Tree325 Retrofit Started

[PeterTheWolf]

....
.....

I gain virtually forces overshoot. This is required in order to have zero average error. The amount behind and the amount ahead must be equal. Mathematically the area between the red and blue plots above and below must be equal over all time.

Looking at the plots above there is a significant persistent following error while moving. By adding I gain we should be able to prevent it from growing so large and then to drive it back to zero.

I Gain values are normally very small because they rapidly (every 90us) accumulate. Start with a very small number such as 0.00001. Or the smallest number necessary to see little or no effect. Then increase. Errors should reduce. Change the Plot type to Error and Output to see Error magnitudes better. At some point the system will go unstable. As usual back off by a significant margin and until the motion is most desirable.

Regards

Tom, I tried applying the Gain "I" for the X-axis. What I first found was that I=0.0000001 would give no effect on the "Position, Command, Output vs Time Plot Type".
Then I started increasing the by one. I was not completely sure what I was looking for the "Plot: Position Error, Output vs Time, Sec" screen.
I did take the "I" value up until I saw it start to oscillate (about 1-2 degrees in each direction servo axis pulley), which was at a value of about I=.01 I=.05 really started to make the servo go nuts. Then I started to back off from about I=.001 until I saw no physical over-shooting on the servo-motor-axis. I did see this over-shooting on the tail-end of the Error plot.

Hopefully this video will help to clarify what I was doing. Video of Gain - I (ZIP file 112mb)

This video shows where I think I need to be on the final Gain "I" setting ... please correct me if I am wrong here. Video of final settings on Gain I (ZIP file 47mb)

Here is the saved Plot File "GAIN_I_TEST.zip" of, what I think is the final setting (all settings can be seen in the "video of the final settings on Gain I" above).

I am finding that setting the Gain "I" is more difficult to tune correctly and interpret correctly form the Error Plot screen.

.....
.....

J325 Quick Links:

Machine
Existing Machine Schematics
Electronic Cabinet-Right Side
Electronic Cabinet-Back SIde
Existing Drive Board SD1525-10
J325 Servo Drive-SD1525 Manual
3-Phase Rotary Convert Used
RickB's J325 Retrofit Wiring
KFLOP 5VDC/15Watt/3A Power Supply
KANALOG Mounted & Connected
Kmotion - Axis Encoder Manual Test of Position via Manual Movement
Kanalog-Encoder Voltage High/Low Checks & 1KOhm Resister
Kmotion Configuration Screens "RUN-AWAY"

.....
.....

[TomKerekes]

Hi PeterTheWolf,

Oops. My bad. I forgot to tell you to increase the max Integrator from 200 to ~2500. Limited to 200 means the Integrator can't to what it needs to do.
Dynomotion.

Otherwise you are doing the correct process, but don't get any positive results because the Integrator is not being allowed to do what it needs to do.

Please repeat the test with max Integrator set at 2500.

Here are some comments regarding your videos.

The idea is to get the red to follow the blue and equivalently the blue error plot to be minimum. Right now the max error peaks at ~1200 encoder counts. We should be able to get that much smaller.

It doesn't make sense to try to match the error to the output. They are using different scales (the left and right) and the software is auto scaling so the plots jump around relative to each other. Don't concern yourself with this.

Dithering 1 encoder count (or even a few is normal) and good! That is only 1/20480th of an inch. You might read this article on dithering:
Dynomotion

Notice that even though the Integrator was limited to 200 DAC counts it was able to drive steady state errors to zero. While at rest, instead of always being off target by 5 or 12 counts as before because of some offsets, it is now off by 1 or less! However while moving the Integrator needs to be allowed to do way more (1600+ DAC counts) to be able to pull the error to zero.

Regards

[PeterTheWolf]

.....
......

Please repeat the test with max Integrator set at 2500.

Here are some comments regarding your videos.

The idea is to get the red to follow the blue and equivalently the blue error plot to be minimum. Right now the max error peaks at ~1200 encoder counts. We should be able to get that much smaller.

Notice that even though the Integrator was limited to 200 DAC counts it was able to drive steady state errors to zero. While at rest, instead of always being off target by 5 or 12 counts as before because of some offsets, it is now off by 1 or less! However while moving the Integrator needs to be allowed to do way more (1600+ DAC counts) to be able to pull the error to zero.

It would seem that the Max Limit on the Integrator set to 2500 did not do much for me in regards to reducing the max error pecks of ~1200 on the blue line (from the Position Error Plot).

