OT: PowerFlex 70 Torque Variation

kamenges

Member
Join Date
Nov 2002
Location
Green Bay, WI
Posts
4,331
We have been using PowerFlex 70 drives around here for some time for various web handling applications. Last month we came across an issue in an application that we seem to have traced back to an issue with PowerFlex 70 drives.

The application is a two-roll extrusion casting stand. the two rolls are independently driven through AC motors by PowerFlex 70 drives. This is something we have done before without any real issues. However, the material being cast in this application is much more sensitive than we initially believed. One of our engineers began seeing scuffing on the formed product and traced it back to casting roll speed variations. More specifically it is a speed variation between the two casting rolls and the variation seems to match up with motor rotation.

We initially thought it was a mechanical issue since it timed so well to motor revolutions. But we went through the drivetrain and didn't find anything. Our guy finally tracked it back to motor speed variations.

This is where it gets interesting. Our guy couldn't get the speed variations to go away. So we set up a test back here with just a motor and drive. We saw the same issues. It is effectively a speed pulsation that is timed to motor rotation. We get two speed pulses per motor rotation (this is a 4-pole motor). The magnitude of the pulsation seems to decrease as the motor speed increases. At speeds below about 30 RPM you can physically see the shaft pulse. The pulse frequency is tied to shaft speed.

We have tried the drive in sensorless vector and FOC. the pattern is still visible in sensorless vector but isn't as pronounced. We have tried multiple motors with the same PF70. Same result. We have tried multiple PF70 drives on each motor. Same result. We have tried other manufacturers' drives on these motors. These other drives run the motors we have without issue.

We have Rockwell hunting this down but so far all they are having us do is change gains and various other parameters. Nothing really helps. Also, I just found out today that we seem to be having this issue at another installation that is going on right now.

I guess the motivation for this post is twofold. First, I am trying to see if anyone else is seeing this kind of thing. I am looking for some ammunition to shoot at Rockwell. Secondly, I would suggest being a little careful with the PF70 at this point. I wouldn't be too concerned with general pump, fan or conveyor type applications. But anything requiring steady torque output, especially at low speeds, might be a challenge.

Keith
 
Keith, by FOC did you mean FVC (flux vector control with encoder feedback)?

Have you tried this with a Powerflex700? They are very similar so the result might be the same.
 
No, I mean FOC (field oriented control, with an encoder).

We have not tried a PowerFlex 700. We have enough of these drives in the field that our preference would be for Rockwell to figure out what is going on and get us a fix. If we end up going PF700 because Rockwell can't get the PF70 to settle down I would Rockwell would cough them up. But we aren't quite there yet.

Keith
 
do you have IGBT long lead reactors installed I found in some applications this greatly smoothed out the Sinusoidal/PWM translation. On a scope the PF 70 were not really all that smooth on its switching In one critical application I removed the PF 70 and installed an Omron G5 drive with reactors on infeed power and outfeen power and when I record it with a scope Its usually a clean sinusoidal waveform usually better than the incoming power. To date I have yet come across any other drive that had such a close representation of PWM to Sinusoidal and I've scoped over 20 different drives.
 
Acknowledging first of all that I might be prejudiced (I have a long history with ABB as a certified startup tech and territory manager), I think you have probably hit the nail on the head when you suggested that the PF70 be used with care.

While other leading manufacturers of high performance drives have chosen to design their systems with exceedingly fast and sometimes parallel microprocessors along with innovative motor control algorithms, AB has hung back on this commitment except in their very best systems drives. This is especially apparent when operating them in sensorless vector mode.

I am certainly not an experienced designer of drive electronics so the details are beyond me, but, even the published data is often not up to state-of-the-art levels. It's dated, I know, but, for example, the GV3000 published data for sensorless vector speed regulation was not any better than V/Hz with slip compensation (something around .5%)

Specific to your problem, I would try to run the PF70 in Flux Vector Control mode with a 1024ppr encoder, tighten up the speed loop gains as much as possible, and see what you get. That's about as good as it gets with a specific piece of hardware. Since it is unlikely that AB will redesign the PF70 just for you, if that doesn't give you what you need, its time to move on to a higher level model or another brand. You won't be the first or the last to do that.
 
I think I'm confusing some people with the whole FOC thing. Field oriented control is the flux control method used by most drive manufacturers today. With the processors available today it makes things much easier. Take a look at this for a decent overview of FOC
http://focus.ti.com/lit/an/bpra073/bpra073.pdf

DickDV can confirm or deny this but I think ABB's DTC still uses the front end concepts of the FOC but go their separate way with the generation of Iq and Id at the motor.

The GV3000 specs don't surprise me. That is a 15 - 20 year old design. The published speed regulation spec for the PF70 is 0.001% of base speed with 1000:1 speed range when operated in FOC. The PF70 requires encoder feedback in order to operate in FOC. I'm pretty sure their sensorless algorith is a completely different beast. That spec is 10 times worse (0.01%).

Originally posted by DickDV:

Specific to your problem, I would try to run the PF70 in Flux Vector Control mode with a 1024ppr encoder, tighten up the speed loop gains as much as possible, and see what you get.

We've tried that. We can reduce the magnitude and duration of the deviation but it is always there. We are running in FOC using a 4096 PPR encoder so I don't think this is a velocity control issue at it's core. It is more like the drive really doesn't know what the flux is twice per rotor revolution.

I think we will end up going down the road of a different manufacturer unless Roickwell can fix this pretty quickly. We have a customer waiting to get a machine running so we can't be a guinea pig in this case.

I'll keep everyone up to date on what we find.

