Need help debugging small SMPS

[Martinn]

Hello all,

I have a defective Haidenhain display unit here. PSU is a SMPS with > 7 V output and a linear post regulator. If I feed the linear regulator with 7 V from a lab PSU, the unit works. Current draw is around 450 mA. So putting in a 7+ V module is a fallback.
I think I reasonably well understand how these supplies work, but I have never repaired one. So here we go. I have connected a HV lab supply, set to 200 V, to the input (makes me feel better than unlimited power variac). Unit is labeled as 100-240 V, so I assume 200 V DC to be fine. The design looks very German - the original Siemens&Halske logo on many components. Controller TDA4605-3 stamped "Austria". I think I have never seen this. Date code on other chips is around 97.
Controller datasheet: https://mm.digikey.com/Volume0/opasdata/d220001/medias/docus/1135/TDA4605-3.pdf
I removed the clip on heatsinks and also unsoldered the switch FET (BUK 446), which was fine according to my transistor tester.
The startup supply resistor chain looks pretty burnt up, some resistors have even cracks, but resistance value is OK for all of them.
Overall behaviour: VCC capacitor charges, controller starts at about 12 V, switches for around 40 ms, then stops. This repeats itself indefinitely.
To me, all switching and secondary side waveforms make somehow sense, no short or otherwise (obviously) dead component.

I attached a scope plot of this behaviour, it shows the VCC pin 6 of the controller. What I find strange is that at some point the primary feedback/bias winding should power the controller, however switching seems to stop before that happens.
Also strange is that there is a high 4Vpp ripple on the voltage once the controller switches, as the controller supply is bypassed with a 470 uF capacitor. What looks like some greyish goo having leaked out of the capacitor is present on the output cap as well, is rock hard and seems some epoxy glue.

Any ideas? Thanks!

[drvtech]

Almost certainly the small electrolytic on the Vcc rail to the chip (pin 6) has dried out. Number one problem on switch mode PSUs. The goo you mentioned is probably just glue to hold the capacitor down and take the stress off the pins.

[max.wwwang]

... The goo you mentioned is probably just glue to hold the capacitor down and take the stress off the pins.

A bit off-topic but this is something I have been wondering about for long. Why is such goo necessary, or why are these caps so susceptible to excessive stresses that they need such means of protection? Just because of their size? Appreciate any insights that would help clear my wonders. Cheers.

[MathWizard]

With some form of current limiting, like a resistor even, I would apply power to just the low voltage Vcc rail, after the diode from the transformer winding tho of course. Not that it's hard to remove a cap to check, but maybe you could see where any current is going.

[kripton2035]

+1 with drvtech

[drvtech]

... The goo you mentioned is probably just glue to hold the capacitor down and take the stress off the pins.

A bit off-topic but this is something I have been wondering about for long. Why is such goo necessary, or why are these caps so susceptible to excessive stresses that they need such means of protection? Just because of their size? Appreciate any insights that would help clear my wonders. Cheers.

In an assembly where there's vibration (and especially in subwoofers) it is possible for the solder joints to fatigue and crack. By stopping movement of the body of the cap relative to the PCB you prevent this. It's a genuine problem, not just a belt-and-braces thing. I've seen it lots of times.

[max.wwwang]

... The goo you mentioned is probably just glue to hold the capacitor down and take the stress off the pins.

A bit off-topic but this is something I have been wondering about for long. Why is such goo necessary, or why are these caps so susceptible to excessive stresses that they need such means of protection? Just because of their size? Appreciate any insights that would help clear my wonders. Cheers.

In an assembly where there's vibration (and especially in subwoofers) it is possible for the solder joints to fatigue and crack. By stopping movement of the body of the cap relative to the PCB you prevent this. It's a genuine problem, not just a belt-and-braces thing. I've seen it lots of times.

Thanks. That makes sense. So that's pretty much about mechanical reasons such as vibration and perhaps thermal stress, etc., not something electrical or anything else. Largely I wanted to know if there are any considerations other than mechanical. I understand from personal experience that repeated/prolonged exposure to something that looks trivial may have an effect far greater than what we would imagine from intuition.

[Martinn]

Almost certainly the small electrolytic on the Vcc rail to the chip (pin 6) has dried out. Number one problem on switch mode PSUs. The goo you mentioned is probably just glue to hold the capacitor down and take the stress off the pins.

Great! Managed to remove the glued in cap (heat shrink broke, glue was strong as new), measured 100 uF (labelled 470 uF) ESR 40 Ohms.
Rummaged through my old caps box and found a matching one, measured OK, but had visible leakage signs. I think I should simply throw all >10yr old electrolytics in the bin. I already attempted this once, but there are so many different voltages and capacities that purchasing a replacement assortment seems to get quite excessive. Found a (new) Chinese low ESR one and it works!
I think one could even have seen some darkish contamination on the - trace, see detail image.

Would you also replace the primary bus and secondary output capacitor (glued in)?

