[seapacer2]

Tom,

I moved my table 12" and the hand crank turned 60 times,

My servos are brand new DMM 7.5K AC and they are powered by their Dyne 4 drives and the motors spin freely in both directions.

Tonight, I connected the drive belt on the X axis and reran the tests, hopefully I have everything you requested, and tonight the drive had some movement. Please see the attached files.

Let me know what else I need.

Thanks,

Mark

[seapacer2]

Tom,

A little closer?

Thanks, Mark

[Jim Dawson]

Mark,

Are you running your DMM's in Analog torque servo or analog speed servo mode?

[seapacer2]

Jim,

I will go in and check that tonight.

Thanks,

Mark

[Jim Dawson]

The reason I ask that is because you should do some ''tuning'' and setup in the DMM software before you try to tune the system in the Dynomotion software. In analog torque mode the only parameter that really has any effect is Torque Filter Constant. That and your ''lines'' output. This should be set to an integer value taking into account your ball screw pitch and pulley ratio. I hope that makes sense.

[TomKerekes]

Hi Mark,

I moved my table 12" and the hand crank turned 60 times,

If we also knew the number of encoder counts per crank revolution then we could calculate the resolution. You might just look at the Encoder Position on the KMotion.exe Axis Screen before and after turning the crank 60 times. As Jim points out the number of encoder counts per motor revolution that the Drive puts out is probably configurable in the Drive. That should normally be configured as high as possible without exceeding KFLOP's max count rate of 1 million quadrature counts per second.

As Jim points out you should tune the Drives as good as possible so they respond quickly to what they are told to do by KFLOP. It is hard to tell what mode they are operating in. Torque (which is basically acceleration) or Velocity. If you made a long move this would be more obvious. The Output (green) would either be relatively constant high value while moving at constant speed or it would be at large values during the start or the stop of the move. A car analogy would be is KFLOP's Output controlling the cruise control of the car or the gas pedal of the car.

The plot is starting to look reasonable. Really only P gain is currently in use. Again it would be useful if you provided feedback on what you see in the plots. Its ok to be wrong as this is probably your first time tuning.

I always smile when I see a Gain value something like 4.015625. That's sort of like seeing a speed limit posted for 60.015625 MPH. Usually things like gains need to be changed by 10% or more to have any noticeable effect. So use numbers like 4 or 5.

The I gain (Integrator) is having no significant effect because it is being limited to a max limit of 1 DAC count. Since it takes hundreds of DAC counts to move the limit should probably be set to 1000. The limit should not normally have an effect. The limit should only come into effect if something abnormal happens ie Crash. The Integrator is a very powerful tool to eliminate persistent errors. If there is any error it will ramp the output (up to the limit) as long as there is any error. An analogy might be with a car. Say the car stops one foot short of the stop sign. And the Driver is told to keep moving the gas pedal down as long as the car is not exactly at the stop sign. Can you see that eventually the car will move to the stop sign? But that being said an Integrator can be very destabilizing. Can you see pushing the gas more and more right up to reaching the sign it is likely an overshoot may occur? That is why the I Gain is normally small so that it gradually moves the gas pedal. But it shouldn't be so small that errors aren't corrected in a reasonable time. And that's why other controls are needed to avoid instability. One of which is D gain which acts proportional to velocity. So as the Integrator is depressing the gas pedal as soon as the car starts moving the Derivative gain will quickly start backing off on the gas to slow down. The P gain also helps avoid overshoot but to a somewhat lesser extent. It looks not at speed but rather at the distance to target. As the distance reduces it then reduces the gas.

HTH

[seapacer2]

Hi Mark,

If we also knew the number of encoder counts per crank revolution then we could calculate the resolution. You might just look at the Encoder Position on the KMotion.exe Axis Screen before and after turning the crank 60 times. As Jim points out the number of encoder counts per motor revolution that the Drive puts out is probably configurable in the Drive. That should normally be configured as high as possible without exceeding KFLOP's max count rate of 1 million quadrature counts per second.

As Jim points out you should tune the Drives as good as possible so they respond quickly to what they are told to do by KFLOP. It is hard to tell what mode they are operating in. Torque (which is basically acceleration) or Velocity. If you made a long move this would be more obvious. The Output (green) would either be relatively constant high value while moving at constant speed or it would be at large values during the start or the stop of the move. A car analogy would be is KFLOP's Output controlling the cruise control of the car or the gas pedal of the car.

The plot is starting to look reasonable. Really only P gain is currently in use. Again it would be useful if you provided feedback on what you see in the plots. Its ok to be wrong as this is probably your first time tuning.

