VFD motor feedback?

[SiriusC1024]

I've been working on a problem with the VFD/motor combinations used on extruders at our plant. These are 300hp-400hp Baldor RPMAC series motors driven by various VFD's. Whether the drive is a Toshiba, Yaskawa, or Allen Bradley doesn't seem to have an effect. The common cause is the motors themselves. Product literature shows that these RPMAC motors are extremely power dense, and I'm wondering if there are special concerns when it comes to them. These are not the permanent magnet rotor type.

VFD's are set to constant V/Hz mode.

I've identified that the motors are running at high temperatures due to stator saturation. In fact, measuring voltage at the motor terminals using a Fluke 123b oscilloscope shows that the V/hz ratio is in the range of 9V/Hz. The motor nameplate calls for 460V@60Hz, or 7.667V/Hz.

The trick I've done so far is changing the drive parameters to be 460V@66Hz to bring the voltage down to the correct levels at the motor terminals. It's proven effective. For example, a 400hp motor was shown to have a 23kW reduction in power consumption, which was almost entirely iron losses. Basically, that amount of heating is equivalent to 23 space heaters set on high. Temperatures dropped from 220F to 185F.

Still, I'm uncomfortable with this hack. I want to know why this is happening. Looking at a pulse from the drive, the reason for the high V/Hz ratio can be seen. The pulse isn't a clean square wave. There are large overshoots, and the area under them is what's boosting the voltage reading. Now what is going on here? Impedance mismatch would explain the rapid oscillations, but the large-area overshoots?

I have an image of the pulse, but I can't figure out how to upload it. https://photos.app.goo.gl/xcCNFDBHhwm3tvU16

 

[synchro]

Looking at a pulse from the drive, the reason for the high V/Hz ratio can be seen. The pulse isn't a clean square wave. There are large overshoots, and the area under them is what's boosting the voltage reading. Now what is going on here? Impedance mismatch would explain the rapid oscillations, but the large-area overshoots?

Is this a line-to-ground measurement? If so, at least a part of that overshoot might be common-mode and not contributing to the L-L voltage across a motor winding.

 

[Besoeker3]

Series motors? My experience is just typical cage motors.

 

[synchro]

Was the scope picture taken after the V/Hz was reduced? Also, did you take scope measurements both before and after the V/Hz change, and if so was there much of a difference in the waveforms?

 

[SiriusC1024]

This measurement is line-to-line motor-side.

IDDRPM364004R1

By series motor I meant product series. Sorry about that. Squirrel cage rotor. Single drive/motor combination. ~15 ft run of vfd cable.

Waveform shows same anomaly with reduced V/Hz. In fact, this scope shot is with the reduced V/Hz change applied.

Longer duty cycle pulses show the same voltage rise. Every pulse has it. Carrier frequency from 1.0 to 4.0 kHz (currently set to 1.2 kHz) makes no difference.
It occurs at the same time interval (roughly 20 microseconds after gate-on), and for the same duration and magnitude. What's telling is that, if the pulse is short enough, there is still voltage present at the terminal even after gate-off. I posted an example of a shorter pulse to illustrate that. The mechanism of action is beyond my current expertise.

 

[synchro]

Are these measurements being made relative to ground? If so, have you put a scope on the DC bus to see if the voltage to ground changes in a similar manner to the waveform that you showed. It looks like you can see the output switching from the negatve to the positve side of the DC bus, and back to the negative. But then you also see the rise you mentioned.

 

[SiriusC1024]

No, sir. These measurements are line-to-line. T1 to T2 for example using the oscilloscope probe and set to PWM mode.

Here is a longer duty cycle pulse like I mentioned. https://photos.app.goo.gl/LwXy5iACCvkM9fpG7

 

[synchro]

Is there a filter of some type on the VFD output that has capacitors, e.g., a dV/dt or sine wave filter?

 

[SiriusC1024]

That's a good thought, but no filters. I did try a 5% load reactor one time, but then removed it because it was ineffective. All it did was get so hot that the clearcoat started to burn off the laminations after a week.

 

[synchro]

Typically there are clamping diodes across the IGBT's in a drive to prevent any significant reverse voltage across them. Because of this, it would seem that the DC bus voltage would have to follow the overshoot voltage waveform that you show in your pictures. I think it could be worthwhile to put a scope on the DC bus and see if the overshoot is also present there.

 

[Besoeker3]

This measurement is line-to-line motor-side.

IDDRPM364004R1

By series motor I meant product series. Sorry about that. Squirrel cage rotor. Single drive/motor combination. ~15 ft run of vfd cable.

Appreciated - I thought it was a system I hadn't experienced !
If the cage motor is operating at constant torque the reduced frequency could well be an issue.

 

[SiriusC1024]

Appreciated - I thought it was a system I hadn't experienced !
If the cage motor is operating at constant torque the reduced frequency could well be an issue.

