AC leakage currents

Discussion in 'General Electronics Chat' started by Skeebopstop, Oct 30, 2009.

  1. Skeebopstop

    Thread Starter Active Member

    Jan 9, 2009
    358
    3
    Dear all,

    I am working with a servo drive which has reasonably high leakage currents at high frequencies (due to inductive/capacitive coupling of high switching currents).

    When done with a true RMS multimeter the currents are within limit ( < 4mA) when earth bonding is secured.

    When we remove the earth bonding link, the chassis floats to 110VAC which is to be expected when operating a motor switching energy into and out of a DC Bus which sits at +-150V relative to EARTH (i.e. chassis). So when the EARTH is disconnected, this couples a waveform onto the chassis with no path to ground.

    We have had an incident where someone got a tingle, definetely not electrocution but we believe their earth bonding was unacceptable and thus this 110VAC found its way onto the chassis and the leakage currents enough to give them a nice fright.

    The question boils down to this however. Under the scope, the most of the leakage is actually represented in ~100mA leakage spikes at the switching frequency of our inverter stage. One might expect this as it will be coupled through the motor cable shield etc.. back into chassis earth, so when there is no earth bond, I believe this is what they are feeling.

    All the standards I have access too typically only refer to < 4mA a.c. with no mention of spectral content for compliance (without requiring further protective measures). So when I do the RMS of the currents I see, the value is quite low because there is very little low frequency content and only short pulses of high frequency content.

    Does anyone know any governing standards on leakage coupling or may perhaps at least advise me on some typical figures? I would say the large component of this leakage is around 1MHz @ 100mA a.c. pulsed every 50uS (i.e. 20kHz).

    Regards,

    James
     
  2. someonesdad

    Senior Member

    Jul 7, 2009
    1,585
    141
    James: I'm not an EE nor do I consider myself knowledgeable about AC leakage stuff, but I've read a little about it and I'm slowly working on a white paper about it. So take the following information with a grain of salt...

    I haven't seen a copy of it, but one relevant standard might be IEC60601 (I know it's for medical equipment and may not apply to your equipment, but it's likely to contain lots of information and there's a large body of information concerning it on the web). If I was in a commercial environment, I'd want to get my hands on a copy (I think there's a similar ANSI document for folks in the US). Here's a link that might provide useful information. You can also peruse wikipedia for other relevant IEC standards.

    I'd recommend making your measurements with a wideband RMS meter, such as the HP 3400A/B, the Fluke 8920, or the Ballantine Laboratories' 323. I'm not sure I would trust a digital multimeter, even if it was labeled as a true RMS meter. But I'd trust the scope measurements, assuming proper use of probes, etc.

    The human body apparently has a mean threshold of sensation of current of 1 mA at 60 Hz. This, in turn, causes manufacturers to typically set a 60 Hz AC leakage current spec of 500 μA or less; some manufacturers set smaller specs to be more conservative. This sensation threshold is used because even though the shock may not be electrically hazardous, it can cause a person to react violently, causing a secondary accident. Imagine a person jerking his arm back after touching an appliance and e.g. cutting himself on a sharp sheet metal edge. The spec makes sense.

    This sensitivity falls off with increasing frequency. To simulate this fall-off, one recommended probe for measuring leakage currents that I've seen is a 0.15 μF cap in parallel with a 1.5 kΩ resistor (I believe this came from a Simpson manual); a voltmeter reads the voltage across the resistor. Since the reactance of the cap at 1 MHz is about 1 Ω, you can see most of the 1 MHz energy is being shunted to ground in the measurement. The link above uses a probe with a cap an order of magnitude smaller, but this is probably because it's for use in the context of medical devices.

    You might want to make your measurements with one or both of those modifications to your probing.

    My concern from what you've written is that you said a person got a tingle from your device. You're attributing this to the switching transients -- but based on your scope measurements, you should be able to calculate the RMS value of this high frequency current. Does it explain the tingle the person got, even after allowing for, say, an order of magnitude or two drop in sensitivity? My engineering intuition is suspicious.

    And are you testing with the unit the person got the tingle from? The reason I'm questioning is because if you're not, the unit that gave the person the shock might have some other fault that you're unaware of. And this would be quite bad if it was a manufacturing or design fault -- it could be waiting to bite someone else down the road with more serious consequences.

    This AC leakage testing is done with the earth grounding conductor disconnected. From what you've said, it sounds like your device will indeed cause leakage currents due to some capacitive coupling, based on your measurements. I assume the business problem facing you is trying to decide if you need to take action -- and the resulting engineering effort and cost. I can't help you with that, but I can give some personal experience: 25 years ago I was an engineering manager at HP and my group was responsible for these type of safety measurements on our division's products. If our division was trying to field a product that had AC leakage characteristics like you're discussing, it would have received intense scrutiny from lots of folks, including the division manager (who was a competent EE) and the corporate safety people would have gotten involved. My guess is that it would have been turned back to the lab for redesign.
     
Loading...