Terminator networks for VFD 3

[xj25]

Hi,

We have a problem of voltage spikes in a VFD installation (220V) with 7 induction motor fans in parallel (about 8 amps in total).

We have measured spikes in PWM pulses (50ns rise time) about 800V peak in the farthest motor from VFD.

We are assesing what to do (add reactors, dV/dt filters etc. etc.) and I found a "termination network" discussion about using RC network between phases matching the cable impedance like this:

R about 100ohms
C about some tenths of nF

Supposedly, at the pulse rise the R matches the cable impedance so spike is "cancelled", after that C charges to line value so no permanent power is wasted.

Has any sense for you guys? any experience or supplier with this kind of terminators?
I have thought that might try with three 100ohms 47nF snubber between phases to see what happens...

 

[Skogsgurra]

Do you really have 50 ns risetime at the motor?

That is quite fast even directly at the inverter terminals and dispersion usually makes that risetime hundreds of nanoseconds at the end of the line.

It is not very common to have any problems with voltage spikes at 220 V. Not even at 400 V. Problems with insulation usually starts at 500 V and is very common at 690 V.

Are you sure about the probe compensation? It is next to impossible to have 800 V spikes in a 220 V installation.

Anyhow, the standard remedy is to use a damped motor reactor or a du/dt filter. Try a 2 % reactor first and add a parallel resistor to fight ringing and hot core. I think that three resistors with 50 - 100 ohms and 20 - 50 W will be OK in an installation like yours.

Snubbers are not used very often to reduce reflected waves/overvoltage at the motor end.

Gunnar Englund

http://www.gke.org

--------------------------------------
Half full - Half empty? I don't mind. It's what in it that counts.

 

[DickDV]

Allen Bradley has promoted their "motor terminal protector" for many years but, while it may protect the motor adequately, it does nothing to quiet the motor leads.

As skogs says, a reactor or dv/dt filter mounted as close to the drive output terminals as possible is preferred.

 

[electricpete]

I would go with what those guy say, I'm sure they have a lot of knowledge on this stuff.

I have never played with vfd's, but it sounds like an interesting phenomenon. I have some questions out of curiosity.

You don't have 6 identical-length cables and the 7th one longer, do you?
What are the lengths?
What do you see at the others motors? (I assume you already measured them since you said this is the highest).
Are these cables shielded? (I don't normally see shielded lv cables, but I don't know about vfd).


=====================================
(2B)+(2B)' ?

 

[ozmosis]

xj25
when you take into consideration all motor leads in parallel, what is the total motor cable length?

 

[electricpete]

I did a quick simple simulation, which is not to be taken too seriously since the input parameters were fabricated, but it illustrates a mechanism where you can get unusualy high amplification when 6 cables are approximately the same length and the 7th is longer.

Slide 1 shows the system model:
[tt]Source ===SourceCable===== Bus===Cable1===Motor1
===Cable2===Moto2
etc
===Cable7===Motor7[/tt]

The source is a ramp increase from 0 to 1 with rise time of 50 nsec

All cables are selected to have 100 ohms characteristic impedance.

The source cable is selected very long so that the source does not produce any reflections during the transient. That simplifies the simulation.... the actual source characteristics certainly could modify these results.

Cables 1 through 6 are selected to have time delay of 201nsec through 206 nsec, which corresponds roughly to 201 through 206 feet.

Cable 7 is selected to have time delay of 400 nsec, corresponding to 400 feet.

All motors are 800 ohms (intended to simulate motor impedance much higher than cable).

The results are on slide 2.

When the pulse hits the motor, the reflection coefficient is (800-100)/(800+100) =0.778. This reflection adds with incoming wave to give magnitude 1.778, which matches the plot.

The reflections of the 1st 6 all reach back to the bus at the same time (400 to 450 seconds). By looking at the response of the 7th line after it's time delay elapses, we can say roughly that the reflection from the 1st 6 upon hitting the bus tend to travel toward the 7th motor. It is as if the 6 motor cables act together like one cable with lower impedance 100/6.... when they get back to the bus they have nowhere to go but the higher impedance cable 6 (100ohms), so there is amplification in this path.

In the end, we get close to 3 times the original pulse showing up at motor 7, even though it has the same cable impedance and motor impedance as all the others (just a longer cable). Normally we think of 2 as the highest, but this involves multiple reflections.

Again, it is just an exercise in applying textbook transmission line theory, but would require a lot of further study and discussion to get any feeling for the extent that it would reflect a given real installation.

