Power supply calibration procedure

[tszaboo]

Hi. I just got a task to calibrate all our power supplies, because one of our visiting customer wasn't happy with the "NOT CALIBRATED" stickers on them. Sending them out for cal would be quite pricey, probably scrapping them would be cheaper. So I'll attempt to calibrate it in-house.
The DUT is nothing fancy, U8001A power supply for example, 30V 3A, linear. And a bunch of similar supplies.

My plan was to place 2x 34465A parallel + series with the output, route it into a DC load (Keithley 2380) and verify DC ranges, CC and CV.
However when I checked the user manual of the power supply, there are extra steps to verify CV, CC noise. And step response.
https://www.keysight.com/nl/en/assets/9018-02142/user-manuals/9018-02142.pdf?success=true
They also use a 100MHz differential amplifier, which is about 6KEUR, and the only reason to use it is to AC couple the signal and a gain of 10? And an RMS voltmeter... Seems a bit excessive to just measure the 1mV noise specification.

How would you do a basic check for a power supply noise specification? Am I thinking about this too much, just place a MSO on it and measure the RMS?

[penfold]

You could always try the flippant approach and use "FOR INDICATION ONLY" if they're unhappy with the negative connotations of "NOT CALIBRATED".

In all pragmatism, if you're just interested in knowing whether or not the ripple is just ridiculously out of spec (i.e. not re-verifying manufacturer's spec's and instead just indicating broken/not-broken... on a PSU that you would normally not calibrate), I see no harm in just checking it with a scope+traditional-probe... (if you have a 1mV/div input range) it'll show you whether or not its ~1mV or ~2mV etc. But of course, indicate such uncertainty in your calibration cert. Especially on something like a bench PSU which would be used to power a circuit which itself should have some tolerance to ripple/noise/transients etc.

[BradC]

Just cover the sticker with a “QC Passed” sticker. Seems to work everywhere else.

[MiDi]

Just cover the sticker with a “QC Passed” sticker. Seems to work everywhere else.

 

Customer should be grateful that the psus have a sticker...

[thermistor-guy]

Just cover the sticker with a “QC Passed” sticker. Seems to work everywhere else.

Seems to me that the OP is performing a functional test, and possible adjustment and repair, rather than a calibration.

"Calibration" to me means (a) measurement using equipment traceable to primary standards, using (b) an audited measurement procedure, plus (c) an uncertainty calculation.

Since it is a functional test, yeah, add a sticker with "Pass/Fail: ______________ Tested on:  ________________ Next test due:  _________________ "

[tszaboo]

Just cover the sticker with a “QC Passed” sticker. Seems to work everywhere else.

Seems to me that the OP is performing a functional test, and possible adjustment and repair, rather than a calibration.

"Calibration" to me means (a) measurement using equipment traceable to primary standards, using (b) an audited measurement procedure, plus (c) an uncertainty calculation.

Since it is a functional test, yeah, add a sticker with "Pass/Fail: ______________ Tested on:  ________________ Next test due:  _________________ "

Correct. I'm only trying to ascertain, if the PSU is performing according to it's specification, or the performance is acceptable for the intended application. For example, having 1mV noise specification is nice to have, but unnecessary. Also the output voltage accuracy is also fine being in the ballpark.
For example there is one PSU which is used to power boards, when they are being programmed on the production line. After this they assemble in the enclosure and connect a battery to it. So having 2.5-3.6V on the power supply would be fine, as this is the working voltage of the primary battery.

You could always try the flippant approach and use "FOR INDICATION ONLY" if they're unhappy with the negative connotations of "NOT CALIBRATED".

I wish. The guy apparently worked as quality manager in an automotive company, and wants to have similar standards at us. He even looked at the online database where we keep track of the the cal certificates and dates. Due to the pandemic, a lot of things were due of course, and it's ny job to clean up the mess.

[guenthert]

[..]
They also use a 100MHz differential amplifier, which is about 6KEUR, and the only reason to use it is to AC couple the signal and a gain of 10? And an RMS voltmeter... Seems a bit excessive to just measure the 1mV noise specification.

How would you do a basic check for a power supply noise specification? Am I thinking about this too much, just place a MSO on it and measure the RMS?

     Performance evaluation / calibration sections in reference manuals routinely list equipment the manufacturer had at hand in the lab or likes to advertise.  Substitute as needed.  The noise specification needs to list the bandwidth.  If it happens to be 10MHz, I would use my HP3400A.

[alm]

I'm guessing they specified the differential amplifier to exclude common-mode noise from the measurement. If your spec is looser, this might not be as much of a problem. Depending on the bandwidth you care about, I would consider using a decent handheld DMM with 100 kHz bandwidth on its AC mV ranges, like a Fluke 189/289. This would give good common-mode rejection, and would make the measurement very simple.

[mendip_discovery]

"Indication Use Only" should be ok but if the readings are relevant to your work then either have calibrated multimeters connected to the outputs or calibrate them.

Options are:
1, Use a calibrated meter, depending on the accuracy needed it could be a 3.5 or 6.5 digit meter and do voltage/amp checks. One point is the one you use for work. Record Readings.
2, Same as above but you also test for how much the voltage drops under load at one point.
3, Same as above but you also hook up to a scope and check for ripple.

It gets more fun if you want to add a decision rule and uncertainty for measurement but that isn't too bad.