Not sure what to do with the startup resistor chains. These are pretty burned up, but values are OK and the traces lift off if you barely touch them. I guess I let them be. A shame these heat all day long although they are needed only a few seconds at startup.
Interestingly there seems to be a circuit with an 800 V BJT, a PTC and 3k series resistor near the input, probably to discharge the massive input X capacitor (so you don't get zapped when you touch the plug terminals after unplugging mains).

Thanks a lot, you have been a great help!

[Martinn]

Thanks. That makes sense. So that's pretty much about mechanical reasons such as vibration and perhaps thermal stress, etc., not something electrical or anything else. Largely I wanted to know if there are any considerations other than mechanical. I understand from personal experience that repeated/prolonged exposure to something that looks trivial may have an effect far greater than what we would imagine from intuition.

What stands right out: There are lots of bulky/heavy components on the PCB, why did they only put glue on two (out of three) electrolytics and nowhere else?

My conclusion: This is expensive industrial equipment, built and tested to certain standards. They have requirements for vibration resistance, put their PCB onto the shaker to verify it and it failed at those two components. They put glue on it, passed the shaker test, bingo.

I once worked with electronics to be mounted on automotive engines and gearboxes. Vibration testing was fun. Heavier components like PQFP would disintegrate seconds after shaker was switched on.

[inse]

BS removed

[drvtech]

Almost certainly the small electrolytic on the Vcc rail to the chip (pin 6) has dried out. Number one problem on switch mode PSUs. The goo you mentioned is probably just glue to hold the capacitor down and take the stress off the pins.

Great! Managed to remove the glued in cap (heat shrink broke, glue was strong as new), measured 100 uF (labelled 470 uF) ESR 40 Ohms.
Rummaged through my old caps box and found a matching one, measured OK, but had visible leakage signs. I think I should simply throw all >10yr old electrolytics in the bin. I already attempted this once, but there are so many different voltages and capacities that purchasing a replacement assortment seems to get quite excessive. Found a (new) Chinese low ESR one and it works!
I think one could even have seen some darkish contamination on the - trace, see detail image.

Would you also replace the primary bus and secondary output capacitor (glued in)?

Not sure what to do with the startup resistor chains. These are pretty burned up, but values are OK and the traces lift off if you barely touch them. I guess I let them be. A shame these heat all day long although they are needed only a few seconds at startup.
Interestingly there seems to be a circuit with an 800 V BJT, a PTC and 3k series resistor near the input, probably to discharge the massive input X capacitor (so you don't get zapped when you touch the plug terminals after unplugging mains).

Thanks a lot, you have been a great help!

If the resistor chains are still the right value I'd leave well alone otherwise you might end up with more board/track repairs. I also wouldn't bother doing anything with the caps if the supply is functioning correctly and they're not showing signs of distress (bulging tops, scorching, electrolyte leakage etc.)
I've never seen a BJT used in the way you describe but there's always a first time. Usually any semiconductor that close to the incoming mains is either part of an inrush limiter or a power factor correction circuit. However, there's nothing to drive a PFC circuit and no choke other than what looks like a common mode filter (big blue thing) so not a PF circuit. Also there's already an NTC surge limiter so how much inrush limiting do you need! Unfortunately some of your pictures are no longer visible so can't trace the circuit.

Edit to add:
Regarding inrush limiting - that looks like a pretty big cap to be charging through that puny little bridge rectifier so maybe it does need a lot of limiting. If the unit is supposed to bridge fairly large mains dropouts you would need an oversized reservoir cap so perhaps that influenced the designer's choice.

[max.wwwang]

What stands right out: There are lots of bulky/heavy components on the PCB, why did they only put glue on two (out of three) electrolytics and nowhere else?

My conclusion: This is expensive industrial equipment, built and tested to certain standards. They have requirements for vibration resistance, put their PCB onto the shaker to verify it and it failed at those two components. They put glue on it, passed the shaker test, bingo.

I once worked with electronics to be mounted on automotive engines and gearboxes. Vibration testing was fun. Heavier components like PQFP would disintegrate seconds after shaker was switched on.

Good point and thanks for sharing your first-hand experience. This is the kind of information I like most, specific and definitive.

I once bought something online from another city, which was big and heavy and I needed to improvise a 'cradle' with scrap wood and load it back with my truck. And the tailgate had to be left open. Because of its length extending beyond the tailgate, I was required by the road code to conspicuously indicate the aft end of my vehicle when on the road. I used a hi-viz vest for this. I drilled a small hole on the end of one piece of wood of the cradle and attached to it the hi-viz with wires (thin, but I made several turns). The hi-viz was very light, and it would not be subject to any load other than just suspending there. All good and then I hit the road. When I got home and started unloading. Strange! Where has the hi-viz (and the wire) gone?! I have no definitive answer. Most likely, due to the hi-viz constantly waving in the wind when the truck was running, subjecting the wire to repeated and cyclic stress (though negligible in terms of magnitude) and the wire succumbed due to fatigue and fell off somewhere who-knows with what it was supposed to secure!