In this video I pushed the I-Gain out of control and then brought it back until I could get it as best as I could; however, I still do not have the Position (Red Line) equivalent with the Command (Blue Line). In fact, I not sure it made that much of a difference with the settings I ended up with on the I-Gain with the Max Limit on the Integrator set to 2500.

Video of I-Gain Test with Integrator set to 2500 (ZIP File 139mb)

Saved Plot Data File per parameters I ended up with (ZIP file)

.....
.....

J325 Quick Links:

Machine
Existing Machine Schematics
Electronic Cabinet-Right Side
Electronic Cabinet-Back SIde
Existing Drive Board SD1525-10
J325 Servo Drive-SD1525 Manual
3-Phase Rotary Convert Used
RickB's J325 Retrofit Wiring
KFLOP 5VDC/15Watt/3A Power Supply
KANALOG Mounted & Connected
Kmotion - Axis Encoder Manual Test of Position via Manual Movement
Kanalog-Encoder Voltage High/Low Checks & 1KOhm Resister
Kmotion Configuration Screens "RUN-AWAY"

.....
.....

[TomKerekes]

Hi PeterTheWolf,

Ok things are becoming complicated because we are testing with the default trajectory that might be a bit too aggressive for your system. That is we aren't able to follow the Tesla even with no bricks under our gas pedal.

As a side note it would have been nice to see some I gain plots between I=0.009 unstable and I=0.0009. That's a factor of 10 different with no tests in between. But in this case I don't think it would have made much difference.

Here are some of your plots with comments:









Notice in the last plot the Output (green) quickly goes to max of 2047, but the Position (red) still doesn't quite keep up with the Command (blue).

I think we must slow down the Trajectory (Motion Profile) to something your system can achieve (tell the Tesla to drive a bit less aggressively so we can keep up).

Whenever we can't keep up we lag behind and later the Integrator causes overshoot to compensate. So the strategy to avoid overshoot is to try to prevent lagging behind in the first place. But in these cases our pedal is to the floorboards and we simply aren't able to keep up.

The motion profile parameters determines the desired trajectory. Currently we have:

Velocity = 40000 counts/sec = 40000/20480 = 1.95ips = 117ipm
Acceleration = 400000 counts/sec = 400000/2048 = 19.5in/sec2 = 0.051G
Jerk = 4000000 counts/sec3 = 195in/sec3

Do you have any specifications for what this machine should be capable of? With regard to Rapids or cutting Feedrates?

Let's try cutting everything in half to see how things go:

Velocity 20000
Acceleration 200000
Jerk 2e6

Hopefully you will be able to understand and see in the plots when the system is able to keep up or not and make adjustments and compromises. You should be able to trade acceleration for velocity and vice versa. You might read this article:
Dynomotion

Please try optimizing I again with these reduced goals.

Regards

[PeterTheWolf]

.....
.....

Do you have any specifications for what this machine should be capable of? With regard to Rapids or cutting Feedrates?
....
....
Hopefully you will be able to understand and see in the plots when the system is able to keep up or not and make adjustments and compromises. You should be able to trade acceleration for velocity and vice versa.

Tom, Thanks for the greatly needed feedback. As I make these tests, read the articles you advise, and apply your feedback I am gaining understanding on what it is I am trying to do.
I have been watching some youtube videos on transfer Systems, Bode Plot Stability, Gain and Phase Margins (Stability Margins), etc. explained by people with PhD. who train in Universities.

I am trying my best. I have three months in hands-on training (actually in hours, it is maybe a total of 15 hours) on subjects of electronics and motion (Transfer Systems).
With this forum as my main exposure to getting many questions answered in electronics/motion (Transfer Systems) it is quite challenging for me.
I welcome the challenge and the continued push to do more.
However, without this forum I am completely lost.
I am trying my best to understand the great knowledge you have gained over the, I am guessing, 30 years working with these systems.
Please forgive me if I cannot quite grasp these concepts as fast as expected.

I will test more thoroughly this time.

Thank you for your continue patients and feedback ...... I truly do appreciate it!

By Design, The Tree Journeyman 325 capabilities are as follows:

Rapid Traverse: 250 ipm ( I would more than happen at 150-200 ipm)
Feedrates: .100” to 100” ipm ( I would be more than happy at .100-70 ipm)
Positioning Accuracy: ±.0005” / 10” (I would be more than happy with ±.001"-.002" / 10”)
Positioning Repeatability: ±.00015” (I would be more than happy with ±.001”)
Resolution: 0.0001” (I would be more than happy with .0003”-.0004”)
Feedback by Closed Loop Optical Encoder
Axis Drive (DC Servo Motor) 1700 lbs. thrust
Milling Capacity (Mild Steel) 3.0/5.0 cu. In./min

.... Link to High Res Image

.....
.....