I always smile when I see a Gain value something like 4.015625. That's sort of like seeing a speed limit posted for 60.015625 MPH. Usually things like gains need to be changed by 10% or more to have any noticeable effect. So use numbers like 4 or 5.

The I gain (Integrator) is having no significant effect because it is being limited to a max limit of 1 DAC count. Since it takes hundreds of DAC counts to move the limit should probably be set to 1000. The limit should not normally have an effect. The limit should only come into effect if something abnormal happens ie Crash. The Integrator is a very powerful tool to eliminate persistent errors. If there is any error it will ramp the output (up to the limit) as long as there is any error. An analogy might be with a car. Say the car stops one foot short of the stop sign. And the Driver is told to keep moving the gas pedal down as long as the car is not exactly at the stop sign. Can you see that eventually the car will move to the stop sign? But that being said an Integrator can be very destabilizing. Can you see pushing the gas more and more right up to reaching the sign it is likely an overshoot may occur? That is why the I Gain is normally small so that it gradually moves the gas pedal. But it shouldn't be so small that errors aren't corrected in a reasonable time. And that's why other controls are needed to avoid instability. One of which is D gain which acts proportional to velocity. So as the Integrator is depressing the gas pedal as soon as the car starts moving the Derivative gain will quickly start backing off on the gas to slow down. The P gain also helps avoid overshoot but to a somewhat lesser extent. It looks not at speed but rather at the distance to target. As the distance reduces it then reduces the gas.

HTH

Merry Christmas to everyone!

[seapacer2]

Good morning all,

I hope everyone had a good Christmas!

I went back out into the garage on Sunday to continue turning the "X" axis. Before continuing servo turning in the KMotion software, I went back to look at the parameters that we originally set up in the DMM software. I reached out to someone with a lot more experience than myself on the DMM servos to take a look, they seem to look OK, I have also attached them to this post for your review.

Tom, you asked me resolution, I hope I'm explaining this correctly. I went back to another post, my gear ratio of my "X" and "Y" is 1.5:1 and my "Z" is 1.8 with my current belt and gear configuration, this is what the Anilam Crusader M was originally set up. Since this machine is a 1976 vintage model, I believe that this machine utilizes .200 lead ball screw setup, making it 5 turns of the hand crank equals 1".
The DMM line count range is from 500 to 4096, I had to change the line number in the DMM software form 2048 to 2500.

2500 * 4 = 10,000 * 1.5 = 15,000 encoder counts for each 0.200" of movement ( one rotation of the ballscrew).


Next you will see the attached graph, it seems like I'm a little closer with the exception of small oscillation (Overspeed?) at the end, It seems like my number P / I / D are larger than I was expecting. I tried utilizing the I/R filters (Not even sure I know how to utilize the Filter screen property?), but no matter the values, it seemed to make the graph worse. However, my gas pedal doesn't really seem to smooth out (Green). Any suggestions, anyone??

This is the sequence I utilized every time I ran this test:
- Press the "Disable" button.
- Press the " Zero" button.
- Press the "Enable" button.
- Press the "Move" button

After I pressed the "Move" button, I observed the "X" axis hand crank, it would rotate in one (1) direction and then the other direction. The amount of rotation is very slight, between .500" and 1.00" of movement, the table would move slight as the hand crank turned. The only way I could get the table to move more, is to insert a value larger than zero in the "Feed Forward" V and A, when this happens I always get a large oscillation (overspeed) at the end of the move, if the value is too high, I need to hit the "E-stop" to cancel. What do I need to revise to improve?

[TomKerekes]

Hi Mark,

2500 * 4 = 10,000 * 1.5 = 15,000 encoder counts for each 0.200" of movement ( one rotation of the ballscrew)

Ok so that would be 75000 counts/inch.

Test a move of size 75000 does it move 1 inch? 750000 does it move 10 inches?

I had to change the line number in the DMM software form 2048 to 2500.

Although in the DYN4 Screen shot you posted it shows the line count at 2048. Isn't that inconsistent with what you told us?

I tried utilizing the I/R filters (Not even sure I know how to utilize the Filter screen property?), but no matter the values, it seemed to make the graph worse. However, my gas pedal doesn't really seem to smooth out (Green). Any suggestions, anyone??

IIR Filters are somewhat advanced and complex. You probably won't need to use them. In the Screen posted all the Filters are cleared (off). You just need to post it or tell us that all the filters are cleared.

