The mystery of the Hi-Pot test

[cupoftea]

Hi,
As we know, Offline SMPS’s get tested for high voltage withstand between their primary and secondary sides. The test involves the application of a certain voltage, for a certain time, and during that time, the leakage current between primary and secondary must stay below x milliamps.

How the test voltage, length of test time, and leakage current is calculated is completely unknown.

....Its not possible to calculate these things, and if you want to know such things, then you must purchase expensive standards documents. The number of standards documents that an engineer would have to purchase in relation to an offline SMPS development, is a great many…and no engineer could possibly afford to buy all these standards. The company that an engineer works for will not lend out the standards documents to any engineer, -because if a product gets shipped without passing one of the standards, (which they so often do) then the company doesn’t want people knowing about it…so keeps their own standards documents to themselves.

Sometimes the “Hi-Pot” test voltage is AC at 50Hz, and sometimes its DC.

Its impossible to calculate what [DC test voltage /Test Time /Leakage current] corresponds to what [AC test voltage /test Time /Leakage current].

But , ….supposing you do know that your product must pass "2500VAC for 1 minute, and the leakage current must stay below 3mA for that one minute"…

Then you pick up a Hi-pot tester and set it to 2500VAC….and you count to 60 seconds and hope that it doesn’t flash red (fail) at you. Even if it doesn’t flash red, you still don’t know whether your product really passes or not, because the Hi-pot testers you get never have the leakage current threshold written on them, and they are usually so old that they have no datasheet. So fingers crossed and goodness help you.

Then you get another offline SMPS to test, and it again is [2500VAC for one minute and 3mA]…..but at this company they only have a DC Hi-pot tester, instead of an AC one. So of course, you simply calculate what [DC/time/mA] corresponds to what [AC/time/mA]….

…except you don’t, because there’s no way of calculating that.

So you use your sense of logic…..and realise that you must apply a DC voltage of 2500 x SQRT(2).

But for what time?....well, less time than the AC tester that’s for sure, since the AC tester isn’t always at its peak level. Surely an equivalent length of time would be [0.63 * one minute], ie 38 seconds(?) This is because the average of a sine is [0.63 x peak]. Again, you would have to assume that the current threshold would be 3mA…….and lucky you if your DC hi-pot tester allows you to set the current threshold, most don’t, and you’ve no idea what it is.
Whilst you are fumbling about, wondering about this, the other (time served) Engineers in the company declare to the boss, that you are obviously a 2-bit idiot, and that you should spend the rest of your working life doing nothing more than replacing dirty solder-tip-wiper-sponges on the lab benches, and scrubbing the conveniences, whilst you're re-filling their solder water bottles.

I mean, it isn’t as if the regulatory hi-pot [Voltage/time/current] settings make any particular sense for offline SMPS. For example, take an SMPS with a Boost PFC front end, with its output capacitor bank…….there is virtually no chance that a typical 50us transient (Line_Neutral) is going to lift the voltage of that capacitor bank up more than a few 10’s of volts anyway……and certainly not up to 1kV. So why the voltage level has to be up to 2500VAC is anyone’s guess. Its appreciable that there should be a length_of_test_time, because high voltage events are cumulative in the damage that they do, so its good to apply the test voltage for sufficient time.
Though, suppose a 2.5kV voltage spike came to the offline SMPS between line and earth. Then , if the offline SMPS output was earthed, then there could be a flashover due to the isolation barrier seeing the full 2500V. [remembering mains cable stray L meaning that for the short transient duration v(neatral)-not-equal-to-v(earth)]…..but no…..there couldn’t be…..the secondary earth would be at neutral potential (and obviously earth potential) and so once again, this transient would simply get quenched to <<1kV by the PFC output caps.
So why we always test to 2500VAC is anyone’s guess.

How long do you think the Hi-pot mystery will continue for?

 

[Easy peasy]

Also you need to know the capacitance for P+N to earth, and output ( + & - shorted ) to earth

the application of say 2500Vac will disport itself across pri and sec ( earth floating ) according to these capacitances - if the output cap to earth is too small it will see too many volts and may upset heatsinking to earth on the sec side.