J325 Quick Links:

Machine
Existing Machine Schematics
Electronic Cabinet-Right Side
Electronic Cabinet-Back SIde
Existing Drive Board SD1525-10
J325 Servo Drive-SD1525 Manual
3-Phase Rotary Convert Used
RickB's J325 Retrofit Wiring
KFLOP 5VDC/15Watt/3A Power Supply
KANALOG Mounted & Connected
Kmotion - Axis Encoder Manual Test of Position via Manual Movement
Kanalog-Encoder Voltage High/Low Checks & 1KOhm Resister
Kmotion Configuration Screens "RUN-AWAY"

.....
.....

[mmurray70]

Are you guys sure the counts per inch is right? Toms calculations are based on 20480, which is a resolution of 0.00005 but the specs say 0.0001 which is double that. If you do a move of 20480 does it move an inch?

If the counts per inch is ok id say you will get things under control with less acceleration.

[PeterTheWolf]

....
.....

..... If you do a move of 20480 does it move an inch?

If the counts per inch is ok id say you will get things under control with less acceleration.

I set up a 1" travel dial indicator on the X-axis and then set the "Step Size" to 10090 and hit move. The indicator move to .5002" and then back to zero within a .0001" - .0002" repeatedly.
I tried this 4 or 5 time with the exact same results. So 10090 counts will move the X-axis table .5002"

So then with the 1" travel indicator on the X-axis I then set the "Step Size" to double that 20180 and hit move. This produced 1.009" move. I found that 20050 would move 1.0002" (exactly) repeatedly with the multiply "move" hits.

It looks like 20050 counts will move the X-axis table 1.0002"

In any case, I am obviously missing something in this tuning because I cannot get the the Position (RED) to follow on top of the Command (BLUE) with just changing the "I" Gain and/or Velocity, Acceleration or Jerk.

The Jerk seems to make the biggest difference on the Error plots, that is, if a good plot is what I have in this video (ZIP file 50mb)

Kmotion Data file with Parameters used

.....
.....

J325 Quick Links:

Machine
Existing Machine Schematics
Electronic Cabinet-Right Side
Electronic Cabinet-Back SIde
Existing Drive Board SD1525-10
J325 Servo Drive-SD1525 Manual
3-Phase Rotary Convert Used
RickB's J325 Retrofit Wiring
KFLOP 5VDC/15Watt/3A Power Supply
KANALOG Mounted & Connected
Kmotion - Axis Encoder Manual Test of Position via Manual Movement
Kanalog-Encoder Voltage High/Low Checks & 1KOhm Resister
Kmotion Configuration Screens "RUN-AWAY"

.....
.....

[TomKerekes]

Hi PeterTheWolf,

Please forgive me if I cannot quite grasp these concepts as fast as expected.

No worries. I hope I don't come off as being impatient.

It looks like 20050 counts will move the X-axis table 1.0002"

Thanks for the verification.

I think these plots are better. The errors are now about 330 vs 1000 before. As I said before matching the Error to the Output is not the goal. The goal is to get the errors reduced. If the error was all zero and the Output was all nasty that would still be great.





Try to continue to increase I gain in order to decrease the initial lag and get back on target quickly. That should reduce the errors and reduce the overshoot at the end of the move.

Regards and Merry Xmas!

[PeterTheWolf]

....
....

Ok ... I think I have the I-Gain setting as good as I can get it before it goes unstable. I kept pushing it at an increase of .0001 until it went unstable at .011

I am thinking 0.0055 may be the best I can get without too much oscillation (with a error of 120) ... even though I still see a little oscillation with this setting.
The following two videos should be clear as to what I tried and ended up with.

Not sure where to go with the next step.

Video of Kmotion settings (ZIP File 40mb)

Video of Machine Axis Movement (ZIP File 149mb)


.....
.....

J325 Quick Links:

Machine
Existing Machine Schematics
Electronic Cabinet-Right Side
Electronic Cabinet-Back SIde
Existing Drive Board SD1525-10
J325 Servo Drive-SD1525 Manual
3-Phase Rotary Convert Used
RickB's J325 Retrofit Wiring
KFLOP 5VDC/15Watt/3A Power Supply
KANALOG Mounted & Connected
Kmotion - Axis Encoder Manual Test of Position via Manual Movement
Kanalog-Encoder Voltage High/Low Checks & 1KOhm Resister
Kmotion Configuration Screens "RUN-AWAY"

.....
.....

