IEC 60601-1 clause 8.8.3 - Dielectric Strength

[WRGandolphini]

Hi all,

The standard says:

Compliance is checked by applying the test voltage specified in Table 6 for 1 min:

Initially, not more than half the test voltage is applied, and then it is gradually raised over a period of 10 s to the full value, which is maintained for 1 min, after which it is gradually lowered over a period of 10 s to less than half the full value



Table 6 - Test voltages for solid insulation forming a PROTECTIVE MEANS

In a production line, all equipment must undergo this routine test, correct?

My doubt, would I have to strictly follow the 10s ramp-up of the voltage, then 1min in the voltage value according to table 6, and after 10s of the ramp-down of the voltage?

Or a reduction in these times can be made because we know that solid insulation can be destroyed, right?

Are there any acceptable manipulations that can shorten routine testing time without affecting the requirement of the standard?

Thanks,

 

[WRGandolphini]

Hi all,

Could someone help me with the above question?

Thanks,

 

[Mike S.]

I can't answer your question, looks like no one else here can, either.

Often a standard will contain contact information, I would contact the creator of the standard and ask your question. I have done this before with several standards and it can be very helpful.

 

[Benjamin Weber]

There is no strict requirement for dielectric strength test in production lines, since 60601-1 is for type testing of the design/construction. IEC 62353 describes recurrent tests after repair. But there is no dielectric strength test. It's leakage currents, PE impedance, visual inspection and functional testing. Additionally there is inof on production line testing in IEC TR 62354, Annex K. Here a dielectric strength test is described, yet it only an informative annex, not normative.

 

[Peter Selvey]

Agree with Benjamin, there should be no fixed requirement to test dielectric strength in production. Dielectric strength is actually an ageing test for solid insulation, and really intended to be applied to individual parts, not assembled devices.

The aging aspect is theoretically applicable, and could be used as an argument to avoid production testing since the test could damage or shorten the life of the insulation. However, virtually all modern insulation can hande far longer times than in the standard, so practically it's not an issue.

Rather, the bigger problem is that in an assembled device, there is a huge potential for voltage to be distributed in a way that damages parts not intended to act as safety insulating barriers. Also it's common that the parts providing safety insulation are not actually stressed in tests performed from outside an assembled device.

With that in mind, a high voltage test can still be a useful production test to check for assembly errors. In this case, the 60s time is not relevant, it just needs to be enough to be sure the voltage is reached reliably e.g. 5s. Even the voltage does not need to match the standard. Such a test should be carefully planned with designers involved to make sure it does not damage parts. Mains to earth is usually OK, but mains to secondary, patient circuits, accessible parts usually cannot be done, or should be done in sub-assemblies only (if not already done by the part supplier).

 

[KB2021]

If you have an NRTL report for Canada and the USA then dielectric strength testing is often a routine production test. The duration of the test aligns with the standard and it is a super quick and easy test to set up in a production line.

 

[Loekje]

Yep, our cDEKRAus report of last month also says we must do 100%, and the 1 second DC test for patient connected circuits is quite some value. Not even sure our tester can do such voltage..
Yellow highlighting by me.


I guess our devices will pass these tests, but how does one proof these production tests do not harm the product?

 

[Peter Selvey]

The 1500V primary to ground for 60s likely to be OK in an assembled device although it does not necessarily mean it is being done correctly. For example, if the device has a mains switch or relays there is no guarantee that voltage actually gets to the parts that should be tested.

The 4000V test is very unlikely to be done correctly on an assembled device. It can both stress parts not designed for 4kV and also not stress the parts intended to provide the 4kV isolation. There is simply no reasonable guarantee that the voltage distribution will follow the isolation plan.

Consider for example the typical design which relies on the power supply for the 4kV isolation:

Mains || 4kV PSU isolation || Secondary || 1.5kV Patient isolation || Patient circuit || Plastic cover (functional) || Applied part (patient, foil)

If you apply 4kV between mains and the applied part (foil), the voltage will distribute between the different insulating components (shown in Bold/Ital) according to the capacitance. Typically due to EMC caps, the PSU isolation will have high capacitance >1000pF, while the patient isolation is typically ~100pF and the plastic cover could be as little as 10pF. That means roughly 99% of the 4kV will occur on the plastic cover, which is only functional and therefore can easily breakdown. If it does breakdown, the patient isolation barrier will be hit next, but it is only designed for 1.5kV, it might pass 4kV but it is not designed for it. At the same time, the PSU isolation which is intended to provide the 4kV doesn't actually get tested.

This issue is actually raised by the standard, but it is a little obscure. First, in Clause 8.8.1:
"Only insulation that is relied upon as a MEANS OF PROTECTION, including REINFORCED INSULATION, shall be subject to testing."

Then in 8.8.3 c):
" If it is not possible to test individual solid insulations, it is then necessary to test a large part of the ME EQUIPMENT or even the whole ME EQUIPMENT. In this case, it is important not to overstress different types and levels of insulation ..."

The upshot is that dielectric strength is intended to be applied to individual solid insulations according to the MOP isolation plan. It is not, and never was, intended for testing the correct assembly of a final product. If the test can be applied to the individual isolating components (e.g. transformer, optocouplers, PCB in the power supply) before being installed, this is preferable to testing on a fully assembled device. If you do test on an assembled device care is required to avoid incorrect voltage distribution and may involve shorting parts and disconnecting parts to ensure the voltage is distributed correctly, something that may be impractical and could create additional risks if done in a production line.

Now, it is possible that DEKRA are just giving the high level requirement from 8.8.3 and leaving it to the manufacturer to decide on the detailed implementation. So for example, if the manufacturer uses a cUL marked power supply and the records show they have 4kV on the relevant parts in their production process, this should be enough. Or you could do the 4kV on the power supply PCB before installing (again, making sure voltage distribution is OK, especially via earth).

However, if DEKRA insist on a 4kV production test on a fully assembled device, it's time to get their accreditation body involved. That's really bad and needs to be stopped.

 

[Loekje]

Hi Peter,
Thanks for jumping in again.

Indeed the 1500V test can be done wrong: at first at my new employer this was specified with the power switch off, in which case only the switch itself was tested (guess what? never failed) now it is performed with the power switch on and the PSU unit is tested as well.

As you can see in the green table the first test at least specifies between which points (Primary to Ground) but the second one lacks this specification.
The insulation MOPPs inside are spread out over multiple DC-DC converters and digital isolators that have certificates themselves (and set in the Critical Component List), these are not really testable before assembly. The datasheets do mention some 100% tests but not up to this level.

I guess I will turn to the test house engineer for some clarification.