Touch current, 240 VA limit, and related matters

[RednBlack]

Hi,

We have a new product being evaluated and this clause has come up in discussions with our NB.

The product is powered from the mains by an off-the-shelf 60601-compliant desktop PSU; the 24V DC output is through a shielded cable and a metal-bodied plug.
The shield and plug body are connected to PE within the PSU. The PSU is Class II, so the connection to PE is purely for EMC reasons and is classed as a functional earth.
The product is in a metal housing. The PSU DC output is floating with respect to PE, but the negative terminal is connected to the housing at the DC inlet.

The 24V DC is distributed throughout the product and several accessories, all connected using the same type of metal-bodied plug. The insulation between the 24V and the housing (PCBs, connector internals etc) is classed as 1 MOOP.

When the touch current is measured in normal conditions, as the housing has a connection to PE, the current is close to zero. Similarly, when the touch current is measured in SFC with the PE connection broken, we get a small AC current from the primary-secondary capacitance, well below the 500 uA limit. So far, so good.

The NB then point us to clause 8.1, which states
" NORMAL CONDITION includes all of the following simultaneously:
[...]
- short circuit of any or all insulation that does not comply with the requirements of 8.8;
– short circuit of any or all CREEPAGE DISTANCES or AIR CLEARANCES that do not comply with
the requirements of 8.9;
– open circuit of any or all earth connections that do not comply with the requirements of
8.6, including any functional earth connection."

So the NB are arguing that the touch current measurement should be done with the 24V shorted to the housing. Fine, I said, in this case the PSU output will be short-circuited and it will just current-limit for the duration of the test. I tried this myself in the lab and got a DC touch current of a few tens of microamps. Similarly, if the functional earth connection in the PSU output cable is broken, there is no return path for the touch current.

The NB are now arguing that 8.1 requires them to find the least favourable combination of shorts / opens in all the functional earth, DC + and DC - connections throughout the system, and measure the touch current in that configuration. Which, with the resulting 24V DC between the housing and earth and a 1K measuring device impedance, it will fail.

We have several similar products, going back many years, all using the same power topology, that have all passed evaluation by the same NB, so this is a bit of a new one on me. They are refusing to be drawn on the previous product evaluations, not surprisingly.

Another similar issue relates to 8.4.2(c) which states

"The limits specified in b) above do not apply to the following parts if the probability of a
connection to a PATIENT, either directly or through the body of the OPERATOR, through which
a current exceeding the allowable TOUCH CURRENT could flow, is negligible in NORMAL USE,
and the instructions for use instruct the OPERATOR not to touch the relevant part and the
PATIENT simultaneously:
– accessible contacts of connectors
[...]
For such parts, the voltage to earth or to other accessible parts shall not exceed 42,4 V
peak a.c. or 60 V d.c. in normal condition or in single fault condition. The d.c. limit of
60 V applies to d.c. with not more than 10 % peak-to-peak ripple. If the ripple exceeds that
amount, the 42,4 V peak limit applies. The energy shall not exceed 240 VA for longer than
60 s"

We do have such a warning in the IFU, as the probability of the user disconnecting the DC plug, poking their finger into it to touch the +ve pin, and simultaneously touching the patient, is indeed considered negligible. But the NB then point us to the 240VA limit. The PSU has a 250W rating, and the max continuous power before it goes into limiting will be somewhat higher than this.

Comments welcome.

 

[Peter Selvey]

OMG you need to find another test house. This is crazy. This opinion is coming from an old guy that started testing back in 1987. There are reasons for all the limits and so on, and if you understand the context and rationale, there's no way you would go to this extreme to try and fail a power supply. I could write a essay on this but the upshot is: Get a new lab. Now. These guys have no perspective.

 

[RednBlack]

OK thanks, my first reaction was this can't possibly be right, good to have it confirmed.

The NB is a major multinational, not some Fred-in-a-shed, and are usually quite reasonable. My approach so far has been 'test it and be damned' but not if they start going way beyond the intent of the standard. The next conf call with them is going to be interesting.

Changing NBs is way above my pay grade, but getting a 3rd party test lab to do just the electrical safety part of the evaluation might be a possibility - assuming we can convince the NB to accept their report, of course.

 

[Peter Selvey]

This is a case of accumulated "what if's".

In safety (and in general for reliable design) it's normal and reasonable to overgame the situation say by a factor of 10 just to be sure. It's not overkill in the sense that there are thousands of decisions to make in a design and if you play the game of running everything just on the edge of OK it'll never work as within those thousands of decisions there's bound to be something stuffed up. So overkill as a general game play is essential to successful designs, both in function and safety.