For a 5 : 1 split, e.g. 2kV and 500V ( the earth is essentially in the middle ) you need 4 x the mains to earth cap on the sec side,

the current can be easily calculated knowing the various C's and voltage and freq.

--- Updated Nov 12, 2022 ---

ideally your psu should survive 2500Vac (P+N) to earth and 500Vac output to earth as well as the input to output spec.

 

[FvM]

How the test voltage, length of test time, and leakage current is calculated is completely unknown.

I understand that you want to demonstrate your confusion about applicable safety standards and test procedures for a specific product. If we read this post, we woud hardly guess that you are in the electronic design business since a long time. I feel it's more a matter of cultivating ignorance.

I concede there's a problem of often having multiple safety standards that can be potentially applied to a specific product. Often it's the application or country of use which decides.

Regarding AC versus DC tests and required test time, you need to distinguish between type approval and routine tests as well as component and device tests. A transformer is usually tested with AC, always 1 min for type approval, typcally 5 sec in series production. The device, e.g. SMPS may not withstand 2.5 or 2 kV AC input to PE because the Y capacitors are only specified for DC test. So you go for DC device routine test. And so on.

IMHO you can hardly avoid to study relevant safety standards when you design SMPS.

 

[cupoftea]

I feel it's more a matter of cultivating ignorance.

Thanks, i would say years of unsucessfully trying to find out the "logic" behind the hipot standards (knowing the logic in the hope it makes it easier to remember). That is,
1.....Exactly what "situation(s)" the standard is trying to instigate protection against.
2....What exactly is the standard in terms of V/t/i...and hopefully...why that v/t/i.

...Though thankyou, as i will be storing away this post, one of the best things ive read about standards in many a year.

ideally your psu should survive 2500Vac (P+N) to earth and 500Vac output to earth as well as the input to output spec.

Thanks,, i worked in a place producing millions......they had an offline SMPS (with L.N.E input) which had earth tracking just 0.3mm away from line and Neutral tracking, on the SMPS PCB. When i questioned this, they said it didnt matter, because if Line shorts to earth, the fuse will blow and it'll be safe....The earth tracking, on this PCB....started, expectedly, at the input connector, and came across the isolation barrier.........connecting to the heatsink solder PTH's of the primary fet and secondary diode heatsinks on its way (it was a flyback).....then tracked into the secondary ground. On the primary side, this same earth tracking, as i said, was just 0.3mm away from the line (230VAC) tracking.
___--___

Below for example is what one customer specified for a 500W Offline 115vac PSU...

********FROM CUSTOMER ****** The ACDC PSU shall be designed to withstand a single application of 2000V AC RMS at frequency in the range 25 Hz to 100 Hz for one minute.
The test voltage is to be applied between the power input terminals and the earthed chassis. All the conducting circuit outputs of the unit under test are connected together and connected to chassis. During the test, the unit under test is to be kept under continuous observation. Breakdown of any insulation or flashing over because of leakage paths will constitute a failure of the unit under test.

.....as you can see, no specification of test current threshold, so its completely meaningless.

The customer also requests...

*****from customer****** The test is to commence at a voltage of approximately one-third of the maximum value and increased to the maximum value as rapidly as possible consistent with its accurate attainment without transient overvoltage. The full test voltage is then maintained for one minute. At the end of one minute the test voltage is to be decreased rapidly to one-third of its full value before switching off the test voltage.

.........I have never seen a Hipot tester that could be programmed to ramp the voltage in that way....they dont even say what dv/dt they want.

 

[FvM]

A linear power-on ramp 0-100% is provided by most Hi-Pot testers. It's particularly useful for DC tests with larger DUT capacitance to avoid false tripping. Discharge at test end is required not leave the DUT with charged capacitors.

I agree that the described "1/3 of maximum" start voltage is no usual feature, I also don't see the purpose. It can be of course implemented with remote controllable testers.

If the customer is requiring unusual test procedures, I would ask 1. which safety standard the requirement is based on, 2. which approved tester is able to perform the test

--- Updated Nov 12, 2022 ---

Ramping is required in some test standards also for AC, as an example IEC 60598-1 (for luminaires)

Initially, no more than half the prescribed voltage shall be applied, then it is raised gradually to the full value.

As you see, a linear 0-100% ramp fulfills the spec.