Yep I thought to check that, too. These motors have a 1000:1 turndown ratio, so unless the motor is absolutely crawling things should be stable. This reading was taken at 38.1Hz. Similar readings show at 50Hz, 42Hz, etc.

 

[winnie]

I have to admit this is very interesting. I don't understand it, and it is the sort of thing that I _should_ understand.

Both of the pulses that you show are fairly short duration; the 'long' pulse' is about 80 microseconds long. You are at a carrier frequency of 1.2 kHz, which means you are looking at about a 10% pulse duration.

1) Do the polarity of these small pulses change depending upon the where you are in the over-all AC cycle? In other words, if you look at say a 50% or 90% duration PWM period, do these small humps still show up, and does their polarity change?

2) Have you looked for corresponding pulses in the DC link voltage, or current pulses on the

3) Does the timing of the small pulses change with motor speed or motor load?

Just throwing ideas around:
Some sort of deadtime distortion
Measurement interaction with other phases
Instrument error (what is the scope grounding connection? scope power supply connection?)

-Jon

 

[synchro]

1) Do the polarity of these small pulses change depending upon the where you are in the over-all AC cycle? In other words, if you look at say a 50% or 90% duration PWM period, do these small humps still show up, and does their polarity change?

I was thinking along the same lines. If these added pulses always have the same polarity as the desired PWM pulses throughout an AC cycle, then they would always be adding to the desired sine wave amplitude. However, for example, if their polarity is always in one direction throughout the AC cycle then they would contribute a DC offset that could saturate the stator core during part of the cycle.

 

[SiriusC1024]

Here is the DC bus at the same time resolution 20us/div: https://photos.app.goo.gl/TyWwgmK9TpD7E8FV8

At a larger scale, in auto mode, there are voltage spikes.

Measurement setup is a Fluke 123b operated off battery power. The probe used is a Fluke STL120-IV.

As far as measurement error, I ruled that out. In the link to the motor above there is a performance tab supplied by the manufacturer. I plotted slip rpm vs amperage on the load performance (constant V/Hz) graph. Before manually rescaling the V/Hz ratio the line amperage and slip rpm values weren't co-linear with respect to the X-axis (can't physically have two different torques). After the change, and V/Hz ratio reading nominal on the oscilloscope, the data points on the two curves lined up.

I'll check polarity as requested and try to grab a longer duty cycle pulse.

 

[winnie]

Here is the DC bus at the same time resolution 20us/div: https://photos.app.goo.gl/TyWwgmK9TpD7E8FV8

At a larger scale, in auto mode, there are voltage spikes.

Given that the system is working now that you've adjusted the V/Hz ratio, and that you are chasing this down for your understanding on new year's eve, I think you and I have similar ideas of fun....

Are you saying that you see voltage spikes on the DC bus, but they are not resolving when you have the scope set to 20uS per division?

Are they much shorter duration pulses, or could they plausibly be the same duration as the pulses added to the PWM waveform?

Jon

 

[winnie]

Interesting that the meter section of the reading on the DC bus shows different DC and AC values

 

[SiriusC1024]

Yeah lol, but then again we probably both can't handle not understanding something. Also, I think there's potential for substantial efficiency gains, as well as drive and motor life, if this gets resolved. People will be able to reference this work, too.

When this scope gets charged up I'm going to try to log dual channel and overlay the AC drive output pulse with the DC bus.

 

[synchro]

When this scope gets charged up I'm going to try to log dual channel and overlay the AC drive output pulse with the DC bus.

That will be good for a comparison, and you should have something better to trigger on vs. the "noise" on the DC bus.

 

[SiriusC1024]

I figured it out. Measurement error.

Yep.

The initial measurements were taken using the STL120 probe on T1 to Channel A and a regular fluke probe on T2 to scope ground. I was looking at the manual on how to set up dual channel measurement without killing myself (don't ground each channel to separate potentials) and saw something about shielded probes. Used the STL120 probe to T1 again, but this time with the little alligator ground lead that plugs into it and attached it to T2. Both these are read by Channel A exclusively. Nothing to scope ground. The overshoots didn't show up. Turns out what was causing incorrect readings was noise from the high current PWM cross-talking to the unshielded neutral probe lead.

So yeah, I feel like an idiot, but I wanted to share anyway so someone can learn from my mistake. Don't trust the accuracy of unshielded meter probes within a drive cabinet.

And if anyone still has patience for this perhaps they can answer the basis of the problem I'm having. Why does decreasing rated voltage lead to lower power consumption and a cooler motor? As reported by the drive HMI:

18.4Hz commanded frequency

460V/60Hz (7.667 V/Hz) = 53.73kW 277A
460V/66.5Hz (6.917 V/Hz) = 43.39kW 288A