Keith
 
We had a couple of Rockwell guys at our shop yesterday looking into the PF70. After a few hours of jacking with the gains they pretty much gave up. They asked our guy what level of speed regulation was required. Our guy said "What is says in the manual". He was laughed at. The Rockwell guys said that was unachievable as far as they were concerned.

These guys were tech support out of Mequon. They took our drive, motor and encoder home with them to play with.

The thing that disappoints me the most is that they really didn't look into the core issue. We have this repeating speed variation whose frequency is tied to motor rotation. Changing velocity loop gains changes the size of the variatin but not the wave shape or the frequency. Why would they think that jacking with the velocity loop gains would make it go away?

I'll let you know if the Rockwell guys come back with anything. We have decided to move on to a different vendor on this project.

Keith
 
Keith,

Interesting thread

Im curious- Do you see this pattern in plain old V/Hz mode?

(yes, I know the regulation goes down hill, Im curious about the pattern you see)

Last question- is there a "time slice" for the update time on the current and speed control loop?
 
Last edited:
I can't remember if we tried V/Hz or not. I know that SV and FOC were both an issue, although SV seemed to act a little better. I think that is just because the speed feedback was a little smoother in SV than with the encoder.

I'm not sure about the time slice. I'm pretty sure the velocity loop is on a 10 msec update, but I may be remembering that from the 1336 Impact days. The current loop update should be quite a bit faster than the speed loop.

Keith
 
Just another update.

We heard back from the Rockwell tech support guys. Apparently all the PF70 is supposed to be capable of is 0.1% of base speed. The published specs are "misleading" (their word, not mine). So they have pretty much thrown their hands up and said "Sorry". They are going to try a PF700 with our motor and encoder and see what they get. But if that is our out we have much less expensive options.

You may want to scribble in your PF70 manuals that the true speed regulation spec is 0.l% of base speed.

Keith
 
OK, let's put some real numbers to this. Let's say a four pole 60hz motor. That would be a base speed of 1800rpm. 0.1% of 1800rpm is 1.8rpm. If the motor is a modern premium efficient motor, then it likely has around 1% slip from no-load to full-load so it would be nameplated around 1782rpm.

That means that, with an encoder, the drive is able to regulate speed to 1.8/18 or one-tenth of motor slip (which is by far the largest component of speed loop error).

That level of speed regulation is acheivable with several premium sensorless vector drives (that's without an encoder). With an encoder, the same premium drives would claim speed error not to exceed 1 pulse plus/minus 1 pulse non-culumulative. In fact, with an encoder, I've never seen a drive claim anything less than that. 0.1% of base speed is really poor, in my prejudiced opinion.

One other comment about speed regulation and this is not drive specific. When you see a speed regulation spec, the error is the same over the whole speed range. So, if, as in the example above, the speed error is 1.8rpm, then, at 1800rpm it is .1%: at 180rpm it is 1%, at 18rpm it is 10% and at 1.8rpm, it's all error. See why operation at slow speed is so challenging?
 
Originally posted by DickDV:

When you see a speed regulation spec, the error is the same over the whole speed range.

You're right. A significant percentage of drive users don't fully appreciate what this means. And it can bring them heartache.

But, again, in this case I don't think AB is doing themselves any favors. They appear to be inducing a velocity error into their own system. When I get home I'll see if I can post a plot of what we were seeing. She's a beaut.

Keith
 
I found this DOC in one of my emails. It shows the speed feedback as read from a CLX processor over Ethernet of an unloaded motor driven at 10 and 20 RPM by a PowerFlex 70. This is with the drive in FOC with an encoder. I personally spun this motor by hand. There are no 'sticky' spots. And as previously stated, two other drives from different manufacturers spun this same motor without issue.

Any ideas?

Keith
 
Keith, I only deal with drives as assembled devices and am not equipped to discuss the internals or how the processing is done. In the past, Rockwell and especially Reliance before them had notoriously slow processors which would cause periodic control errors as the microprocessor either gets behind or is off doing some other housecleaning work.

Maybe there are some other VFD designers out there that can comment on the internal performance issues but that's not me! I just reach for a different brand.
 
One of the issues I will face, being employed by a small OEM, is that I am doing to have PF70s crammed down my throat. Rockwell has done a very good job of marketting to end users. So I get told what to use by my customers. All I can do is present evidence to my customer and hope he accepts it. Most often they do not unless they are a repeat customer. Even if a specific plant were to believe me we often don't get involved at the corporate level with our customers. So if Rockwell is taking the corporate head of engineering on golf outings every other week why would that same head of engineering believe me when I tell them to switch away from Rockwell?

So ultimately it is in my best interest to have Rockwell get this figured out. I'm going to have to deal with it either way. But maybe this can help someone else.

Keith
 

Similar Topics

Hello, Would anyone happen to know what parameter in a PowerFlex 525 to have the drive trip if the torque permissive set point is exceeded...
Replies
3
Views
391
Hi all, I'm looking at an application with a Powerflex 755 with encoder feedback, being run as a standard VSD (as opposed to a motion axis), with...
Replies
5
Views
1,679
Hello all, I am using a PowerFlex 700 (Rev 10.001) to maintain tension in a web. We're reading lbs from load cells under one of the rolls...
Replies
85
Views
11,295
Hi all, I am having issues with a Powerflex 527 STO. I am hard wiring a safety relay (guard master MSR27TP) to S1 and S2 and also the common to...
Replies
5
Views
1,629
Good Morning , I'm building a control panel with PowerFlex 525 drives. I have black wire for the 480 vac , red for 110 vac , blue for 24 vdc...
Replies
7
Views
2,866
Back
Top Bottom