The thing is in safety, we often have a (long) sequence of events. If we overegg the numbers by 10 for each item in the sequence, it's easy to end up with factors of 1000 out of whack. If there is a sequence, the workable approach is to use reasonable numbers for each item in the sequence, and then throw in a factor of 10 at the end to be sure.

In this case there are the probability of the faults in the 24V circuit, which the test lab is (in effect) assigning P = 1, crazy. Real world figures are <0.001/year for each fault. Then we have the assumption of very low impedance of the operator which is rarely the case, normally the body is >100kΩ at 24V unless there is large contact areas and saline solutions sloping around (at both entry and exit points). Then we have contact impedance of the patient which faces the same issues, assumed to be 1kΩ but in reality will be much higher at 24V. Plus, you need a well timed high current of about 100mAdc to start to have a risk of something bad - the limits in the standard are based on sensitivity, not stopping the heart. Sensitivity is important by the way, for example a surgeon doesn't want to have little electric shocks going on, but it's not heart stopping stuff at 0.5mA. But 24V ... I challenge anyone to sense it without special preparation of the contact point. Oh, and don't get me started on how the cr/cl limits in the standard don't make sense for distance within a circuit (not between isolated circuits). Anyway, it's a big pile of overkill each one of which is OK if it was the one any only line of defence against somebody dying, but not if there is a (much) longer sequence.

 

[RednBlack]

Thanks, that all makes perfect sense, and is presumably part of the rationale for voltages below 40 - 50V, depending on which set of standards you're working with, being deemed 'touch safe'.

The part I'm struggling with is being able to point to a particular clause in the standard which shows the NB their interpretation is wrong. As you hinted, it may be less pain in the long run to find a different NB.

 

[Tidge]

The part I'm struggling with is being able to point to a particular clause in the standard which shows the NB their interpretation is wrong. As you hinted, it may be less pain in the long run to find a different NB.

A possible question for the NRTL, although it may be unwelcome, is to ask which other devices they have tested in this exact way. Part of the ground rules for NRTLs is that they are supposed to treat all clients equally. You won't get to see test reports (without arm twisting) but if you think they are being unreasonable they have an obligation to demonstrate neutrality.

 

[Peter Selvey]

I think Tidge is on the right path. The interpretation of the NB is not something that could be applied consistently. For example, in a switching power supply there is a resistor that senses the current and shuts down the power supply in case of short circuit or overload. According to 8.1a, any distance that doesn't meet 8.8 and 8.9 can be shorted in "normal condition", which means this resistor could be shorted as PSU designers would surely not have the required creepage distance. Which means the power supply is no longer protected against short/overload. Every power supply on the planet would fail with an overzealous interpretation of 8.1a.

It gets even more problematic when you combine 13.2.1 with 8.1a which then allows the "normal condition" faults to apply to a wider range of hazards, not just electric shock.

The intention of 8.1a is in the context of mains isolation where you might have a mix of functional insulation and safety insulation. Clause 8.1a allows the assessment to bypass the functional insulation and focus on the safety insulation. Mains voltages are genuinely dangerous, the voltages are high enough to punch through skin insulation and access the blood circuit which is relatively low impedance (~1kΩ). So it's reasonable to expect manufacturers to have clear, identifiable, testable barriers around mains voltages. Clause 8.1a is there to protect the public against those manufacturers that want to try and sneak in a bit of dodgy functional insulation to get them across the line.

Usually this is no problem as most standards treat voltages below 60Vdc as safe (as long as they are double insulated from mains).

However, in the medical environment, there are special cases where even 1V could be a problem. Rather than identifying these special cases, IEC 60601-1 decided treat all voltages above 0.1V as "live". This then creates a problem when combined with clauses like 8.1a that were written around mains voltages. So we end up with little electronic thermometers running off a 1.5V coin cell being hit with 1000Vdc.

What is clear is that any attempt to use 8.1a to justify insulation for secondary circuits will be inconsistent - even within the same product. There is a risk that pointing this out to a test lab could cause more trouble, like the lab asking for insulation around the PSU current sensing resistor. But that's just one example. Modern PSUs have a bunch of protective features for overvoltage, temperature in addition to overcurrent. All of these could be defeated with 8.1a. And that's just the PSU. Lithium batteries, or the main function of the medical device (accuracy of measurement, control, functional protective features) can all be defeated with 8.1a. So eventually the lab should understand that they can't selectively apply 8.1a just to this case.

I wonder if the gods working on the 4th edition are going to fix this?