Next you will see the attached graph, it seems like I'm a little closer with the exception of small oscillation (Overspeed?) at the end, It seems like my number P / I / D are larger than I was expecting. I tried utilizing the I/R filters (Not even sure I know how to utilize the Filter screen property?), but no matter the values, it seemed to make the graph worse. However, my gas pedal doesn't really seem to smooth out (Green). Any suggestions, anyone??

It looks very unstable. I think the P and D gains are too high. But the strangest thing is after the positive move from about 0.5 seconds to 0.8 seconds the output remains at 2047 (again pedal to the metal). Can you see this? Please answer this question. Its hard to see as I think the green line is behind the blue line. You didn't post the raw data as I asked you previously to include so we could zoom in to see clearly. It seems the DYN4 shut down for some reason. You will need to determine why this is happening. Maybe an over current condition or something. Possibly if you reduce the gains which should result in a smother, slower, less aggressive move then maybe the DYN4 will not shut down.

This is the sequence I utilized every time I ran this test:
- Press the "Disable" button.
- Press the " Zero" button.
- Press the "Enable" button.
- Press the "Move" button

That's the safest approach but in most cases after a successful move you can get away with just pushing the "Move" button.

The amount of rotation is very slight, between .500" and 1.00" of movement, the table would move slight as the hand crank turned.

The size of the movement is set by the "size" parameter. 1000 counts on a system with 75000 counts/inch would only be 0.013 inches. Try a bigger move such as 75000 which should be 1 inch if your earlier information was correct.

The only way I could get the table to move more, is to insert a value larger than zero in the "Feed Forward" V and A, when this happens I always get a large oscillation (overspeed) at the end of the move, if the value is too high, I need to hit the "E-stop" to cancel. What do I need to revise to improve?

For now leave the Feed Forwards at zero and change the size parameter to change the size of the move.

[seapacer2]

Tom,

2500 * 4 = 10,000 * 1.5 = 15,000 encoder counts for each 0.200" of movement ( one rotation of the ballscrew)


Ok so that would be 75000 counts/inch.

Test a move of size 75000 does it move 1 inch? 750000 does it move 10 inches?

Yes,

I had to change the line number in the DMM software form 2048 to 2500.


Although in the DYN4 Screen shot you posted it shows the line count at 2048. Isn't that inconsistent with what you told us?

That is before I changed the value from 2048 to 2500. I will make sure I have the correct screen shot, I posted from work today.

tried utilizing the I/R filters (Not even sure I know how to utilize the Filter screen property?), but no matter the values, it seemed to make the graph worse. However, my gas pedal doesn't really seem to smooth out (Green). Any suggestions, anyone??


IIR Filters are somewhat advanced and complex. You probably won't need to use them. In the Screen posted all the Filters are cleared (off). You just need to post it or tell us that all the filters are cleared.

I will keep the filter cleared for now until I can grasp this better.

Next you will see the attached graph, it seems like I'm a little closer with the exception of small oscillation (Overspeed?) at the end, It seems like my number P / I / D are larger than I was expecting. I tried utilizing the I/R filters (Not even sure I know how to utilize the Filter screen property?), but no matter the values, it seemed to make the graph worse. However, my gas pedal doesn't really seem to smooth out (Green). Any suggestions, anyone??


It looks very unstable. I think the P and D gains are too high. But the strangest thing is after the positive move from about 0.5 seconds to 0.8 seconds the output remains at 2047 (again pedal to the metal). Can you see this? Please answer this question. Its hard to see as I think the green line is behind the blue line. You didn't post the raw data as I asked you previously to include so we could zoom in to see clearly. It seems the DYN4 shut down for some reason. You will need to determine why this is happening. Maybe an over current condition or something. Possibly if you reduce the gains which should result in a smother, slower, less aggressive move then maybe the DYN4 will not shut down.

Yes, I know that the value are extremely high, I will post the raw data as requested next time.

But the strangest thing is after the positive move from about 0.5 seconds to 0.8 seconds the output remains at 2047 (again pedal to the metal). Can you see this?

What am I looking for? I can see movement in the hand wheel. Should I be reducing the output? My drives and motor are brand new, I will try adjusting the output.

This is the sequence I utilized every time I ran this test:
- Press the "Disable" button.
- Press the " Zero" button.
- Press the "Enable" button.
- Press the "Move" button


That's the safest approach but in most cases after a successful move you can get away with just pushing the "Move" button.

Thanks,


The amount of rotation is very slight, between .500" and 1.00" of movement, the table would move slight as the hand crank turned.


