stepper runs better in Stepconf than in Axis

More
06 Dec 2012 20:56 #27302 by BigJohnT

Did you isolate one core with isolcpus?

wiki.linuxcnc.org/cgi-bin/wiki.pl?The_Is..._Parameter_And_GRUB2

John


No, but the pentium 4 is a single core isn't it? Though I never checked if after disabling Hyperthreading in the bios only one cpu showed in /proc/cpuinfo.
I'll check that later.

regards, Peter


I think your right, it's so hard to remember all the different ones... and yes it was a WAG.

John

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

More
06 Dec 2012 21:03 #27303 by BigJohnT

Quickly looking at your .ini file, what you really need to do is tune your system.
You have a default value of 100,000 for the base thread which does no fit with your reported max jitter of 26,000 and does not mesh well with your very high acceleration values and quite high velocity values.

Your driver timings for a M542 as per your stepconf file are all wrong (too high) save one , see
wiki.linuxcnc.org/cgi-bin/wiki.pl?Stepper_Drive_Timing

Then read
wiki.linuxcnc.org/cgi-bin/wiki.pl?Tweaki...ftwareStepGeneration

regards


I think Stepconf calculates the highest base period possible based on velocity of the fastest axis or something like that. I wonder if the "test this axis" uses that same number?

John

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

More
07 Dec 2012 17:44 #27378 by peter76
Hello, thanks for all the replies!

I double checked and /proc/cpuinfo lists only one cpu; good.

@ArcEye, thanks for your faq, had overlooked that till now. On e thing I'll definetely try is running 8.04 instead of 10.04 on the pc; had a spare at home for testing and that gave much nice results with 8.04.
As fo the Base period in the ini file, I had Stepconf wizard do everything for me and haven't had a proper look around there ( shame :-).
I'll try setting the base period with the calculations mentioned in the wiki.
As fo the driver time settings; I started with the settings from the wiki, but after being on the phone with the company who sold the drivers / steppers to me, I increased the a bit to see if that would solve the problem.
The settings mentioned in the wiki are the minimum timings for the driver to respond as per the datasheet; the company advised me to err on the save side and increase the values a bit. Any thoughts on this?
The velocity and acceleration I have now is what I would like for the plasma; I don't need torque, but speed. Or am I missing something here.
Like mentioned it works fine now, also in Axis with these settings, as long as I run the cpu hog.
I'll post with updates when I have 8.04 running. BTW, would it also be possible to use the kernel from 8.04 with 10.04, or would that break a lot? Or compile a kernel in 10.04 without smp support?

Regards, Peter

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

More
07 Dec 2012 20:34 - 08 Dec 2012 23:31 #27381 by ArcEye

One thing I'll definitely try is running 8.04 instead of 10.04 on the pc; had a spare at home for testing and that gave much nice results with 8.04.
BTW, would it also be possible to use the kernel from 8.04 with 10.04, or would that break a lot? Or compile a kernel in 10.04 without smp support?


Well worth using 8.04, no you can't swap kernels, I have compiled non SMP kernels for 10.04 in the past, they work better but still double the max jitter of the kernel with 8.04

The settings mentioned in the wiki are the minimum timings for the driver to respond as per the datasheet; the company advised me to err on the save side and increase the values a bit. Any thoughts on this?

Use those settings and calculate the base period, it won't be 100,000 I'm sure

The velocity and acceleration I have now is what I would like for the plasma; I don't need torque, but speed. Or am I missing something here.


Max velocity is governed by the max number of steps that can be generated per time period, and the actual distance that equates to.
Acceleration only governs how quickly you reach max velocity, but is likewise affected by maximum step generation.

In practice too high acceleration is the single easiest way with a stepper system to lose steps. This is especially true when doing rapid retractions with 180 degree change in direction.
Torque comes into it, the faster a stepper goes, the less it has of it and the easier it is to stall.

Others can explain in great technical detail the factors involved, I just know from building stepper powered machines that conservative acceleration figures, give soft start and stop, accurate positioning and
minimum problems.

I build machines to do my work for me whilst I do something else, I don't want to have to watch over them in case they hit problems because the speeds are set too fast.

regards
Last edit: 08 Dec 2012 23:31 by ArcEye.

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

More
11 Dec 2012 00:44 #27522 by peter76
Ok, did some more testing; made a dual boot system with 8.04 and 10.04. After some fiddling I had SMI disabled ( was already disabled ), the nic disabled in the bios, and had disabled two kernel modules to do with smi on Dell systems.....
8.04 gave some spikes over 100000 ns, with also showing overruns when testing with the command line tool.
In 10.04 I also checked the latency both with the command line tool and the gui tool. In both the latency goes slowly up to a bit less than 40000, no overruns showing up.
But when running the gui latency-test, I get an unexpected realtime delay message in dmesg. Does this matter if the command line tool doesn't give overruns, or should I ditch this computer and get another for Linuxcnc?

Regards, Peter

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

More
11 Dec 2012 01:31 #27525 by ArcEye

......should I ditch this computer and get another..........


Other than it is a P4 Dell, don't know what computer you have.
Does it have an entry in the latency tests page?
wiki.linuxcnc.org/cgi-bin/wiki.pl?Latency-Test

40,000 is pretty poor, but I am surprised it did not run 8.04 much better than 10.04.

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

More
11 Dec 2012 02:17 - 11 Dec 2012 02:18 #27528 by peter76
Sorry, it's a Dell Dimension 3100 C, not sure if it's mentioned in the wiki, there's a Dimension 3100 there, but I'm getting this feeling this is a different one.
It's a P4 with 512 mb. HT disabled. Last night I tried the command line latency test with SMI disabled in rtai.conf to see the smi thingies turn up, but nothing showed.
Also using an old ati rage pci card makes things much worse than with the onboard graphics.
I know it's poor, that's also why I'd like to try something else. Got another pc here, which I can have a go at, and else I'll see if I find something mentioned in the wiki.

regards, Peter
Last edit: 11 Dec 2012 02:18 by peter76.

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

More
14 Dec 2012 05:10 #27703 by peter76
Ok, hooked up another board I had lying around.... Jitter around 25000, but stable there; good enough for the moment. This problem solved for now.

Regards, Peter

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

Time to create page: 0.828 seconds
Powered by Kunena Forum