[TomKerekes]

Hi PeterTheWolf,

I'd agree and say around 0.005 is probably a good balance between stability and following error.



Note minor tip: I noticed in the video you mentioned increasing gains by fixed amounts such as 0.0001. Better would be to increase by a certain factor. ie double. Or 30%. Also you might round the numbers to simple values. For example where you had 0.00997 you may as well of had 0.01. If you see a difference when making such a small percentage change it is likely due to something else.

With those settings the error is ~120 (6mils) peak when accelerating and decelerating. This seems somewhat large for this moderate speed (~60ipm). But I don't see an easy way to get better other than by using feedforward. The problem with using feedforward is that it is open loop. We might tune things perfectly but if things change such as mass (big part on the table), friction, cutting forces, then the result will not be as good. But I'd say go ahead and tune it in as it should help. Just be aware of its limitations.

Feedforward looks at the trajectory and immediately applies the nominal expected amount of output to follow the trajectory. It is sort of like driving a car around a racetrack that you know in your mind so well you can almost drive it with your eyes closed. You know how much to turn the steering wheel immediately at the entry to a curve. Driving without feedforward is like ignoring the upcoming curve, continuing straight, and waiting until you see yourself running off the road before quickly attempting to make a correction. That's what we've been doing so far. The best result happens when feedforward and feedback work together. The feedforward does the majority of the output without any delay, and then the feedback monitors how things are going and only needs to make slight corrections.

KFLOP offers Velocity and Acceleration feedforward. The assumption is that going at a specific velocity requires a certain amount of output and acceleration/deceleration requires a certain amount of output. The amount of output required is assumed to be simply proportional to the amount of velocity or acceleration needed. The process of tuning the feedforward involves trial and error adjustments to the V and A feedforward values to find what gives the best results.

Its important to understand what parts of the trajectory (blue plot) have high acceleration/deceleration and what parts have high velocity. Acceleration causes the Trajectory to have curvature. High velocity causes the trajectory to have a steep slope.

I'm expecting you understand that currently our largest errors occur at points of high acceleration/deceleration. I'm expecting Acceleration feedforward to be most helpful to reduce those errors.

Sometimes it is helpful to reduce the feedback while tuning the feedforward. Mainly turn the I gain to near zero (so it has nearly no effect throughout the motion). This is sort of like learning to drive the racetrack with your eyes closed. If you are able to follow the track reasonably well you can be sure you have learned the track well. So then after you open your eyes you will only need to make the minimal corrections.

'V' feedforward numeric values are typically small numbers. 'A' feedforward numeric values are typically even smaller. This is because the velocity and accelerations they are multiplied by are typically large numeric values in counts/sec or counts/sec2 and the result in output is not typically that large of a number. Start with small numbers such as 0.0001. Increase until an effect is seen on the error. Then continue to increase until the error reaches a minimum. Much like making a corner in a car and anticipating a certain curve and turning the steering wheel too much will cause you to veer to the inside of the curve rather than the outside of the curve.

As a side question can you explain the mechanics of your machine? I'm assuming the motor, tachometer, and encoder are all on the same shaft? Or is the encoder on the lead screw? Also right now it seems impossible to reach the original specification of 250ips. Based on our output of 600 DAC counts to go ~60ips. Then full DAC of 2048 (10V) would only be about 180ips. I'm wondering if someone reduced the size of the motor pulley. That would increase the resolution (like we see)and reduce the speed (like we see). Is the belt at the proper tension?

HTH
Regards

[mmurray70]

Tom, is there any possibility of improving things with the adjustments on the servo drive? Or would you just end up with similar performance, and a different set of PID numbers in Kmotion after going through the tuning process again?

[TomKerekes]

Hi mmurray70,

Good question. I was reluctant to suggest that because it would be confusing and if we mess things up it would be hard to get back to where we are. I was thinking that if the machine was working ok before and no one changed anything it should be usable as is. The fact the Amplifiers can't reach the speed in the specification is concerning. That indicates something must be off mechanically or electronically. Theoretically if the amplifier has some excessive lag then it should be possible to compensate for it by adding lead in KFLOP as long as everything is linear. But there may be some subtle side effects. The Tachometer (available to the amplifier but not KFLOP) has an advantage of being theoretically infinite resolution vs the encoder which has only moderate resolution (I think it is 1024 count/rev).

