Help debugging a non-isolated mains supply

[gxti]

My current project includes an on-board switcher that produces 5V from the mains input to run some analog circuitry on the mains side. It doesn't work. I've already decided that the next revision won't have a "mains side" at all and will be externally powered, but I can't stand the thought of leaving it non-working. So maybe someone experienced with switching power supplies will recognize the symptoms and give me some pointers.

I'm using a LNK305DG from Power Integrations in a high-side buck configuration with direct feedback. I've basically cut-and-pasted the reference design and just tweaked the reference voltage divider to 5V (instead of 12V) and replaced parts with ones more easily attainable from Digi-Key, but the values are otherwise identical. The behavior I'm seeing is that the supply won't start when fed with anything more than about 30VDC. It's "hiccuping" which means the IC is seeing an overload condition either due to too-high switching current or because the feedback is not working. With 16-30VDC input the supply starts and roughly holds regulation but is switching at 2KHz or less (66KHz is normal). This doesn't surprise me too much since the part requires 70VDC minimum to operate, but at least it shows that the fault is sensitive to input voltage. Under all conditions I see no more than 50mA average current at the input but I don't have a handy way to see what the peak current into the switch is.

With light load (e.g. a 1K resistor, still more than required per the datasheet) once it starts at < 30VDC I can crank up the voltage a little and it doesn't stop, but the output voltage also increases with it. This seems to indicate a problem with the feedback. With moderate load (100 ohm resistor) it immediately starts hiccuping when I increase the input voltage. With no load the output voltage increases with each hiccup, but it still keeps hiccuping and does not start. Again, this is not a normal operating condition so maybe that's not too indicative of a problem.

Also worth noting is that both diodes get fairly hot. In particular, the feedback diode gets to 50C which seems really odd. If I measure the voltage across it, the peak forward voltage is only 0.68V which according to the datasheet corresponds to a forward current of only 1mA. The flyback diode sees 1.36V peak forward voltage but doesn't get as hot (35-40C). These readings were taken with the 30VDC input, since that's the only condition where the supply will operate continuously.

I've tried reducing the capacitance by a factor of four but it makes the output rougher and doesn't otherwise change the behavior. I've also tried increasing it by a factor of four, again with no improvement. I originally specified placed a ceramic cap for the feedback capacitor but it made quite a lot of noise at 2KHz so I replaced it with a tantalum. I can still hear switcher noise if I touch either diode, or the main inductor.

Here's the schematic: http://hg.partiallystapled.com/circuits/ctservice/raw-file/6b2146b43551/out/production.PDF (see page 5)
Switcher datasheet: http://www.powerint.com/sites/default/files/product-docs/lnk302_304-306.pdf

Any ideas?

[nctnico]

What type of inductor did you use for L101?

[gxti]

This is the inductor: http://www.digikey.com/scripts/DkSearch/dksus.dll?lang=en&KeyWords=AIUR-02H-102K-ND&cur=USD
Full bill of materials here: http://partiallystapled.com/~gxti/trash/2012/12/09-bom.html

[nctnico]

That inductor looks like a filter choke (lossy core material), not an energy storage inductor. I'd use something like this:
http://uk.farnell.com/murata-power-solutions/49102sc/inductor-single-1000uh-20-0-50a/dp/2062863

[T4P]

Looks like a inductor to me 

[muvideo]

Hello, some thoughts:

- diodes: at what voltage will you use the switcher? Anyway the minimum is 70Vdc and probably better
stay above 85V, I would use higher (much higher) voltage diodes, 4148 are good up to 100V maximum.
- by the way, Voltage: minimum recommended is 70Vdc.
- DS calls for a 2kohm R101, because feedback network needs 0.8mA to work properly (pag7 of the DS)
- layout: did you try to setup the same layout that is in the DS? The components are few, I would try.
- the inductor seem ok, but you are running the LNK305, the DS example is for the 304, did you try to
use PIExpert spreadsheet to see what component values come out for your design?

Hope this help, good luck!
Fabio.

[gxti]

You are right about the diode, that was a silly oversight on my part. I was paying too much attention to the requirement to match the diode drops and decided to use the same part, but neglected to look at the reverse voltage. However, in my testing using bench power I'm at not even a third of that 100V rating before it starts hiccuping. Yes that's below the minimum recommendation but the behavior is identical to when running with the full 170V, and this way I can actually probe around the voltage where it starts failing to operate. I suppose I might have damaged the flyback diode while operating at full voltage, but they're both getting warmer than I'd expect yet only the flyback diode would have been exposed to high voltages.

Minimum load isn't a problem, I'm testing with 100 ohms or 1000 ohms which draws above the recommended minimum current. I did not build a load resistor into the supply because the actual load is always on. It's only partitioned while I diagnose this problem. Layout is not likely to be an issue I think, I would expect to see it performing less than perfect but not fail to start entirely by what seems like a large margin. In any case I followed all the recommendations to the degree that I could.

I'll take a look at PIExpert. Their website does not seem to be working well tonight so it'll probably be tomorrow.

[nctnico]

Looks like a inductor to me 

Not every inductor is suitable for use in a switching power supply! Its suitability depends on the gap and the core material. If the core material has too much loss then the power supply won't be able to deliver any output power. If the OP didn't make any mistakes in the layout and/or component directions then I'd look into the inductor.

[gxti]

The datasheet claims this series of inductors is for "switching power supplies" etc. and the saturation current limit of 510mA is surely met, so I don't know what else to look for.

Here's the layout, it's not going to win any awards but again I don't see why the circuit would totally fail to start because of that:

[nctnico]

Much of the black art of SMPS is due to poorly specified inductors. Just try the inductor I pointed to in my first post. I have bad experiences using the long bobbin kind of inductors in SMPS as a storage inductor.