A-M-C A12A100 Servo Drive with SouthWestern Ind ProtoTrak Plus

More
21 Feb 2021 16:21 - 22 Feb 2021 18:17 #199635 by new2linux
PCW, Many thanks for all the help. I have attached a volt meter to the motor leads at the driver, with the power unplugged from wall and turning the hand-wheel of the x axis, the faster you turn the higher the meter reads, depending on what direction meter would read + 23 volts to - 23 volts.
With e stop released, linuxcnc is off, meter reads -105 (+- floats around, depending on direction of table) but will go in to positive range when hand-wheel is moved.

all help warmly welcomed.
many thanks

Edit: I have attached volt meter to the "power in" (P1) from power supply, of driver and used arrow key to move table to see if voltage drop when load is applied and shows very small, like 0.4 change from the 38 volt range. Seems to have good resistance to motion at low feed rate, naturally depending on what settings are used when tuning. Many thanks
Last edit: 22 Feb 2021 18:17 by new2linux. Reason: 2nd test

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

More
22 Feb 2021 20:26 #199762 by Todd Zuercher
I don't think those tests tell you anything really useful.
The following user(s) said Thank You: new2linux

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

More
22 Feb 2021 20:50 #199764 by new2linux
Todd, Many thanks! I tried to tune the Y axis over the weekend, and the very similar looking hal trace appeared. As with tuning of both axis, when tuning, clicking the "test button" after every change, I could go back to the exact same settings and have great difficulty getting to repeat exactly, a lot of the time it would error out. Or needed to try the arrow key several times before the feed would work. I thought it was key board, swapped out key board, no change. Several times started feed rate low and worked back up. Many thanks for the help!!

Many thanks!

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

More
23 Feb 2021 14:48 #199834 by Todd Zuercher
That is the problem with trying to use torque command input with a PID. The FF1 setting is very speed dependent. What works well at a low speed will be way way too much a higher speed (enough to cause following errors.) In a velocity command system the command required to move increases linearly with the speed, but in a torque commanded system it is anything but linear. In fact it is even possible that the amount of command input needed to maintain a low feed rate might be more than is needed at a higher feed. For the most part, at least theoretically, once you've overcome friction the amount of force (torque) needed to maintain an achieved velocity is zero. In practice it still usually takes more electrical force (current) to spin the motor faster. That is where I thought that limiting the FF1 by setting a value for "pid.N.maxcmdD" might be useful.
If you want to add that as a tunable item in the Axis Calibration window, add these lines to your hal and ini files
In the Hal file in the x axis section add:
setp   pid.N.maxcmdD       [AXIS_0]FF1LIMIT
and in the ini file in the [AXIS_0] section add this line
FF1LIMIT = 10

After restarting Linuxcnc you should see the new line in the Calibration window for FF1LIMIT with an initial value of 10. I really have no idea what might be a good starting value. start with a higher value, then with the FF1 working well at a low speed lower the ff1limit incrementally until you start to see some effect from it. Then set it to the smallest value that doesn't do much. Then test to see what happens when you increase the velocity of your moves. If this works/helps, then you can make the same changes for the other axis.

I really have no idea what this is going to do, it may not work at all. I have never tried it myself. I gave up trying to make a single PID loop work with a torque command on my machine years ago. (But I hadn't thought of trying this back then and I was much greener at setting up Linuxcnc.)
The following user(s) said Thank You: new2linux

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

More
23 Feb 2021 16:49 - 23 Feb 2021 18:14 #199855 by new2linux
Todd, Thanks, Made changes as per your suggestion. looks good, see pic.
Attachments:
Last edit: 23 Feb 2021 18:14 by new2linux. Reason: add pic

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

More
23 Feb 2021 19:01 - 23 Feb 2021 19:07 #199877 by new2linux
Todd, Many, many thanks! Attached are some pic of the most current changes, look good 29"/min, The file name has settings, if pic is not clear. Only most recent pic are attached. many thanks!!

Edit: Todd: This is from prior post: "FF1 working well at a low speed lower the ff1limit incrementally until you start to see some effect from it."

The effect I am looking for, is the making the "axis 0 f erreo" trace, horizontal, strait and as small as possible, correct? many thanks.
Attachments:
Last edit: 23 Feb 2021 19:07 by new2linux. Reason: Ask question

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

More
23 Feb 2021 19:41 #199884 by new2linux
Todd, many thanks. This is with "ff1 limit" from 10 reduced to around 7. Just looking for 2nd opinion. many thanks!!
Attachments:

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

More
23 Feb 2021 19:59 #199888 by Todd Zuercher
You still don't have enough FF1 or FF2. Try setting the FF1=1.3, FF2=0.2 and leave the FF1limit =10. Then record traces of moves at 10ipm, 20ipm and 30ipm. So we can see how the different speeds are effecting the f-error for a base line.

Then testing at 10ipm, lower the FF1limit until you start to see the f-error increasing. Reset the FF1limit to approximately the smallest value that doesn't increase the f-error. Then record traces at each speed, 10, 20, 30ipm (with all the settings the same as the last one.)
The following user(s) said Thank You: new2linux

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

More
23 Feb 2021 20:49 - 23 Feb 2021 20:56 #199889 by new2linux
Todd, Many thanks!! This ia a selection of traces as per your suggestions: They are in ruff order as to follow your directions. File name may not be complete.
Last edit: 23 Feb 2021 20:56 by new2linux.

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

More
23 Feb 2021 22:50 #199898 by new2linux
Todd, Many, many thanks for your help!! The attached files are the latest and look the best so far, feed at 23.8"/in, watch the units! I have other pics, these show small change in FF1 that will move the f error below "as to invert". I plan to work this tomorrow, any suggestions are warmly welcomed. Many thanks!
Attachments:

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

Moderators: piasdom
Time to create page: 0.807 seconds
Powered by Kunena Forum