It does seem like performance is significantly less then many other systems I've been involved with. It might be just the combination of the slow switching speed of the Amplifiers, the somewhat low resolution of the encoders, the belt drive, belt reduction, low leadscrew pitch (0.1 inch), etc...

As is we should be able to get around 180ips Rapids (where accuracy isn't that important) and with lower acceleration and feedrates for cutting I would expect accuracy to be reasonable. So I was thinking to just move forward as is. If we find accuracy is not acceptable we can always revisit things later. Any input welcome.

Regards

[PeterTheWolf]

I'd agree and say around 0.005 is probably a good balance between stability and following error.

Its important to understand what parts of the trajectory (blue plot) have high acceleration/deceleration and what parts have high velocity. Acceleration causes the Trajectory to have curvature. High velocity causes the trajectory to have a steep slope.
I'm expecting you understand that currently our largest errors occur at points of high acceleration/deceleration. I'm expecting Acceleration feedforward to be most helpful to reduce those errors.

Sometimes it is helpful to reduce the feedback while tuning the feedforward. Mainly turn the I gain to near zero (so it has nearly no effect throughout the motion). This is sort of like learning to drive the racetrack with your eyes closed. If you are able to follow the track reasonably well you can be sure you have learned the track well. So then after you open your eyes you will only need to make the minimal corrections.

'V' feedforward numeric values are typically small numbers. 'A' feedforward numeric values are typically even smaller. This is because the velocity and accelerations they are multiplied by are typically large numeric values in counts/sec or counts/sec2 and the result in output is not typically that large of a number. Start with small numbers such as 0.0001. Increase until an effect is seen on the error. Then continue to increase until the error reaches a minimum. Much like making a corner in a car and anticipating a certain curve and turning the steering wheel too much will cause you to veer to the inside of the curve rather than the outside of the curve.

Understood.

My Fast-Forward test steps were made.

I set set the I-Gain to close to zero, where it had no real effect ... I=0.00001
The I started out with Fast-Forward V=0.0001 & A=0.0001
I then (first) only increased the Acceleration feedforward to equalize curvature (RED on BLUE) ... error seem to correspond as well.
Then I increased the Velocity to equalize the slope (RED in BLUE) .... error seem to correspond with reduction as well.
This process seemed easier to interpret compared to just looking at the error results only.
Did I do this correctly ... using this process?

Here is the Kmotion videos of the results with I-Gain I=0.0001 (ZIP file 42mb)

Here is a video with I-Gain set back to I=0.0055 and Fast-Forward Parameters (ZIP File 29mb)

My Final Kmotion Saved Data file

As a side question can you explain the mechanics of your machine? I'm assuming the motor, tachometer, and encoder are all on the same shaft? Or is the encoder on the lead screw? Also right now it seems impossible to reach the original specification of 250ips. Based on our output of 600 DAC counts to go ~60ips. Then full DAC of 2048 (10V) would only be about 180ips. I'm wondering if someone reduced the size of the motor pulley. That would increase the resolution (like we see)and reduce the speed (like we see). Is the belt at the proper tension?

1.) Yes, the Servo motor, tachometer, and encoder are on the same shaft (on the back end).
2.) I would have to say that the servo motor pulley are original equipment (with a 2:1 ratio). I once again verified that two revolutions of the servo motor pulley produces one revolution on the Ball-Screw pulley.
So, 2000 counts = one revolution on the servo motor pulley which produces 0.100" (exactly) move on the ball-screw (table) of the X-axis.
This was tested with the "Move". With an dial-indicator set I was getting 0.100" move with it returning back to zero ... exactly (well within .0002")

3.) I also verified, once more, that a 10,000 count move produces a 0.500" move on the ball-screw (table) on the X-axis. Also, 20,000 count move produces a 1.000" move on the ball-screw (table) on the X-axis ... exactly.

4.) Concerning the Servo Motor notched belt drive tension, I will need to look into this with more detail as at this moment I am not sure how to set this. I am assuming it is set correctly.

Tom, is there any possibility of improving things with the adjustments on the servo drive? ....

I was wonder this same question .....

Hi mmurray70,

Good question. I was reluctant to suggest that because it would be confusing and if we mess things up it would be hard to get back to where we are.

More confusion .... I can relate!!!


.....
.....

J325 Quick Links:

Machine
Existing Machine Schematics
Electronic Cabinet-Right Side
Electronic Cabinet-Back SIde
Existing Drive Board SD1525-10
J325 Servo Drive-SD1525 Manual
3-Phase Rotary Convert Used
RickB's J325 Retrofit Wiring
KFLOP 5VDC/15Watt/3A Power Supply
KANALOG Mounted & Connected
Kmotion - Axis Encoder Manual Test of Position via Manual Movement
Kanalog-Encoder Voltage High/Low Checks & 1KOhm Resister
Kmotion Configuration Screens "RUN-AWAY"
Tree Journeyman 325 Designed Specs.


