Machsupport Forum
Mach Discussion => General Mach Discussion => Topic started by: mbele on April 18, 2013, 04:14:35 PM
-
Hi,
I've noticed some time ago that when I increase feed rate via feed rate override in mach3 mill it looks like CV mode turns off (moves are not smooth any more). I'm able to reproduce this behavior now, so I'm posting my g-code and xml file. When program is running at 4000mm/min it runs fine, but when the feed rate is set to 5000mm/min or above (via feed rate override or in g-code program) moves are not smooth any more. What's more interesting is that, at raised feed, the first pass is smoother that the following passes :) I haven't noticed lost steps.
PC is based on AMD Athlon 2400+ 2GHz CPU with 1GB RAM running XP SP3 (Standard PC install). It takes one minute for the attached g-code to initialize (I've always experienced very slow initialization when program is based on arcs).
I get peak power from steppers at ~4000mm/min (~200RPM x 20mm lead direct drive 2.7Nm steppers 1/8 micro-stepping). Kernel runs on 25kHz.
I'd really like to be able to feed faster since my spindle is not very strong and I have to take light cuts.
Thanks,
Mihael
-
You may want to take out all the unnecessary G17 commands. Takes the load time down from 1 minute to less than a second here.
Ian
-
Thank you Ian!!! Not only that program initializes in no time, but the quality of motion is so much better. I'll take out redundant Gs too :)
Now it runs fine at 5m/min with 0.3G accel. For higher feeds I had to lower acceleration to avoid loosing steps. Above 5m/min with 0.1G accel motion is not smooth any more - slowdowns/speedups are noticable, but since accel is relatively low, the machine doesn't vibrate too much. Would higher acceleration make the motion smoother (actually, I tested that and the answer is no, but I cannot be sure since I'm on the limits of available torque) ?
Thanks,
Mihael
ps. I'm putting rhino script that generates g-code from arcs and lines in downloads section
-
I would guess that the G17 on every line was killing the planner.
*generally* the higher the accel the better as long as the *robustness* of your machine can handle the increase in jerk.
Ian
-
Can someone run attached test file @ 8m/min with 0.1G accel ?
I just managed to run it in the air @ 8m/min with 0.3G accel, and the motion is smooth, but these are not safe settings for my machine.
If I lower the acceleration (tried 0.25G, 0.2G, 0.15G and 0.1G) motion is not smooth any more - making slowdowns/speedups between segments more obvious as accel is more decreased :)
So, my current conclusion is that there are "valid" acceleration and feed rate settings for CV algorithm to work properly, and that they are somewhat proportional (at least on my system/setup, running latest version of Mach3)
Actually, could it be that, because of the lower acceleration, CV is turned off so that cut accuracy wouldn't suffer ?
Thanks
Mihael
-
Can someone run attached test file @ 8m/min with 0.1G accel ?
Set my accel down to 1000mm/s/s to get 0.1G, heres the video.
http://youtu.be/HwoikeEF2N0
Hood
-
Well.... maybe not quite as impressive ;D but I ALSO had to turn mine down to get 0.1G :P
here it is at 0.1G
http://www.youtube.com/v/dDY0QXkL0ig?hl=en_US&version=3
and here at 0.5G
http://www.youtube.com/v/B03MQNLUuNc?hl=en_US&version=3
Although it's OK at 0.5 you can hear it *bangs* a tad so that's why I normally have it backed off to 0.25G.
Ian
-
Thank you Hood & Ian! I don't know what to say... guys, you're great :)
When I watched Hood's video, I saw the same behavior as in my system: machine slows down/stops three times during execution of one loop - only, on Hoods video it looks smoother than on my machine. I allready wrote the response, thinking that it's normal behaviour and that I'll have to live with that, but then I saw Ian's videos that do not show those 3 glitches :)
Ian, what are your CV settings / lookahead ? Maybe there's a catch...
-
Ians does look smooth.
Are you using the parallel port?
Hood
-
CV is plain "Mach out of the box" - lookahead 20, everything else off. System is an old AMD 1.5GHz / 512MB with onboard graphics and parallel port.
Ian
-
I played with lookahead and kernel frequency, but with no results. Ian, would you please, if it's not too much trouble, post your settings xml file, so I can do the comparison ? BTW. your machine is quite impressive, handling theese feeds and accelerations with no vibrations (being built from alu extrusions, if I'm seing right) - do you have build/specs thread ?
I ran test one more time with 0.3G accel, and though I cannot see that machine slowsdown/stops, I can feel it if I touch the machine frame. So, my theory about low acceleration causing CV to fail is not correct. Also, I presume that PC is fast enough, since Ian is getting good results with slower PC.
Thanks,
Mihael
-
Ian what version of Mach are you using? If I recall some versions had CV issues and some didnt.
Hood
-
Hi Hood, on the machine in the vid I'm using 42.20. (never got round to updating it)
Coincidentally I have an eye on CV being "the problem". (or more to the point a MYTH about CV being the problem).
When Mihael asked me to post my xml I did a quick compare of mine and his. There's basically just one difference. I notice he has "Stop on CV angles > 89.9"
Mihael please turn this off (untick it in general config) and try your toolpath again.
Ian
-
Well I cant try that version as it will be too old for the CSMIO that I use and regarding your Stop CV, I dont have that enabled in my profile so possibly not that.
Hood
-
I'll whack 43.62 on it and get back
Ian
-
Thanks, would be good to know. I am using 056 or 057 I think.
Hood
-
Yep - looking like you're absolutely right Hood. With 43.63 I get exactly the same problem.
Ian
-
Ok thanks, I am thinking it may be 022 or before that is needed, seems to after that version that a lot of things got screwed up.
Dont suppose there is a chance of you trying 022? :)
Hood
-
will do but it'll have to wait till tomorrow - unfortunately I have to go out and drink some beer!
Ian
-
Ha ha as good an excuse as any I suppose ;D
Hood
-
I think I was using 043.022 when I noticed the problem, and then upgraded to latest version. What I found weird, is that I couldn't find reference in the changelog for the version I was using prior to update....
During testing I turned off all CV related extra features, and I'll keep them that way until I find good scenario to use them :)
Guys thanks again for looking into it. Where can I download older stable builds ?
Cheers,
Mihael
-
ftp://machsupport.com/Mach/
-
OK - tried 43.22 - no go, 42.40 - no go. The latest version that works is the one I started out with i.e. 42.20.
Ian
-
Thanks Ian, looks like its even further back than I thought :(
Hood
-
Thank you guys. Lot of issues got fixed since 042.020, I hope those will not prove to be serious :)
-
No worries - It makes a refreshing change to have someone ask a question and who then SUPPORTS and THANKS the people who take the time to try to help them. Any time Mihael.
Ian
-
No worries - It makes a refreshing change to have someone ask a question and who then SUPPORTS and THANKS the people who take the time to try to help them. Any time Mihael.
Ian
agree 100%
-
:) Thanks guys, I have made a note of this topic and will be reporting it.
Brett