To those devoted to VFD:
I am looking for any document that says something about acceptable values of high frequency common mode current spikes flowing in motor cable between a frequency converter and an induction motor and leaking towards earth through the stator winding insulation capacitance.
My question refers to the capability of a (low voltage) stator winding insulation to cope with such a di-electric stress.
Motor manufacturers appear very reluctant to reveal there know how or may not even have any.
Most info I have seen so far deals with acceptable voltage spikes, not current spikes.
Maybe in other words the question is: what is maximum acceptable stator winding capacitance value (minimum required high frequency common mode impedance of motor) versus power rating, no. of poles, star or delta, etc.?

Regards,
Arie Mol, NL

Arie,
i have a document that recommends no more than 0.35A/mm. The document does not talk about the capacitance. I know of a few cases that the current was measured, mitigation measures developed and implemented to decrease this current and problem went away. Now, I also know that in some cases the current was higher and there was no damage to the bearings nor insulation. This topic is a sort of taboo to motor manufacturers and you know perfectly why.

Not sure if this helps you.

I have never taken any measurements like you are talking about, but here is my take fwiw:

It is more important to characterize the voltage spikes (magnitude, rate of increase, and repetition) than the current spikes. If you ask someone to surge test a motor... you tell them what voltage, not what current. That voltage (along with rise time) tells us how severe the surge will be for the motor. Also as you know, there is a lot of literature characterizing the surge environment of motors based on voltage spikes (see www.irispower.com)

But the current measurement is of course more accessible. Which is probably why you are asking the question. I assume you have unshielded low voltage cable so you are putting a clamp-on around all three phases to measure the high frequency common mode current spikes.

So the real question I think is this: what do we do with that current spike information to "estimate" the corresponding voltage spike at that location.

I think the answer lies in cable surge impedance. Surge impedance is to traveling waves as regular impedance is to regular circuit analysis... it is the thing you multiply by current to determine voltage. Cable surge impedance is well tabulated and accessible for various configuration because this info is needed when selecting surge protection. I will look to see if I have any info handy... in the mean time any info about the cable would be helpful... size, 3 conductors in one jacket or three separate conductors.

But I think there is one big problem. The very high frequency current content will be hard to measure accurately without attenuation. So if you multiply your measured current times surge impedance you might end up with a number a lot lower than actual voltage spike magnitude due to the measurement filtering effects. I would imagine that with a study of the measurement setup and the characteristics of the spikes, someone could develop a correction factor to apply to the current peak and rate to account for measurement attenuation, but it might not be a trivial task to do it accurately.

This message has been edited. Last edited by: electricpete,

You had previously posted current and voltage waveforms here:
http://maintenanceforums.com/e...912/m/5531041793/p/3

There was a question why the ringdown frequency of the current didn't match the ringdown frequency of the voltage. It didn't make much sense at the time. Now I'm thinking maybe it is a limitation of the current measurement. The inductance limits the high frequency repsonse even in an air-core coil. I'm thinking what we saw in that indicated current signal is an initial indicated spike corresponding to a high actual di/dt, followed by an indicated ringdown which is more a characteristic of the measurement circuit than it is of the actual ringdown current flowing in the cable (sort of like seeing an impulse/impact test of the measurement circuit which excites the natural frequency of the measurement circuit).

Do you have specifications for your Rogowski coil in terms of frequency response?

Electricpete,

There are two separate ‘phenomena” – the voltage spikes and the current spikes. Though they are related. The voltage spikes are very well known and as you pointed out they were responsible for damaging motor insulation. Hence the changes in the motor insulation. I do not think this is an issue anymore.
However, with the introduction of IGBT based drives we discovered new problems – bearing damaged due to current spikes. These are caused by high common mode current spikes and incorrect installation practices. As you pointed out with high frequency we need to look at cables in a different way. How to measure the common mode current? An easy task. By measuring current in all 3 conductors using high frequency current clamp. However, what is more important is what happens to this current. I mean which paths it “chooses” to come back to the inverter? And come back it must. There are ways of measuring where the current goes and making sure it does come back through cable not the bearings.
Various capacitances inside the motor do play the role though from practical point of view they are not important as calculations would be very complex. Hence it is easier to measure common mode currents, shaft currents and a few others. If there are no bearing damage – no problem.

To answer Arie question – there are no known standards, at least I never seen one. However, the spikes amplitude is not that important as the current should come back through the neutral conductors and the cable shield. How to do it? High frequency grounding, shielded cables, filters increasing cables high frequency impedance, active filters etc. It all depend on the severity of the problem and the “pocket” of the owner……

quote:

Do you have specifications for your Rogowski coil in terms of frequency response?

I will have to check but if I am not mistaken the Rogowski coils we used had range up to 10 MHz.

Pete

I just checked it:
Frequency range: 0.8Hz - 10MHz.
di/dt - 20kA/microsec

Hope it helps....

This message has been edited. Last edited by: Kris,

Hi Kris

Arie has clarified in his original post that his interest was stator winding insulation, and that is the context of my comments, although I'm making some assumptions about what he's really trying to get at. Those are interesting comments about bearing monitoring as well.

