" C3 which is connected to the transistor side of the opto) is dried out and has high ESR.

 

[ronsimpson]

We do not have a real schematic. Post # 13 might be close.
To test the TL431:
Unplug the supply from the wall. Apply 11 and 4 volts to the 12 & 5V outputs. The TL431 output should be as high as it can be. No LED current.
Then move the supply voltages up to 12.5 and 5.5 (or something slightly high). The TL431 output should be as low as it can be. Two to 2.5V.
The TL431 should flip high/low at about 12 and 5V. You should be able to know if the TL431 is watching just the 5 or just the 12 or both.

 

[jward]

While waiting to receive the replacement capacitors, I prepared to continue troubleshooting based on the advice given. It was then that I discovered that the circuit deviated from the example even more than I had initially mentioned, particularly with the addition of Q1. Because of the kind efforts of those who took the time to read about the problem and especially of those who offered advice, I felt obligated to begin the laborious task of partially reverse-engineering the board to have a more accurate schematic, which is shown below.

In the schematic, all components are accounted for but I didn’t go as far as examining the value of each component, which isn’t always necessary for troubleshooting anyway. By the time I got this far, I had the replacements. I first replaced C11 and C12 because C12 seemed most likely to cause my symptom. They were also identical in value and were adjacent to each other so for convenience, I replaced them at the same time before testing. This fixed the problem. Unfortunately, because I did both at the same time, I don’t know which one fixed the problem so I partially lost my academic exercise in properly troubleshooting this circuit. Most importantly though, the unit works and is in service. While I was at it and had new electrolytic caps, I replaced the others as well which wasn’t absolutely necessary because the old ones still worked fine but maybe the replacements will postpone another repair.

Thanks for the troubleshooting tips. For others who may benefit from this, this schematic is of the switched-mode power supply (SMPS) for a NuVo tuner model NV-T2DFG. I was unable to find a schematic of it from other sources. The tuner is discontinued and the company was sold to Legrand.

 

[Nigel Goodwin]

I first replaced C11 and C12 because C12 seemed most likely to cause my symptom. They were also identical in value and were adjacent to each other so for convenience, I replaced them at the same time before testing. This fixed the problem. Unfortunately, because I did both at the same time, I don’t know which one fixed the problem so I partially lost my academic exercise in properly troubleshooting this circuit.

If you had an ESR meter, then you could easily have checked all the capacitors in circuit quickly and simply - it's been a VERY useful devices over the last couple of decades. It's quite possible BOTH were faulty anyway, as they were sub-standard quality in the first place, they should both be replaced - as well as all others.

You could also have simply tested the capacitors using your scope as well, BUT you would have needed to take suitable isolation precautions (which I'm not going in to, as it's an often argued situation on these forums).

 

[unclejed613]

Unfortunately, because I did both at the same time, I don’t know which one fixed the problem

i used to do computer monitor repairs (CRT), and there were several brands that were OEM by Tatung (later versions of the IBM 8514 for example, as well as tatung branded monitors). these monitors had a power supply module, and all of them had basically the same schematic, but some component values were different between brands. there were 3 small electrolytics on the primary side that would dry out and have high ESR. i called them the "3-stooges". it didn't take long to figure out it was best to replace all three. if one of them were dried out, chances are the other two aren't far behind. in a production environment, you figure out what a) fixes the problem, and b) what eliminates follow-on failures, and you take care of both problems by replacing the common failure items.

the large caps (like the 47uF/400V one) don't fail as often, but if they do, they generally dome or split on the top pressure relief seams (looking like something out of "Day of the Triffids" or an egg pod in "Alien". for some reason, those large caps smell like chocolate when they fail, the smaller ones usually smell like burnt fish.

 

[Nigel Goodwin]

the large caps (like the 47uF/400V one) don't fail as often

No they don't, but they aren't under the stress of the others as they essentially run at 100/120Hz, where all the others run at a MUCH higher frequency, it's the high frequency running that particularly seems to kill capacitors.

When they do fail though, it's almost always that they are totally O/C - and in such a case they commonly blow the PSU in the process, although some sets (in particular the above mentioned Tatung) seem to happily survive it, and even still work but with a hum bar.

You can't even seem to buy 105 degree versions of the big caps, or low ESR ones.

 

[jward]

Nigel Goodwin: Thanks for the ESR meter suggestion. It prompted me to look online for one. The last time I used one was so long ago that the only ones available (e.g. Fluke, B&K Precision, etc.) were prohibitively expensive for as infrequently as I’d use one. I ordered a cheap off-brand ESR meter just to have on hand.

unclejed613: All of the replacement electrolytic caps I ordered were small and rated for 105°C even though the enclosure doesn't get much above ambient.