The size of the movement is set by the "size" parameter. 1000 counts on a system with 75000 counts/inch would only be 0.013 inches. Try a bigger move such as 75000 which should be 1 inch if your earlier information was correct.

Adjusted the size to 75000.


The only way I could get the table to move more, is to insert a value larger than zero in the "Feed Forward" V and A, when this happens I always get a large oscillation (overspeed) at the end of the move, if the value is too high, I need to hit the "E-stop" to cancel. What do I need to revise to improve?


For now leave the Feed Forwards at zero and change the size parameter to change the size of the move.

Will do.

[TomKerekes]

Hi Mark,

Thanks for the detailed responses. Those are mostly all correct.

Ok so that would be 75000 counts/inch

Test a move of size 75000 does it move 1 inch? 750000 does it move 10 inches?

Yes

I don't know what "Yes" means. Did you test both of those? Can you use more words in your response?

What am I looking for? I can see movement in the hand wheel. Should I be reducing the output? My drives and motor are brand new, I will try adjusting the output

You should be looking for whenever the green plot is at a big value and the axis should be accelerating at a big value not hardly moving. You didn't answer my question of whether you can see the green plot stuck at 2047 (pedal to the metal but car doesn't go). Again if you reduce gains this effect may go away, But first you need to understand when such crazy behavior is happening.

[seapacer2]

Tom,

I went back to troubleshoot why I could not tune the servos through the DMM software, I found a bad servo on the "X" axis. Swapped the "Z" axis and installed on the "X", I then was able to perform the DMM autotune procedure without the drive faulting. I then reconnected the 25-pin connector to the KLOP and Kannalog, I started with opening the console window, and tried to run a command with DAC, nothing. I validated the wiring from the servo axis swap. It is correct, why did it work before? Am I missing something? Why would it work before and not now?

Thanks,

[TomKerekes]

Hi Mark,

I don't really follow that. But it should be straightforward to troubleshoot. Command a DAC value. Verify with a voltmeter on the Kanalog DAC output that the voltage is correct. Verify there is Voltage on the input to the amplifier. If there is command voltage going into the amplifier but the motor isn't moving then the problem is in the amplifier/motor. If there isn't any voltage going into the amplifier then the problem is with KFLOP/Kanalog or wiring.

[seapacer2]

Tom,

It looks like I'm getting voltage in and out from the analog I /O screen. All the components are brand new nothing it left from the Anilam system. You telling me now the wiring is wrong, or I have more bad components? It operated before, that is what is so confusing. See the attached file.

Thanks,

[TomKerekes]

Hi Mark,

You don't have a voltmeter?

[Jim Dawson]

Tom,

I had a phone conversation last evening with Mark. We found a wiring error in the drive enable circuit that will soon be corrected. This would have been corrected some time ago, but the apparently bad X servo motor caused a bit of confusion. Mark is on the road for a few days, and will get back to this soon.

Jim

[seapacer2]

Tom,

Please see the attached screen shot and saved file of the "X" axis tuning. We might be getting closer. All the filters are clear.

P = 1
I = .001
D = 200

Max limits:
Output = 200
Integrator = 500
Error = 2000

Motion Profile:
V = 20000
A = 100000
J = 4e + 06

Step:
Time/Secs = 3.5
Size = 35000

We also increased the size to 140000 and moved the axis 2.00"

I'm I headed in the direction? Just working on the "X" axis.

Thanks,
Mark

[TomKerekes]

Hi Mark,

That's following within 0.004 inches.

Can you see the Output (green) has 200 count spikes? This is because the D gain is 200 and there is not any Filtering. Add a 2nd Order Low Pass filter Freq=500 Q=1.4 and see if the spikes are reduced.

Can you see the Output is being clipped at 200?

[seapacer2]

Hi Mark,

That's following within 0.004 inches.

Can you see the Output (green) has 200 count spikes? This is because the D gain is 200 and there is not any Filtering. Add a 2nd Order Low Pass filter Freq=500 Q=1.4 and see if the spikes are reduced.

Can you see the Output is being clipped at 200?

Tom,

Anything higher than 200 for the output, the hand wheel wanted to oscillate really bad. I know that the output should be somewhat higher, will adding the filter reduce oscillation and allow me to increase the output?

Are we making progress?

Thanks,

Mark

[seapacer2]

Tom,

I added the filter as requested, it seemed like to took care of the spiking. i was also able to increase the output and integrator. It seems that the position is still following the command fairly well. Should I try to increase the output and integrator some more?

Thanks,

Mark