=====================================
(2B)+(2B)' ?

 

[electricpete]

It is as if the 6 motor cables act together like one cable with lower impedance 100/6.... when they get back to the bus they have nowhere to go but the higher impedance cable 6 (100ohms), so there is amplification in this path.

"nowhere to go" was a poor choice of words, since a portion will be reflected back along the original 6. But the important point is that it resembles a pulse traveling on a single cable of impedance 100/6, transitioning to a cable of 100, and we can calculate the transmission factor T=2*100/(100+100/6)=1.71.... and indeed we see that the increase which finally occurs at motor 7 (from 1.78 to 2.9=1.1) is prety close to the transmission ratio 1.78 times the first reflected portion 0.78. 1.71*0.78=1.3. Actually it doesn't work out exactly (the increase of 1.1 does not match calculated 1.3)... anyone want to sharpen their pencil and figure out why? I'm not sure...I must have made an error somewhere. I'm pretty sure the simulation is right and my handcalcs got astray somewhere.

Another note - I did not model the trailing/decreasing edge of the pulse. If pulse duration were shorter than about 500nsec, then this model would need some changing.... the original incident pulse on motor 7 would stop dropping before the reflected pulses from the others reached it.

=====================================
(2B)+(2B)' ?

 

[electricpete]

I probably should mention something else that you guys probably already noticed.... this peak of 3*input occurs in 2 steps, rather than a single rise time. Probably has nothing to do with what you're seeing if the entire rise you see occurs in 50nsec.

If anyone cares, I resolved my hand calc error. One thing I forgot was the source cable attached. It has source impedance 100/7=14.87, selected so that the source pulse passes to the 7 output lines without any change in magnitude. So when that pulse reflects back along the 6 lines, it transitions not to the single line of motor 7, but to the parallel combination of motor 7 cable and the source cable (labeled T7_Para_Tsource below).

Below is sequence of signals parameters that matches the observed increase of 1.18 (2.96-1.78) at motor 7 when the reflection from the other 6 arrives there:

Parameter Value Formula
ReflectionAtMotor 0.777777778 =(800-100)/(800+100)
RefPlusTransAtMotor 1.777777778 =ReflectionAtMotor+1
T7_Para_Tsource 12.50098436 =100*14.287/(100+14.287)
TransmCoefAtBus 0.857181427 =2*T7_Para_Tsource/(T7_Para_Tsource+100/6)
TransmittedPastBus 0.666696665 =TransmCoefAtBus*ReflectionAtMotor
Increase 1.185238516 =TransmittedPastBus*RefPlusTransAtMotor

The 1.18 calculated increase matches the program's displayed 1.18 increase.

I feel better now.

=====================================
(2B)+(2B)' ?

 

[xj25]

Hi, thanks for the replies,

diagram (see pdf attached for actual waveforms)
VFD--MCBs for each fan------(fan1)
---------------(fan2)
------------------------(fan ...)
----------------------------------(fan 7)

length to fan7 is about 20m. The remaining 6 proportional.

Pulse rise time was wrong, is 100ns. See waveforms.

Probe is differential I only have the measurements for fan7. I will check compensation (Actually I don´t remember to see the compensation screw, it is a Chauvin Arnoux diff. probe).

DickV "...motor terminal protector... while it may protect the motor adequately, it does nothing to quiet the motor leads."
Sorry for sure I´m uninformed but I think that I don´t understand this statment, then, what is not protected with terminal network?


Regards!

 

[Skogsgurra]

Thanks for the recordings. If your probe is a C&A DP-25, there is nocompensation. So nothing to adjust. Measurement should be OK.

The recordings shown are not only reflected waves. Looks more like ringing and that, in combination with reflected wave, can explain the unusually high voltage. Pure reflection cannot get higher than twice the DC link voltage, 1.8 is a common value, and in this recording, the factor is 2.5

There are MCCs in the motor paths. They may contribute if the coils are inductive with very little R.

Dick probably means to say that emission from your cables (noisy cables) will not be affected by RC termination. There will be some reduction, but not much.

How is the 800 V affecting the motors? Do you have premature isolation failures? Or bearing problems?

I have recordings from motors with pure reflection. Shall dig and look for them.

Gunnar Englund

http://www.gke.org

--------------------------------------
Half full - Half empty? I don't mind. It's what in it that counts.

 

[electricpete]

fwiw - I tend to agree with Gunnar (that's usually a good bet) that exceeding 2 is unusual and sugests something more than simple reflection. ie we don't expect a simple single reflection to exceed 2, so multiple transitions/reflections are needed... but it's hard to imagine how they would be coordinated to arrive at the same time.

