Random motion and lost steps after upgrade

Hello all and thanks in advance for any help anybody can give me!

I have tried everything i know and am running out of places to look for the problem...

I recently redesigned a gantry style router that i had made a few years ago and it is a combination of new parts and re-using old parts. the original machine ran 2.5 on lucid forever without ever a single problem! it just worked and setup was pretty straight forward.

then i rebuilt this router and all was well, but i started reading about JA and dual homing for auto squaring. so i upgraded and now I'm in a world of hurt. I've been beating my head against a wall or any other hard/sharp object for 2 weeks now trying to get this working (which means 2 weeks of not cutting anything)

here is what is going on: I'm drastically losing steps on my z axis like no other. it is not random. it drops the EXACT number of steps each and every time (the bit carves the exact path every time). i only lose steps on retracts. when i run a program, a simple carbon fiber drone body cutout with lots of drills, i get the same motion every time -- start, rapid to first drill, a little stutter up(????), rapid down to 1mm above piece, spiral down, a stuttering drop at the end, and retract to 2mm or so less than where i started. this program has been ran several times on the original machine with never a hitch. when i run other programs, i get more or less the same results, not consistently between different programs, but consistently within the same program. some cause it to jump drastically, sometimes it spirals down to the bottom of a drill and suddenly speeds up while dropping the z. I'm posting this question here because i think that because it is consistent, that it is a software issue.

here is what i did different from one to the next --

the original controller was xyz with dual y steppers split from one driver on a 5 axis breakout. it was 2.5 on lucid, with an older core 2 duo machine (emachines t5088 with upgraded cpu) with 2gb ram and pci parallel port.

the new controller is xyyz on same 5 axis breakout, 3 of the drivers are the same with 1 brand new driver, and the same computer configuration. the difference is it is a clean install of the wheezy image from linuxcnc.org, with added apt sources and is running the latest 2.8rc from the master branch. i forgot to check which build it is, i just updated last night and the first build i installed was from about 2 weeks ago.

.... and here is where i have to say miserable i am with wheezy. this is my first time really using a -actual- debian release and i must say that it hit my computer like windows vista. in fact, vista would probably be an improvement over what i have now...

i have tried to as much as i can to eliminate any possible hardware issues.

• swapped the breakouts with 4 different ones
• swapped the drivers with brand new ones, working ones from other machines, and other random ones lying around
• checked, rechecked, swapped, replaced, resoldered the cables
• gutted and rewired entire controller

now I'm working on hooking it up to a different computer that is nearly identical and running another router that is a spitting image of the problem one except much larger, but i won't get to that until this evening.

this is not to mention that when it does move i can't even look at it without getting joint following errors. What am i missing? i know it has to be something stupid, but I'm running out of ideas here. is it software? configs? hardware? electrical? mechanical? where do i begin and what method do i use to trouble shoot and track down the culprit?

I'm sure that somebody has been through this before and any guidance would be awesome!!! thanks!!

Did you check your latency with the new installation? You might have forgotten all the tricks you might have used to get good latency on your old Ubuntu 10.04 installation. All those old tricks will likely need to be applied to your new installation.

If the latency numbers are good (or about the same as your old setup), check to be sure you have the step timings, max velocities and accelerations set the same as your old config.

check to be sure you have the step timings, max velocities and accelerations set the same as your old config

The configs were copied straight over from the original install, and (surprisingly) the latency is around the same even with window stutters and refresh problems (wheezy runs like a flaming turd on this machine).

i also have my acceleration dialed all the way down to 1.0 and my ferror and min error are set a 1.0 just to keep from getting joint following errors!

Guess the next question is what are your PC specs. (processor, memory) If the performance of the PC is so poor maybe it isn't meeting the minimum specs for Wheezy.

You should be able to run the latest versions of Linuxcnc on Ubuntu 10.04. If the system worked find before, go back.

Having joint following errors really sounds latency related. The stepgen doesn't deal very well
with servo thread latency but there are some things that can be done to make it more robust,
most importantly:

Have you set the stepgen max acceleration setting to 1.25x the per axis acceleration value?

Todd wrote:

Guess the next question is what are your PC specs. (processor, memory) If the performance of the PC is so poor maybe it isn't meeting the minimum specs for Wheezy.

its a core 2 duo 2.66 gHz , 2gb pc4200, radeon x300 pcie, pci parallel port, 160gb hd. this machine ran 2.5 with no problems and smooth as glass!

You should be able to run the latest versions of Linuxcnc on Ubuntu 10.04. If the system worked find before, go back.

The minimum version of ubuntu i can run is 12.04 because i am using 2.8pre right now and i have not heard a good word about precise, although i haven't tried it myself...

PCW wrote:

Have you set the stepgen max acceleration setting to 1.25x the per axis acceleration value?

right now it is about 20x the per axis because i have the acceleration and max speed dialed down so slow now. could that be the cause because it is a repeatable phenomenon and it always seems to happen at the beginning or end of a move on the z. later tonight i will try and maybe get a video of it and that might help if somebody saw it happening.

A following error plot on Halscope might be more useful than a video

oddwick wrote:

The minimum version of ubuntu i can run is 12.04 because i am using 2.8pre right now and i have not heard a good word about precise, although i haven't tried it myself...

I do not know of any reason you can't run 2.8pre on Ubuntu 10.04.

oddwick wrote:

right now it is about 20x the per axis because i have the acceleration and max speed dialed down so slow now. could that be the cause because it is a repeatable phenomenon and it always seems to happen at the beginning or end of a move on the z. later tonight i will try and maybe get a video of it and that might help if somebody saw it happening.

I have seen having a huge discrepancy between stepgen max acceleration and axis acceleration like that cause some weird things. On the first machine I set up I made that mistake and it cause the axis to oscillate like a badly tuned servo, Pulled my hair out quite a while trying to re-tune the servos before someone pointed out the error of my ways.