Advanced Search

First picture, the black servo positioning something or another is what we today call THC, and all that (but much faster) today fits in half a cigarette box! I would get rid of this so fast....
The attached picture with 4 boards, the left one is the power supply for the next 3, those 3 are the servo drives with lots of electronics, so i would really try to save this, do not worry about the broken encoder connectors, encoders should now be wired to Mesa board.
The box with PLC, i would also get rid of this, fast, all that is now done in software and through a Mesa board.
I would suggest getting a Mesa 7i96S with 7i85 and 7i83 and THCAD for now as this will be good for analog (existing drives) and new drives with analog or step/dir controls.
There are cheaper combos or single board solutions like Mesa 7i97T that everything for analog servos and encoder feedback, or 7i92TM with 7i77 that is the same things functionally.
For new dives with no feedback to LinuxCNC, 7i96S should be enough and you can add 7i85 for encoder feedback at any time later, or single board solution with encoder feedback = Mesa 7i95

P value for stepper systems should never be changed, and it depends only on servo period timing, so for a servo period of 1000000 (1 million) it always should be 1000.
Servo period should also never be changed, except in some situations, like when using torque mode servo systems that benefit from faster iteration, or when using a PC with very bad latency. And whenever servo period is changed, so should the PID values.
The math there is 1/servo period in nano seconds, so for
1000000 servo period = 1000 P value
500000 servo period = 2000 P value
2000000 servo period = 500 P value
That is all for steppers in position mode, for servos and steppers with encoder feedback to LinuxCNC those values must be tuned.

Pncconf initially sets the P value for stepgens to 1/servo_period in seconds
(so 1000 for a 1 ms servo thread)

The physical meaning of this is that any position error (due to differences
between LinuxCNCs and the FPGA card clock, jitter, and the time
between reading the stepgen position and updating the velocity register)
is corrected by the next thread invocation.

A P value higher than 1/servo_period will overcorrect, and a a value higher
than 2/servo_period will result in oscillations.



 

Great news! It works now!, that P value was definitely the issue. I set my P value to 1000 as advised and the machine operate perfectly now!,

Is the P value something i calculate based on my jitter servo value or is that just a fixed value of 1000? 

For example, if my servo thread jitter reads 15586, would that mean my P term should be 15.586? (15586 divided by 1000?) 


I appreciate all you guys help so much , you are amazing and i apologize for being a noob. I'm doing my best to learn and im getting better everyday, thank you all so much!

Now i have to get a few things polished up and tune it all in. I am so excited!!

I'd love to know why the PNFconfig generates that as the P value? All the DIY's i followed never mentioned that P value was something i had to modify/tune.

OK the reason for the stepgen instability is the the P terms are all set to 10000
but should be 1000 for a 1 ms servo thread.

Hmm that's odd ; I uploaded it but it seems it didn't load or attached, apologizes hopefully this time it works .

The main hal file: plasma_v2.hal
has not been attached.

Yes of course thank you so much for your help!

see attached 

Thank you for clearing that up, although should have been at the very first post.
Your servo period set at 25597 was what sent me in that direction, that should not be edited normally.

Can you post your hal/ini files?
 

I used the same exact values for my AXIS setup. I didn't randomly guess them, they operate well on AXIS. My whole machine worked with GRBL for 2-3years straight no issues, i just couldn't achieve THC with GRBL & would sometimes have issues with EMI.

Setup:
7i96 with THCAD-10
Nema 23 495oz motors(1.8degree , 3amps max) with
TB6600 drivers(running 1/4 microstepping at 3amps max) (i will upgrade them later after i sort this out), two 36v supplies(10amp each),
SFU1605 Ballscrew on all axis'. 

(It was a somewhat budget build)

My stepper scale is perfect at 2032, max velocity of 150in/min(~3800mm/min) and max accel of 50inch/sec^2 (~1270mm/sec^2).

I've played around and purposely made the machine loses steps and it happens around 200 inches/min, 180 it stabilizes and 150 is what i chose as extra margin of safety.


I only used chatgpt for coding questions or how to edit bios etc, my master's degree is in AI (i'll be done in May 2025); Ironically I'm struggling to figure this one out..lol , but i will not give up ! Even if l have to read the whole entire LinuxCNC PDF and QTplasma documents or learn how the code itself works, i will do it!

 

 

mal136 wrote:

...the machine moved but the stepper motors just kinda danced back and forth they didnt actually move the gantry at all. 

Any advice on why that is occuring?

Probably because the scaling, velocities and accelerations are not set properly.
My advice: Stop using chatGPT, it can not help with these types of issues, yet, set the values properly in PncConf, not just random values. Read about how to get those values right based on what you actually have on your machine.
Or take some pictures of the machine and upload them here, we can figure out a lot from pictures.

I changed the "servo period" to 1000000 and i re-did the config using PNFconfig and the machine moved but the stepper motors just kinda danced back and forth they didnt actually move the gantry at all. 

Any advice on why that is occuring?

My jitter went up a lot for some reason and was measuring 46060 for servo thread and 44293 for base thread.

Do i have a hardware issue with my computer ? I'm reading through the linxcnc documents and QTplasma docx as we speak but it doesnt really dive into some of these issues so im a bit lost on what i can do to resolve this.

I'm sorry to be a burden and i hope it doesnt come off as annoying but im really trying my best here , dedicating hours of trial and error and endless chatgpt questions to learn.

Thank you again everyone for your help, this is my first linuxcnc machine.

I'm genuinely confused how AXIS works perfectly fine and QTplasma doesn't operate as it should, any idea why?

PCW wrote:

Unexpected realtime delay on task 0 with period 25597

Is this a parallel port setup? (25597 sounds like a parallel port base thread time)

EDIT I see this is a 7I96S configuration but the 25597 thread period makes no sense
in that setup.

 

PCW already mentioned the issue, you have to fix it in your ini file, set the "servo period" to 1000000 ( 1 million), save and start the config again.

It is a Mesa 7i96 via ethernet cable, i'm not sure what is consider normal for that value but AXIS worked fine so im confused how it affects QTplasma.

very odd