.....
.....

[mmurray70]

Well that looks a little better. Save those numbers as a baseline. Probably be good enough as it is.

Yeah i bet the reduction and fine ballscrews is really killing acceleration. Those servos need to spin like crazy to move that table! My machine is a little bigger but uses 10mm ballscrews with direct drive. This one in comparison would need to accelerate the servo motor 8 times faster for the same table acceleration. If you look at the video of the machine, the servo does seem to be starting and stopping fairly quickly. Even though the settings for the axis dont sound too bad at first glance, maybe its still a little too much considering the servo is geared so low by the belt and fine pitch ballscrew?

I wonder is it worth trying with even lower acceleration still before adding the Feed Forward? Maybe try 100000 or 150000. If you do try this, remove the Feed Forward and do a test with current settings, then reduce acceleration even further and see if theres an improvement. If its the same then forget about this test, if its a little better then tweak the Feed Forward again for the lower acceleration.

[mmurray70]

the encoder which has only moderate resolution (I think it is 1024 count/rev).

It does seem like performance is significantly less then many other systems I've been involved with. It might be just the combination of the slow switching speed of the Amplifiers, the somewhat low resolution of the encoders, the belt drive, belt reduction, low leadscrew pitch (0.1 inch), etc...

Understood.

2.) I would have to say that the servo motor pulley are original equipment (with a 2:1 ratio). I once again verified that two revolutions of the servo motor pulley produces one revolution on the Ball-Screw pulley.
So, 2000 counts = one revolution on the servo motor pulley which produces 0.100" (exactly) move on the ball-screw (table) of the X-axis.
This was tested with the "Move". With an dial-indicator set I was getting 0.100" move with it returning back to zero ... exactly (well within .0002")

3.) I also verified, once more, that a 10,000 count move produces a 0.500" move on the ball-screw (table) on the X-axis. Also, 20,000 count move produces a 1.000" move on the ball-screw (table) on the X-axis ... exactly.

My last post was based on Toms post of 1024 encoders and 0.1 pitch ballscrews. Thats why I was thinking the servos must be spinning so fast!

After a closer look at your post Peter, it looks like the encoders are actually 2000 (or 2048) and ballscrew pitch is 0.200"? Your move of 2000 counts moved the Servo pulley 1 full turn right, not the leadscrew pulley, correct? This would make more sense. If this is the case, there might not be anything to gain by dropping acceleration even further.

[PeterTheWolf]

....
....

Just to confirm:

After a closer look at your post Peter, it looks like the encoders are actually 2000 (or 2048) and ballscrew pitch is 0.200"?

Yes ...this Tree Journeyman 325 has a ball-screw Pitch on .200" (per one revaluation) on all axis.

Your move of 2000 counts moved the Servo pulley 1 full turn right, not the leadscrew pulley, correct?

Yes... that is correct. 2000 counts on the servo motor axis (pulley) will move the ball-screw axis One-Half of a revaluation, which is 100".

So, the Servo Motor Axis moving 2 revaluation to a count of 4000 will rotate the Ball-Screw 1 revaluation, which is .200" (on a dial-indicator).

I trust this clears up any confusion on the mechanical ratio of the pulley ratio on this machine.

.....
.....

J325 Quick Links:

Machine
Existing Machine Schematics
Electronic Cabinet-Right Side
Electronic Cabinet-Back SIde
Existing Drive Board SD1525-10
J325 Servo Drive-SD1525 Manual
3-Phase Rotary Convert Used
RickB's J325 Retrofit Wiring
KFLOP 5VDC/15Watt/3A Power Supply
KANALOG Mounted & Connected
Kmotion - Axis Encoder Manual Test of Position via Manual Movement
Kanalog-Encoder Voltage High/Low Checks & 1KOhm Resister
Kmotion Configuration Screens "RUN-AWAY"
Tree Journeyman 325 Designed Specs."


.....
.....

[TomKerekes]

Hi PeterTheWolf,

I think you did an excellent job with the Feedforward tuning. You should realize the maximum +/-6 counts of error only corresponds to +/- 0.0003 inches. To get better than that is probably not significant considering all the other mechanical errors. Furthermore in a different scenario the errors are likely to be significantly different. It like teaching a blind robot to make a corner in your car absolutely perfectly, that's great but then next week, on a different curve, with different road conditions, and the trunk full of bricks, the result is not likely to be as good. We are only testing one move size, at one speed, at one place in the travel, with no payload. As an experiment you might try clamping a big part on the table to check the effect on the error.

btw you might also look at the plot and observe the dither. Notice the errors are ironically actually somewhat worse when sitting still as compared with while moving at high speed. It is very difficult to hold near zero error at zero speed because of stiction, backlash, and compliance. You might imagine the motor stuck at a position 0.0003 inches from the destination and the servo trying to correct this tiny error. Finally the servo applies enough motor torque to break free, but then it overshoots. I wouldn't worry about this at this point. Here is a related article:
Dynomotion

I'd suggest the next step to find two separate Motion Profile parameters for Rapids and Feeds. You might read this article:
Dynomotion

For rapids we just want to move from point A to point B as quickly as possible. We don't care so much about accuracy but we don't want a large amount of oscillation and shock at the end as we may be cutting at feedrate soon afterwards. I'm thinking you should be able to achieve 180ipm which is the limit of the amplifiers with their current settings. Rapids make use of the Jerk parameter. Jerk is like limiting the speed that you apply the gas or brakes in a car.

For feeds we likely want to use lower accelerations because accuracy is important while moving/cutting. It is up to you to decide what is the fastest feedrate you will be using for cutting and what error is acceptable. Reduce the acceleration until that criteria is achieved. Always keep the Jerk setting at 1000X the acceleration value to simulate the no-Jerk limit that will occur while feeding.

You might then also repeat all the previous steps for the other two axes. At that point we should be ready to move to KMotionCNC to actually run a GCode Job.

Regards

[PeterTheWolf]

...
....

I'd suggest the next step to find two separate Motion Profile parameters for Rapids and Feeds. You might read this article:
Dynomotion

Sorry, I just want to confirm my understanding on this next step.

I have read and re-read what you advised on ... "Velocity, Acceleration, and Jerk".

When you say I will need to come up with two separate motion profile parameters.... one for Rapid (moves) and one for Feed (moves) ....


Now I am going start testing different "Motion Profile" values (V, A, & J) according to these four step below to find two sets of values used in KMotionCNC based on the tuned servo values ("P", "I", & "D") I have already found for the X-axis?


1 - Find Max Velocity at low Acceleration and infinite Jerk
2 - Find Max Acceleration for your system for the chosen Velocity
3 - Find optimal 3rd order Jerk limited motions.
4 - Reduce Max Following Error Limit to small value


.....
....

[TomKerekes]

Hi PeterTheWolf,

That is basically correct. The idea is to find two sets of motion profile parameters (V A J). One set used for accurate cutting and one set for Rapids (point to point moves).


Set #1 - accurate cutting

You will need to choose some criteria regarding maximum cutting speed and precision. You can then find the maximum Acceleration that will allow you to achieve those goals. For example you might decide you will never need to do accurate cutting over 30ipm and you are ok with 1mil precision. Convert the 30ipm to counts/sec and set that as your V to test. Convert the 1mil precision to counts so you know what to look for in the plots. Now try different Accelerations to find the level that meets your precision goal. With lower Acceleration the system should be able to follow the trajectory more precisely. Each time you change the Acceleration set the Jerk to 1000X the Acceleration number. This simulates switching the Acceleration on and off instantly (in 1 millisecond) similar to what will occur when cutting.

When you are finished record the Acceleration and convert it to inches/sec^2 (divide by the resolution) to be entered later into KMotionCNC's Trajectory Planner.

Discard the Jerk Value

The Velocity setting might also be converted to inches/sec and entered into KMotionCNC's Trajectory Planner. Or you might enter a bigger value if you want to allow higher GCode Feedrates that might be used when you don't care so much about accuracy.

Set #2 - Rapids

For this set we probably don't care much about accuracy. The criteria is more that we can reasonably follow the trajectory without falling grossly behind and that we don't rock, shock, or shake the machine excessively. From previous tests it seems your system should be able to move 180ipm so you might choose that as your max rapid speed. Then find Acceleration and Jerk settings you are happy with. Jerk relates to the rate the acceleration is applied. Its like how fast the brake is applied in a car. A reasonable value for most machines is around 0.1 seconds. By always setting the Jerk Value to 10X the Acceleration value full acceleration will be applied in 0.1 seconds. Make sure that you test moves big enough (size) that both full acceleration and velocity is achieved. The Velocity plot is good for checking this. Obviously the max Velocity in the plot should be the same as your max V setting. The Velocity plot should also have a straight ramping section. This represents the constant and maximum acceleration period. If the plot consists entirely of curves then we never ramped to maximum acceleration before it was time to start ramping acceleration back down.