10mhz is higher than I expected. 10 Mhz range on the coil would be higher than the oscillation frequencies and would not be consistent with my discussion above. Although it may not be high enough to capture the peaks. I would still be very interested to know the specifications for the particular coil that Arie used to collect that data posted in the other thread.

========

Arie,

What I have said above I believe is a correct starting point for analysing the problem under the assumption that the signals act as travelling waves. I didn't talk about conditions for that assumption and I should have because it may not apply here. It would be the case for longer cable runs on the order of one quarter wavelength or more (at first glance I think one quarter wavelength would be around 25% * 3E8 meterpersecond / 1Mhz = 75 m, but the frequency content of that step increase may be quite a bit higher than 1 Mhz and the wavespeed in the cable may be slightly slower than in a vacuum). If the cable is long, thent case they should be related by the cable surge impedance. What is downstream (for example motor capacitance) would not enter into the calculation for converting between current and voltage (except due to reflections) in this case.

For a shorter cable, we could not rely on the surge impedance and conventional circuit analysis applies and that may well be the case here. The relationship between current and voltage in an R / L / C circuit with very little info on the value of the parameters is tougher to come up with.

How long was the cable for the data posted in the other thread ?

This message has been edited. Last edited by: electricpete,

Pete

I have re-read Arie post and you are right, his concern is about insulation. So your info on measuring voltage spikes, spikes propagations DOES APPLY.
Cable lengths - they do matter. Most of the VFD manufacturers will specify the maximum length.
Rogowski coil - 10 MHz is sufficient to measure the voltage spikes.

Thanks for the replies!
Some replies from my end first:
* Indeed my posting is related to the one mentioned in 06 Jan 2009 posting. Same application.
* Cable length involved is less than 10 meters, normal three phase shielded symmetrical cable, properly connected for VFD application.
* The Rogowsky coil I used is a LEM RR3030. Bandwide is 50 khz @ - 3 dB, but this coil is able to capture much higher frequencies. Even home made coils can do this job. The crucial element however is the integrator, not easy to produce in a home made version.

The idea behind my posting is this: Common mode voltage spikes will always result in common mode current spikes. It is the severity of current spikes that may possibly harm the insulation. Now the big luck here is that current spikes are easier to measure. More safe also, just clamp-on. Therefore it would be of more benefit to have acceptance limits based on current. That is why I am interested in estimating stator winding leakage capacitance levels.
But this subject is still an open book to me, even to OEM's. Not until I have gathered more information about this subject by doing measurements wherever I can, some more light will be shed. So patience is the name of the game, to be continued.
Arie Mol
Now it is skating time over here!

In addition, just a thought fwiw (VFD only).
I think there could be a relation between stator winding insulation integrity level and circulating bearing currents. For identical motors: if common mode high frequency insulation leakage current is higher for one motor then the circulating bearing current will be lower (because low stator impedance path shortcircuits the rotor impedance path) hence reduced change for bearing damage.
Oops: reduced insulation integrity saves bearings!
IMO, that is why so many identical machines behave non-identical. It is because insulation parameters may vary so widely for 'identical' machines.
Regards,
Arie Mol

The cable is short enough that surge impedance of the cable is not relevant. For all practical purposes we can simplify the circuit and consider this short cable as an ideal conductor. BUT within the motor itself, the speed of the wave is dramatically slowed by the magnetic core and the wavelength increases and wave behavior dominates within the motor (even though it does'n't within the cable). Which is to say, we can't model the motor as a simple capacitance. Think about it... the surge voltage is mostly seen at the first coil, and the full voltage never hits the neutral of the winding, so how would it be relevant to use a ground capacitance measurement which equally weights all turns including those near the neutral? (it wouldn't imo).

If we wanted to use a circuit model, it would look something like the Wagner diagram as Jan suggested in the other thread. But this is really just a means to use circuit elements to represent wave behavior.

The wave behavior of motors can also be represented by the motor surge impedance. The surge impedance of motors is tabulated for purposes of selecting surge protection. Attached is an excerpt from Industrial Power Engineering and Applications Handbook with three different figures providing motor surge impedance vs motor voltage and rating. This seems to provide a pretty good basis for relating the current spike magnitudes to the voltage spike magnitudes when we have a short supply cable. (anyone agree or disagree?).

Again this presumes we know the current spike magnitude... and the current probe frequency response will be something to consider.

Your probe had a 3db point (70.7% of signal passed) of 50khz. I think this does answer the question in the other thread why you didn't see the same frequency current oscillations as the voltage oscillations (600khz). There were probably actual current oscillations at 600khz in the system but they were so attenuated by the probe that you couldn't see them on the output.

The bandwidth limitations of the probe will also limit your ability to see the true current peaks. A higher frequency probe like Kris' would get you closer to the true peak. Again, you might be able to estimate how much the peak would be attenuated with knowledge of the probe frequency response characteristics and actual current spike characteristics or perhaps some calibration experiments such as the data you posted in the other thread showing both current peak and voltage peak... but the height of current peaks and voltage peaks don't seem particularly correlated there... and the behavior does resemble simple capacitance in some respects. I guess we need to study that a little more to understand the pattern.

This message has been edited. Last edited by: electricpete,

MotorSurgeImpedance.ppt (328 KB, 13 downloads)