Attached are two more simulations.

1st simulation as shown on slide 1 is same as previous except:
* I have put in the actual cable lengths
* I got rid of all the other motors except #8.
* I set the source cable length to 1nsec and got rid of the delay... now the source acts to have zero source impedance. That may not be realistic, but if I put in a source impedance, it will tend to drag the final voltage below 1. Have to think about that some more.

Results are on slide 2. What is seen on motor 7 is a peak of 1.78 which we calculated before as 1+ (800-100)/(800+100). The period is about 4 travel times ... 4 * 60 = 240. This makes sense since the pulse will bounce off motor (OC) with same polarity, bounce off bus (SC due to source model) with opposite polarity, return to motor with opposite polarity have completed ½ cycle in 2 travel times.

2nd simulation is on slide 3. Here we have simply added in the other motors.
Results are on slide 4. Addition of the other motors caused motor 7 peak to 1.85, which exceeds previous calc’d values, but still not more than 2. Period still 240 nsec which does not match measured 1mhz ó 1000nsec period. If we wanted to describe this result in qualitative terms, we could say the interaction between motors is minimal, the longer cables tend to have higher peak because the the cable is approaching closer to the half the length of the pulse, where the theoretical reflection coefficient (up to doubling) applies. There is some interaction as evidenced by the fact that motor 7 is slightly higher with others present, but it does not seem to be a big effect in this particular model.

None of this proves anything related to your motor. It is just a wander through the world of transmission line behavior and still a long way away from recreating your results or having confidence in any model (I doubt we will ever get there).

I’m going to think some more about the source model. If anyone has suggestions on that or other aspects of the model, feel free to chime in.
Can you describe the physical configuration of the cables.. what type of cable and is it shielded and which end.

=====================================
(2B)+(2B)' ?

 

[electricpete]

clarification in bold:
what type of cable and is it shielded and which end.
should've been
what type of cable and is it shielded and which end of shield is grounded.

=====================================
(2B)+(2B)' ?

 

[electricpete]

What is distance between the vfd and the MCBs?

=====================================
(2B)+(2B)' ?

 

[electricpete]

fwiw - I tend to agree with Gunnar

Sorry, Gunnar had a different discussion than mine.

=====================================
(2B)+(2B)' ?

 

[Skogsgurra]

Found these waveforms in a report from comparison between an ABB ACS 800 and a Siemens Sinamics VFD. The ACS 800 has no filters and the Sinamics has an integral sine filter.



Gunnar Englund

http://www.gke.org

--------------------------------------
Half full - Half empty? I don't mind. It's what in it that counts.

 

[xj25]

I will check, but think that are used separate cables for each phase, not screened. At least at the end connection where I took the measurement was like this, so very probably is the same all the way from VFD.

The MCBs are just near VFD in the same cabinet (0,5m?).

It is a new installation so hasn´t been detected any problem (yet).

Fans are EBM-PAPST and we have been told (before making the voltage measurement) by EBM representative that should be installed a filter like schematic adjoint. He refers to that should be verified that rated motor temperature is not overcome due to VFD duty, because it could cause insulation problems (maybe he is mixing several concepts here).

The filter is a kind of low pass plus isolation transformer. Do you think that really is needed something like that?. I don´t see the point of the isolation transformer (unless it helps with bearing currents?).

How long (usually) insulation fails are noticed with this kind of spikes?

Regards!

 

[Skogsgurra]

That's a full-blown sine filter with a step up transformer!

Very unusual. And very expensive. ABB have delivered a few of them for large machines. Never seen that used on 1 kW machines.

I think you should reconsider the whole thing. Do you really need that kind of fan? Is this a very special installation? Like clean room for semiconductors or a medical lab?

Gunnar Englund

http://www.gke.org

--------------------------------------
Half full - Half empty? I don't mind. It's what in it that counts.

 

[electricpete]

I don´t see the point of the isolation transformer (unless it helps with bearing currents?).

Again, I’m not a vfd guy. Throwing in my 2 cents and hoping to learn. Anyone correct me if I’m wrong.
It seems the purpose of the isolation transformer would be
1 - limiting high dv/dt pulses at the motor;
2 - limiting common mode voltlage so as to limit bearing current.

Neither of these goals seem to have been accomplished.
#1 – we can see in your graphs that we have very high magnitude and high frequency content of thes pulses.

