LNK364- based SMPS failure (FIXED) - but is the output ripple excessive?

[Lorenzo_1]

The switch-mode power supply in a quite expensive dehumidifier (an NEYL2020A) popped its clogs some time back. Photo attached.  Cost $130 for a replacement in the meantime, so figured it worthwhile to try and fix as a spare and for learning.. I don't have a.differential probe and was wary about being too adventurous on the hot side of it. Did use an isolation transformer for the DUT and lots of reading before  probing. The circuit used is very similar to the one on p.3 of the LNK362-364 datasheet.

Main issue was no 5V or 12V output.  Input side looked all OK - and transformer seemed to have intact windings and be putting out secondary voltage.  However, it was shutting down and powering up every 800ms or so.  It did have two blown rectifier diodes on the output side (why is quite unclear to me). While waiting for those I replaced the optocoupler and the Linkswitch LNK364 switcher (an interesting all-in-one switching device - bottom of picture just to left of optocoupler) just in case it was playing up also - though the regular start-up sequence suggested it was likely OK.  All the caps checked out in spec so I left them be.  It came back to life nicely after surgery.  The 12V rail drives relays (via a Darlington array IC) for compressor, fan and air flap drives. The 5V rail supplies humidity and water level sensors and the control panel.

I had a look at the ripple on the 5V and 12V rails - it's running about 130mV p-p on each.  Does that seem about right or should it be tighter than that? The 5V rail has 2x220uF electrolytics to the low-side ground bridged by a 4.7mH inductor.  The 12V rail has a 330uF electrolytic to ground in parallel with a 110k resistor.  I have roughed up a Tina layout of the main part if that helps (or is otherwise useful to anyone). Wondering if it sounds about right and I can put it away till needed or whether the secondary filtering might need another look?

[Lorenzo_1]

Should have added that I measured noise with a FET 1x probe with a short ground pin on a Tek485. It's low frequency (ca 7.6kHz) and corresponds with what might be a duty cycle frequency in the LNK 364, that has an internal ~132kHz oscillator.  The attached screenshot is lower grade from 10x probes on my other scope but shows the 132kHz switching on a low load (0.2A draw) and the ~7,6kHz noise that corresponds to the longer period cycling.  The shape of the noise pulse is the same and the amplitude (ignoring spikes) 123mV.

[AndyC_772]

That looks like a supply which has no load (ie. not supplying any current), and it's gone into hiccup mode.

In this mode - which is completely normal - it will switch for a few cycles, transferring charge into the output capacitors. The output voltage increases to a point in excess of the set point, so the controller shuts down for a while and allows the output to discharge again. The cycle keeps repeating, and gives rise to the short bursts of switching activity you're seeing.

Try putting a dummy load resistor across the output, and you may see a considerable improvement.

[Lorenzo_1]

Thks AndyC. I don't have a decent dummy load but the shots were taken with 22Ohm 10W load drawing 200mA. With a higher load (0.8A draw) the switcher increases duty cycle and the number of 132kHz pulses in each 7-8kHz cycle and then shuts down (overload?) before starting again a second or so later. With no load on 5V rail, still runs but shuts down after 1x132kHz pulse - cf 3 as pictured previously.

 Perhaps I need a proper electronic load to test these devices? I've been thinking about it for a while.  Power supplies seem to be the most frequent point of failure - five in two months a while back. Can't run everything off a UPS. Our power here is pretty flaky - lots of tropical storms, blackouts from tree-falls and brown-outs. Gets expensive without a bit of DIY.

[Lorenzo_1]

Thks for the hint. Higher resistance load brought the noise down to +/-10mV and voltage is stable over a decent range of loads.  Will put on the shelf for a rainy day.

[Lorenzo_1]

Turns out the unusual signals in the photo I attached earlier were b/c of bandwidth limitation on my digital scope - not picking up the high speed +ve excursions in particular. Made my way into the working power supply on the unit and compared signals between working and repaired unit picked up adjacent to the transformer coils using Tek485 with a 1x probe. Photo here is a bit untidy on the triggering/signal overlap but both units show precisely the same switching pattern. Put the repaired unit back in while I had it stripped down and it's working fine. Compressor unit wouldn't start up with even a 100W current limit globe in line but turned on fine with direct connect after confirming everything else was OK.  Useful bit of learning about SMPS behaviour along the way.