Interesting find - I suspect it passes  the standard test, because the test method does not consider a fault may sneak up slowly.  (and 10uA/second, suggests ramping  up to the 6mA in an 10 minutes or more )

However, it does suggest that a level of DC may be not be detected that has the potential to blind upstream RCD protection, if that upstream RCD is particularly vulnerable.

In practice a fault that comes on that slowly and smoothly is very unlikely (real faults tend to come on with all sorts of current steps discontinuities as things break down) and the risks from using the device will be small.

But it suggests the DC sensing method has some baseline drift compensation that is struggling, and is in effect the DC fault detection is very slowly AC coupled. I presume this is the result on a sample of one, and it would be interesting to know if that means others from the same design will have faster or slower rates of base-line drift, as if some are much faster then it could result in faults being missed.

I think the answer to the exam question is 'probably, but not too sure'

Mike.

There are other issues that cause it not to trip - I think as long as there is less than 6mA of change within 30 seconds the product will not trip (it doesn't' need to be a gradual ramp). I am worried that this will cause upstream RCD binding and I was hoping that there would be something in the regs that cover this situation. I've not yet carried out a long term test and my fear is that over the course of a couple of hours this current could continue to increase to even higher levels. IEC 62955 does not have any long term or absolute test thresholds (only step changes and ramp from <2A to 6A in <30s

I know the very slowly appearing fault is not tested, but part of me does wonder  if it should be.

Equally 'zeroing' a DC current measurement from clamp on magnetic devices is pretty much essential, as anyone with a DC clamp meter will attest, partly due to the earth's magnetic field and also to slow release of deep level states (late relaxing hysteresis)  in core materials. (whee it was magnetized at some point, and then mostly un-magnetizes very quickly, but a few 'sticky' magnetic domains stay aligned  after the stimulus has gone, and flip state on a scale of  hours or even days later)

This does mean that some base-line drift is inevitable, and a current that rises slowly compared to that correcting interval (time) will be mistaken for drift and nulled out.

Flux gate sensors are usually better than hall effect in this regard, as the core is being periodically de-gaussed but they need more circuitry and power and are not as easily miniaturized. I'm not sure what sensor is in these things.

Given the price of some makers options, one feels it could almost be a squid, but I doubt it

Mike.

This product uses a novel technique (i.e. cheaper) to measure the DC current using an off the shelf current transformer. My testing shows that it can't reliably measure DC. It has been design to pass IEC62955 which it does so as far as the manufacturer is concerned it is safe to use in the UK. I just don't want to see them installed without an actual DC 6mA RCD 

This product uses a novel technique (i.e. cheaper) to measure the DC current using an off the shelf current transformer. My testing shows that it can't reliably measure DC. It has been design to pass IEC62955 which it does so as far as the manufacturer is concerned it is safe to use in the UK. I just don't want to see them installed without an actual DC 6mA RCD 

This is very interesting, as it surely cannot be just extending the bandwidth down to 10mHz or whatever it needs to see a ramp or triangle that slow as a normal transformer coupling.Do you know what the novelty in the technique is  by any chance ?

The only way I can imagine is a variation on flux gating, and I'd have thought the extra electronics and the EMI the driving waveform inevitably injects would be prohibitive - given that Hall effect devices are cheap enough to build two into every brushless computer fan.

That said I see that it is being done by at least one maker.

https://www.magnetec.de/storage/2022/07/mag_Broschuere_Fluxgate-BasedResidualCurrentSensors_web.pdf

albeit with a more carefully controlled magnetic material and a better fixed winding geometry than  a plain transformer. (though that may also be to aid automated PCB population as well.)

Mike.

PS if it is flux gating, its not all that novel as an idea - in the late 1950s oil survey companies were using them to spot local variations in the earths magnetic field on the surface to map out the geology beneath.

This is from the Chinese patent application:

Available here:

patentimages.storage.googleapis.com/.../CN207408789U.pdf