I would say #2 is not accomplished either. Here is my simple understanding of common mode voltage: we switch a given phase between + and -. The + and – are voltages with respect to ground since they derive (via rectifier) from input power which is grounded. Since we have 3 phase system, we always have 2 on + and 1 on – or 1 on + and 2 on -….. either case represents common mode voltage that can encourage zero-sequence voltage to ground. The fact that it has such high frequency means it can flow easily through capacitances to make it’s way from stator to rotor. Once in rotor, it can simply complete the path to ground through the bearings.

Since these same pulses show up so distinctly on the output and at the motor, I would say it seems like the zero sequence cannot have been reduced at all (and therefore you lose nothing in terms of bearing protection to remove it).

BUT, notice we have a ground on the secondary of this transformer. That certainly makes it easy to complete the zero sequence path. A question for the gurus: what happens if we remove that ground on output of the transformer?…. wouldn’t that be better for the bearings?

======New subject:====

Looking back to your attached waveform, pdf page 6/7.
The time to rise 800 volts is 0.4 usec = 400 nanoseconds.
It almost looks as if there are two regions of the curve:
First region rises 0 to 400 volts in 100 nanoseconds
Second region rises 400 to 800 volts in 300 nanoseconds.

It is kind of bizarre to start with 100 nanosecond rise time and end up with 400 nanosecond rise time. Just brainstorming why that may be going on to cause that:
1 – multiple pulses (some possibly reflected) superimposed
2 – motor is acting something like a capacitor (surge capcitor spreads out the pulse).
3 – dispersion due to the individual frequencies seeing different impedances and speeds
Beats me… just thinking. Any other explanations?

======New subject:====
Note in my previous modeling, I had assumed the wave was traveling at the speed of light in a vacuum, which resulted in my calculated 240 nanosecond period of ringing from the source (by the way I don’t think viewing source as short circuit is too unrealistic with reflect to reflections from the source because the multiple lines tied together at that location in parallel result in very low impedance). But actually, the speed would likely differ from speed in a vacuum by 1/sqrt(EpsilonRelative), where EpsilonRelative reflects not just the insulation but an average of everything between conductor and ground (or conductor and other conductor). If I pick the highest conceivable number EpsilonRelative =4 (probably unrealistically high for the average), even then we end up with speed of 1 / 2 speed in vacuum, and the period would double to 480 nanoseconds…. still a long way from the 1000 nanosecond period of your measured ringing.

The source of the ringing is a curiosity to me. It is common in measurements Gunnar and others have posted (although not necessarily with overshoot exceeding 2). I have heard it referred to on the forum as LC ringing. It may be the case, but I can tell you in large volumes of surge study by EPRI, I don’t recall them every referring to anything like that. The model of the motor for purposes of predicting voltages within the system up to the terminals of the motor is an R / C circuit (See EPRI 5862 Volume 2, Section 5), and an R/C of termination of a transmission line does not create ringing.

Probably there is something I am missing. I have a completely open mind to the cause of the ringing, because nothing I know of explains it. I would be curious to know what is the frequency of the ringing occurs on motor # 3 which has about half the cable distance. If it is same frequency, then I am leaning toward this “L-C ringing” even though I don’t understand how that applies in this context. If the ring frequency is factor of 2 higher, then we expect something to do with reflections from the source (although I can’t reconcile the frequency, and I’m not sure why we wouldn’t see evidence of it on the source voltage trace).

Another question: the vfd output voltage traces you took are on the output of that transformer?


=====================================
(2B)+(2B)' ?

 

[ozmosis]

My experience with EBM-Papst motors in this scenario is that a sinewave filter will be very necessary. Check with EBM but typically the design of the fan/motor is usually of an 'external rotor' construction and the insulation class of this motor is typically very poor, hence the need for sine-filters between the VFD and motor(s).

 

[Skogsgurra]

Pete: The waveforms were taken without filter and transformer.

The transformer is usually there to step up voltage after the sine filter, where you lose so much voltage that the motor doesn't work well. ABB has (as always) made a virtue out of necessity and given the combination a name: 'Step Up Filter'

Patrick: I see, I have one of those in my ceiling. I haven't tried with a VFD. And will not. These motors usually have a high rotor resistance and can be easily controlled with a triac. Cheaper and easier than a VFD. And better for the windings.

xj: Was the VFD recommended by the fan/motor vendor?

Gunnar Englund

http://www.gke.org

--------------------------------------
Half full - Half empty? I don't mind. It's what in it that counts.