Author Topic: How to test a SMPS?  (Read 742 times)

0 Members and 1 Guest are viewing this topic.

Offline jwhitmore

  • Regular Contributor
  • *
  • Posts: 137
  • Country: 00
  • Software head strays into the Hardware.
How to test a SMPS?
« on: December 01, 2017, 09:57:21 am »
I have a SMPS based on the Analog Devices ADP2303, that's giving me a 5V Rail, (eventually for USB Host), and there's also an MCP1825S which is giving me a 3v3 rail for the micro-controller. I've produced a board just with the power supply just so I can test is and make sure I'm happy with it, before I connect it to uc's and the like, but how do I test it? I know that's a very vague question but I'm thinking at a high level.

So for example the MCP1825s is rated at 500mA but if I used it to supply 450mA would it get hotter then I'd like? That's a LDO Regulator, as far as I remember, so that might be fairly predictable but the ADP2303, which is a 3Amp part. Will it get hot and is my layout design good enough for 3A or should I say that it's good enough for 2A and leave it at that. I'm sure that if you were creating a serious product and had the money a certification body would do all that needs to be done.

A question I would have is how long is long enough to test? So if I decided that I'd test pulling 400mA out of the 3v3 rail should I leave it drawing current for an hour and see if it got hot, or should it be left 2 hours, a day, a few days? I'd assume that if you connected a load it'd fairly quickly stabilise, in terms of temp. And should I be looking at Caps and other components as well for any signs of stress.

I'm now thinking certification would be cheap.
I could test that the unit blows up when I pull 5Amp but I'm not too concerned where it gives up I want to find where it's happy to be left on 24/7 pulling current. So if I pulled 500mA from the 3v3 rail and the 3A from the 5V that it'd sit there for a week without issue. Or perhaps my layout would need a bit more copper to get rid of a little bit more heat.

A BK Precision current load would be nice for Christmas, and certainly would be a great help, but I guess I'd need two.  If you don't have a BK Precision I guess some of the eBay cheapos would be a start.

You can probably tell that I've not got a clue so any pointers would be very helpful, thanks a million in advance.

Offline T3sl4co1l

  • Super Contributor
  • ***
  • Posts: 13966
  • Country: us
  • Expert, Analog Electronics, PCB Layout, EMC
    • Seven Transistor Labs
Re: How to test a SMPS?
« Reply #1 on: December 01, 2017, 10:29:59 am »
So you're doing...what? ???  You didn't actually say what you'd power the LDO from..

If you're powering the LDO from the 5V, then it will dissipate P = Iout * (5 - 3.3)V, give or take a small error.

The thermal resistance of the LDO is well known, from ~80 C/W (ballpark, I haven't looked it up) for a minimal footprint on FR-4, which is enough for about a watt at thermal limits (usually ~150C, which gives you an ambient limit of 150 - (80 C/W * 1W) = 70C, not bad at all).

Don't take my ramblings seriously: exact numbers are all in the datasheet!

Heh, I used ADP2303s like six years ago.  I recall availability kind of sucked, we had to buy a reel from Avnet of all places.  I suppose it's not any worse, probably much better available, nowadays?...

I'm not sure how you'd propose to draw 5A from either regulator -- the LDO will have a built in current limit and thermal limit, while the buck regulator will simply current limit at its rating (sometimes with foldback or hiccup, but I think that one does continuous limit?).  Neither will deliver that much current into a short circuit.  To draw more current, you'd have to "forcibly" draw current out of the circuit -- using a negative source, which might destroy the LDO (you'd be reversing the OUT to ADJ/GND pin polarity, which is usually protected with an ESD diode, nothing that'll handle 5A), and might damage the buck converter as well, but not so much because of the regulator chip (it's still limiting current) but because you'd melt the poor schottky diode beside it!

As for time, thermal time constants range from milliseconds (silicon die) to seconds (package) to minutes and hours (PCB, enclosure).  Just cook it until its temperature stops rising.  Keep in mind that LDOs have notoriously inaccurate current limits (again, don't take my word for it, read the datasheet and see!), so you should test many samples of the LDO, each with different lot code, to be sure your sample includes some reasonably-worst-case devices.

Better to conduct the design, from first principles, with adequate guard banding between maximum current limit rating, maximum voltage drop, thermal resistance of the layout (including whatever heatsinking you've provided for the device), so that the worst case condition remains sound.

Seven Transistor Labs, LLC
Electronic design, from concept to prototype.
Bringing a project to life?  Send me a message!

Online treez

  • Super Contributor
  • ***
  • Posts: 1097
  • Country: gb
Re: How to test a SMPS?
« Reply #2 on: December 02, 2017, 02:55:16 pm »
Well i could bore you to heck with a list of smps tests, so i wont, and ill just mention one...
Do a thermal test with a k type  thermocouple.
Glue  it to the metal tab of the fet, but you dont want it electrically connected, so put a tiny bit of that transparent orange tape (forgotten what its called) over the metal tab first.
Then glue the thermocouple end to that.
Use 2 part fast setting glue.
Thats not very professional, but proper thermal glue takes about 2 days to dry and is extremely expensive.
Make sure the test is done with the smps in the intended enclosure......ensure you  block out the exit hole of the thermocouple, for which you will have drilled a dont want any drafts in there which normally wouldnt be there.
Put the test assembly in a thermal oven and turn up to your maximum ambient.
Oh and by the way, ensure to have a second thermocouple in the enclosure but just in the general ambient, as a kind of reality check.

Other tests include:
Start up output voltage overshoot
Undervoltage input
Response to prolonged input  undervoltage
Response to output short cuircuit (hiccup or latched shutdown)
Transient response (sudden no load to full load and vice versa)…beware overshoot of vout when no load to full load.
Ripple voltage on vout (use proper probe)
Any staircasing of inductor current when vout is low, say in overload?
Etc etc etc

...oh and what about sudden application of vin as a step input...does it make your input LC filter resonante and produce a damaging overvoltage?


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo