Set up using AMC AB15A100 drives, brushed DC motor Prototrak Plus with Encoders

More
17 Jun 2022 13:59 - 17 Jun 2022 14:05 #245322 by new2linux
Todd, Thank you! I have tried several P values, attached are the example of just P & FF1 & DEADBAND = .00001. The file name has settings. Trying to use FF1 to move the traces closer together going from .497 to .05 are attached pic. If I increase the P it starts to osculate or increase the FF1. I have had the exact same set of variables & have 1 time I will get a trace & next time the osculation will start & trace is not usable. One of the pics of the traces has no FF1 at all.

units:
The axis.0.joint-vel-cmd = 100m/div
axis 0 f-error = 5m/div
pid.x.output = 200m/div
hm2_5i25.0.encoder.00.velocity = 200m/div

Referencing the oscilloscope approach. I am going to reach out & see if my buddy's father has a scope with all needed accessories and time. This can be preformed off the machine? It would be bonus if he has a signal generator as well?

Many thanks!!
Attachments:
Last edit: 17 Jun 2022 14:05 by new2linux.

Please Log in or Create an account to join the conversation.

More
17 Jun 2022 14:39 #245328 by new2linux
Todd, this set of pic with only P changing (same units as prior post) & slowing down the feed rate so it would no osculation or having the motor jerk/serge.
Many, many thanks!
Attachments:

Please Log in or Create an account to join the conversation.

More
19 Jun 2022 12:09 #245426 by Todd Zuercher
In halscope adjust the vertical gain slider for the encoder velocity and the velocity command have the same scale. Right now you have one set to 100m/dev and the other to 200. Making them look closer than they are.
The following user(s) said Thank You: new2linux

Please Log in or Create an account to join the conversation.

More
19 Jun 2022 17:19 #245439 by new2linux
Todd, Many thanks for your help. I tried the exact same settings (P, DBAND) & feed-rate, with the units uniform on the attached pics, but was unable to repeat as prior post. I did not do extensive/exhaustive selection, just slowed the feed rate..

Many thanks
 
Attachments:

Please Log in or Create an account to join the conversation.

More
22 Jun 2022 12:18 #245590 by new2linux
Todd, and all the people that make this forum so informative, many thanks! In addition many thanks to all the great help from A-M-C, the engineering staff as well as the product, excellent all the way through.

I reached out to A-M-C engineering, there suggestion follows:
"If there are velocity issues, I would adjust the velocity loop integral gain. This will reduce the response time as you increase the capacitance, though will help in eliminating any velocity following errors.
You could do this in the same manner by trying each switch configuration one at a time."

This is as is: Currently, POT 1 is 5.5 turns CW: Going back to when this was set, the suggestion was to adjust POT 1, CW until motor squeals, back up (CCW) 3 turns. I will track change to see if this changes. If this does not seem to move the halscope trace then try the SW3 (7, 8, & 9 ) switches.

Many thanks.

Please Log in or Create an account to join the conversation.

More
22 Jun 2022 15:52 - 22 Jun 2022 19:11 #245603 by new2linux
Many thanks, for all the great help. I have a pic showing SW3-7, SW3-8, SW3-9 with the best trace, one change at a time, units are uniform. The sw3, does not appear to make much of a difference. When tuning sw3-9 it was more difficult than -8 or -7, to keep the osculations low or motor axes joint error. The white trace does not seem to get closer to the red trace.
Many thanks

Edit: In the beginning of the file name, a "sw" is the switch with numbers after the sw 7 = dip switch 7: 7 8 = dip switch 7 & 8; 7 8 9 = dip switch 7 & 8 & 9. RefFactory file has the original AMC setting.
Attachments:
Last edit: 22 Jun 2022 19:11 by new2linux.

Please Log in or Create an account to join the conversation.

More
27 Jun 2022 18:09 - 27 Jun 2022 18:22 #246016 by new2linux
Many thanks for all the help! I have reviewed the suggestions from A-M-C & members of the forum, including tommylight's tuning tutorial as I understand them. I am listing the approach I have tried with the sw3-7,8 & 9, in the off (factory) & on settings. Starting with sw3-7 add 8, & eventually 9 (all on), going through the complete tuning process for each, starting low, watch as the traces move in response to the adjustment. None would allow the white trace (because channel 4 is selected, normally purple) "hm2_5i25.0.encoder.00.velocity" to get close to the red trace "axis.0.joint-vel-cmd", with 1 exception, to lower the feed rate to 52"/min. I did not try to improve the trace at 52", unless this is the only way. Using the procedure below, have I approached this correctly?? Power was turned off for sw3 changes, this took several days for all the different settings.

