Repeatable Velocity Error on Encoder

[DaveW]

I have a quadrature incremental encoder measuring the speed of a conveyor and it is hooked up as a feedback only axis on a Kinetix 6500 servo. I was looking at the feedbackaxis.actualvelocity parameter and the error is a repeatable pattern starting out small, gradually increases, and then abruptly back to small again. The encoder cable length is about 40 meters. I've replaced the encoder with no luck. I've check grounding and the cable is in a low voltage wire tray. Can this be noise? Any idea?

 

[Peter Nachtwey]

I don't think that is a wiring noise problem.

I have a quadrature incremental encoder measuring the speed of a conveyor and it is hooked up as a feedback only axis on a Kinetix 6500 servo. I was looking at the feedbackaxis.actualvelocity parameter and the error is a repeatable pattern starting out small, gradually increases, and then abruptly back to small again. Any idea?

Yes, where is the feebackaxis.actualvelocity parameter computed? Is it computed in the drive or does the graphing software compute it from the difference in position? It looks like some sort of beat frequency between the sampling and the computations.

I could not see a scale that meant anything. How fast is the speed and how much is it varying?

 

[DaveW]

The parameter is available from the CIP axis (AXIS__CIP_DRIVE) when you create the axis in ControlLogix.
The speed average about 4 inch/sec plus/minus 5. The pulses are about 1 per second.

 

[Peter Nachtwey]

The pulses are about 1 per second.

Is that one encoder pulse per second? It better not be.

 

[DaveW]

No, the beat frequency you are talking about is about 1 per second.

 

[MichaelG]

what is the motion group task coarse update time and the RPI connection to the 6500?

 

[DaveW]

what is the motion group task coarse update time and the RPI connection to the 6500?

Coarse Update Time = 2ms
RPI = 0ms

both are default values.

 

[Lancie1]

No, the beat frequency you are talking about is about 1 per second.

Yes, I agree. Blowing up the picture 1000 times, there are about 15 peaks each second, or 30 cycles per second, or a 30 Hertz repeating error. If the program is missing every other peak, then that would be a 60-cycle rate. Sound familar? I would add more shielding to filter out the AC induced voltage.

 

[leitmotif]

Do you have this in shielded cable AND if so is shield grounded on only one end?

Somewhat wild BUT do you run this cable thru an area with TIG welding?

Shut machine down and measure the cable with the scope to see if it is getting EMI ??

Dan Bentler

 

[Peter Nachtwey]

This isn't a noise problem, at least not an electrical noise problem. This could have to do with a beat frequency between sampling at the drive and when the positions are relayed to the PLC to compute the velocity. If the velocity was calculated at the drive there shouldn't be this problem but the velocity is calculated at the PLC or maybe even in the RSLogix.

I bet this is one of the nasty little problems one has when sampling and calculations are done asynchronously.
Look at the trend showing the positions increasing as expected. The problem is not the encoder but how the velocity is calculated.

 

[DaveW]

...would be a 60-cycle rate. Sound familar? I would add more shielding to filter out the AC induced voltage.

Lancie, I thought about that but the 15 peaks gives 15hz and I would have to miss 3/4 peaks to get the 60hz. We use Accu-Coder cables with the double foil and shield and it's grounded at the servo drive's end.

Do you have this in shielded cable AND if so is shield grounded on only one end?
Somewhat wild BUT do you run this cable thru an area with TIG welding?
Shut machine down and measure the cable with the scope to see if it is getting EMI ??

Dan, yes ground at the servo end and no TIG welding. For EMI measurement, just take the leads off the servo end and measure to ground?

If the velocity was calculated at the drive there shouldn't be this problem but the velocity is calculated at the PLC or maybe even in the RSLogix.

Peter, I believe it's calculated in the drive since there's some settings for the axis where you can request the drive to send the velocity and position the cpu ever cycle.

Accu-Coder want to sell us a repeater to be installed somewhere in the middle of the run because they think the cable length is too long. AB says the cable length is not a problem.
I'm going to go through the cable tray and see if there's any AC cables or wires. Thanks for all the suggestions.

 

[shooter]

it looks a timing problem, do you reset any counter?
if so change it and sub the value.

 

[Peter Nachtwey]

Peter, I believe it's calculated in the drive since there's some settings for the axis where you can request the drive to send the velocity and position the cpu ever cycle.

Then the drive can't calculate velocities accurately. It wouldn't be the first time. That is why I am so fanatical about algorithms and especially those that calculate velocities and accelerations.

Accu-Coder want to sell us a repeater to be installed somewhere in the middle of the run because they think the cable length is too long. AB says the cable length is not a problem.

I believe AB. Why would the positions look good and not the velocity?

I'm going to go through the cable tray and see if there's any AC cables or wires. Thanks for all the suggestions.

Again, this is not a cable noise problem. If it were the positions would look bad too.

Ask your suppliers the same question I just asked.

It would be good to have the trend exported and zipped to a CSV file so I could look at the raw data.