When finished leave these settings so they will be used for rapids and also any internal KFLOP motions such as Homing and such.

At this point you should know the worst case following error that occurs under all normal circumstances. Set the Max Following error to a slightly bigger value. By doing this if anything goes wrong for example you hit an obstacle, servo goes unstable, inadvertently commanded to too high of a speed, etc. then the Axis will immediately fault and disable.

HTH
Regards

[PeterTheWolf]

....
..... We are only testing one move size, at one speed, at one place in the travel, with no payload. As an experiment you might try clamping a big part on the table to check the effect on the error.

Today I loaded the X-axis Table up with about 140lbs of steel and tested the Error base on the current final working X-axis servo "Step Response" Values (P,I,D, V, A, J,) over a the 10,000 count move.
The Error plot and Command/Position Plots look the same as the plots without the load. However, I did see less dithering at the pulleys (physically oscillating) but the error plots still is hunting at zero.
I also noticed with a load on the table when I move with a count of 10,000 I no longer get an exact move of .500" on my dial indicator. It moves .493" and then back to zero.
Can I assume that this will be addressed later?

....
I'm wondering if someone reduced the size of the motor pulley. That would increase the resolution (like we see)and reduce the speed (like we see). Is the belt at the proper tension?

I was able to barrow a Gates Sonic Tension Meter (model 508C) to check the tension on all three servo belts.
Based on the information I found the X-Axis and Y-axis checked within a couple of Hz; however Z-axis needed adjusting.

Span Length Formula:
S = √(CD^2- (D-d)^2/4)
Where: S=Calculated Span Length (mm)
CD=Center Distance (mm)
D=Large Pulley Dia. (mm)
d=Small Pulley Dia. (mm)


X-Axis:
Brand Belt: Bando Synchrolink
Number on Belt: 270L-0756 (SB-26)
PowerGrip Timing Belt – L (0.375”)
Mass Constant = 3.2 g/m
Belt Width = .750”
Frequency on Correctly Tensioned Belt: 52Hz
Speed Ratio =0.500
Pitch Diameter of the Large Pulley (26 tooth) = Ø3.104”
Pitch Diameter of the Small Pulley (13 tooth) = Ø1.552”
Center to Center axis span (used in formula above) = 9.8125”
Checked Belt Tension was at a Frequency of 50-51 Hz.

Y-Axis:
Brand Belt: Bando Synchrolink
Number on Belt: 322L-0756 (0949)
Mass Constant = 3.2 g/m
Belt Width = .750”
Frequency on Correctly Tensioned Belt: 52Hz
Speed Ratio =0.500
Pitch Diameter of the Large Pulley (30 tooth) = Ø3.581”
Pitch Diameter of the Small Pulley (15 tooth) = Ø1.790”
Center to Center axis span (used in formula above) = 11.8125”
Checked Belt Tension was at a Frequency of 52-53 Hz.

Z-Axis:
Brand Belt: Bando Synchrolink
Number on Belt: 225L-100G
Mass Constant = 3.2 g/m
Belt Width = 1.00”
Frequency on Correctly Tensioned Belt: 128Hz
Speed Ratio =0.400
Pitch Diameter of the Large Pulley (30 tooth) = Ø3.581”
Pitch Diameter of the Small Pulley (12 tooth) = Ø1.432”
Center to Center axis span (used in formula above) = 7.250”
Checked Belt Tension was at a Frequency of 128-129 Hz.


Microsoft Word Doc of the above Belt Frequency Settings

Please advise if errors exist.

.....
.....

J325 Quick Links:

Machine
Existing Machine Schematics
Electronic Cabinet-Right Side
Electronic Cabinet-Back SIde
Existing Drive Board SD1525-10
J325 Servo Drive-SD1525 Manual
3-Phase Rotary Convert Used
RickB's J325 Retrofit Wiring
KFLOP 5VDC/15Watt/3A Power Supply
KANALOG Mounted & Connected
Kmotion - Axis Encoder Manual Test of Position via Manual Movement
Kanalog-Encoder Voltage High/Low Checks & 1KOhm Resister
Kmotion Configuration Screens "RUN-AWAY"
Tree Journeyman 325 Designed Specs."
Tree Journeyman 325 Axis-Tension Frequency Settings"

.....
.....