Starting (POT 2 is set as per motor requirements & POT 4= mid-point) with (all settings=0.0, except FF1=1.0) POT 1 (factory is full CCW) turn CW until motor squeals, back up 3 turns & POT 3 (factory is full CW) turn CCW until motor firms up, this is used to get as close as possible. This can be tweaked after P, & FF1&2 are ruffed in, then set as good as possible.
When adjusting P & FF1 & 2, starting with P until motor oscillation, work on say the FF1, until can't get any better, then go back to P & try this a bit higher, then FF1&2.

It is important to have all the settings in both (x&y) axes in the "calibration screen" set to 0.0 before settings any of the "sw3" settings or any of the, "hal scope P, I, D, FF1, FF2, DEADBAND" settings are adjusted. The attached pic has the "calibration screen" showing the "Y" axes tune 1, (with additional lines that are not on the "X" axis (tune 0)) could this cause tuning issues? This is in the .hal file & I use the "#"?

Attached is 1 pic of the best P, FF1 & FF2, in each sw3-7,8&9 settings. All units are set identical, as I flip through the around 60 different (use arrow key to view 1 after the next) screenshots, with the same units it is difficult to see the change! Now, it very well could be I don't know what to look for. My efforts were to get the white trace as close to the red trace as possible, but I don't think it's very close. The file name starts: 7=sw3-7: 78=sw3-7&8; 789=sw3-7,8,&9, The "Y" CalibrationScreen.png shows tune 1 (Y axis), could this be the reason why the tuning is difficult?
Any thoughts are warmly welcomed, I will link this post to the A-M-C engineer. I am praying for a solution. Many thanks
Attachments:
Last edit: 27 Jun 2022 18:22 by new2linux.

Please Log in or Create an account to join the conversation.

More
27 Jun 2022 20:05 - 27 Jun 2022 20:10 #246021 by Todd Zuercher
Tuning in Linuxcnc is difficult because the drive isn't tuned properly yet.  Tuning the drive is difficult because you keep trying to do it with variable command signals.

While you are tuning the drive you need to be sending it a consistant constant signal. That is why you should be using an open loop command signal (all of Linuxcnc's PID settings =0 except FF1=1) while you are changing the settings in the drive.

Also you need to be consistent with your trace gain scaling to try to compare apples to apples. Your green trace is the signal that is being sent to the drive via the analog +/-10v signal. How that signal to the drive is scaled is in your ini file by a combination of the "MAX_VELOCITY =" and "OUTPUT_SCALE =" settings. In your case 120ipm. The problem is when Linuxcnc is commanding near the full speed of 10v your drive is only moving the motor about 50ipm (0.84inches/sec). But this is a little hard to tell in your Halscope images because you are not setting the gain scaling the same for the different velocity commands and feedbacks that you are reading with it.

Adjust your gains for the Halscope signals so that 1inch/sec is the same number of ticks for each signal you are reading.

Then you need to work at adjusting the drive settings so that a 60ipm commanded move, actually moves 60ipm (or 1ips) this will be a 5v analog command to the drive.
Last edit: 27 Jun 2022 20:10 by Todd Zuercher.
The following user(s) said Thank You: Clive S, new2linux

Please Log in or Create an account to join the conversation.

More
28 Jun 2022 12:48 - 28 Jun 2022 12:49 #246075 by Todd Zuercher
Maybe I should clarify, Your latest Halscope images I can't read the scale factor for most of the traces, so I can't say if the scaling is now correct or not. In post #245143 the gain scaling for the pid command out and encoder vel feed back were very wrong.
forum.linuxcnc.org/12-milling/44454-set-...ers?start=120#245143
So when you tuned the drive to match those 2 traces you limited the maximum velocity that the drive could be commanded to do to about 50ipm.

You need to go back and redo that step with the trace scaling corrected. Also this is when you should be trying to adjust the other drive settings to get it to respond as best as possible.

Then and only then can you proceed to tuning the other Linuxcnc PID settings, But the FF1 setting should remain 1 (or very close to it.)
Last edit: 28 Jun 2022 12:49 by Todd Zuercher.
The following user(s) said Thank You: new2linux

Please Log in or Create an account to join the conversation.

More
28 Jun 2022 13:36 #246076 by new2linux
Todd, many thanks! I am careful about the units being all exactly the same, at 200m/unit.
As I start to tune drive to its self, I will follow this method. The POT 2 settings should be set to motor requirements, POT 3 set to factory (full CW), POT 4 set to factory (mid-point) only POT 1 starting from factory settings (full CCW) turning CW, to get the best trace. Best trace looks like the smallest of over-shoot, with the flatting out in the "cruse" portion of trace w/o oscillation. Watching the green trace to purple (white, because it's been selected) trace. Using this method each time, as the way to try each of the sw3-7,8 & 9.

When I get this part correct, then start with P, FF1 & FF2 starting with DEADBAND set to .000001. Then the I & D setting can be adjusted.

Many thanks!!

Please Log in or Create an account to join the conversation.

Moderators: piasdom
Time to create page: 0.248 seconds
Powered by Kunena Forum