So I'm going to throw a bunch of Ideas out there for you to think about.
Maybe try the indicator idea first....
Well I looked at the file again and stepped through the start of it in mach....
This is what has been lightly nagging at me about it...
When you plunge then move in the first short y and x the tool has a full load on it... as it moves forward there is a heavy load and a lateral force on the bit from the rotation of the cutting edge, after the first two little cuts you are now only cutting on a partial step over.
This is a long shot but is it possible that the lost steps are a result of this first heavy load cut of each layer???
As an experiment, could you just do a small square/circle say 1/2inch all the way through, then run the full code afterward to rule out the heavy cut theory??
Also just curious on what your acceleration values and steps per inch are?? to high an accel could lock up the motors on a high load, but on the short first move you might not notice??
Again, another idea to eliminate variables, could you try doing a plunge then move above the work say 6 inches then plunge again to verify the distance traveled, then move over an inch and plunge then do a full load cut for the six inches. This would show you if the axis losing steps from the load (cut path is shorter than the two plunges,) or drifting sideways due to the lateral load on a heavy cut (one end of the path is closer to the plunge.) Might need to repeat in all 4 directions to tell for sure if it a directional problem.
If neither of those are a problem then it has to be a Z axis problem.
How about putting a dial indicator on the X add Y and then run the Z up and down at various speeds to see if you are getting "bleed over" steps to the other axis??
I very much appreciate all of the suggestions from you guys BUT:
There are no heavy loads or heavy cuts going on here. Please forget this theory. The file I'm using now only plunges into the wood (pine plywood or soft poplar for these tests) twenty thousandths of an inch at 15 ipm for each cut with a 1/4" router bit. That's nothing at all for the torque this machine can generate. The machine has the power to bend a lot of things if the bit goes 3/4" into pine plywood and tries to cut at 1,000 ipm. That is far from the case here in these tests. Theoretically, the dual drive X axis can generate close to 3,000 oz-in and the Y can generate close to 1,500 oz-in at these speeds.
Ideally, if there were no missing steps or other interfering signals, and the bit was was bending due to flexing in the Z axis the bit would be making a hole that is undersize by the amount of the flexing, not stepping over by the same amount and in the same direction with each pass. The walls would have a slight angle on all four sides. The walls would have the same angle if the bending is equal in all four directions of travel.
One more test - If I swap the X and Y motor cables at the Geckos will the step over change direction? If yes, then it is an electrical problem.
Pandinus:
Adding the 1/2" square isn't a problem.
Right now, the accel value is 20. Plunge rate is 15 ipm. Feed rate is 20 ipm. It takes 9 minutes and 54 seconds to complete a full cycle of these two small pockets. I can cut those in half and I believe the step over will probably not change perceptibly. That will be an easy test of bit loading but it will double the cycle time. Cutting at 150 ipm at 20 thousandths mignt be a more interesting test of loading to see if the step over changes. Maybe 300 ipm also.
Your last test is similar to one I have considered but have not done yet. A number of parallel straight lines at different feed rates, spaced apart by an inch, and at the same depth of cut. Maybe some in the X plane and the same number in the Y plane. Then measure the lengths.
CarveOne
CarveOne
http://www.carveonecncwoodcraft.com
Possibility #1,000,001...
My machine had unequal backlash up and down in the Z axis, and this caused the Z to rise over time in designs that have a lot of Z motions like the Bullhead pub sign from Vcarvepro. This drove me crazy until I figured that out.
Could your R&P machine be prone to unequal backlash on the X and Y? I appreciate you tightened up the tension on the pinion spring. I presume the pinion is attached to a plate that causes it to move in an arc until it makes contact with the rack teeth. If this arc moves the pinion to where it makes hard contact against the "front" tooth and still allows clearance to the "back" tooth, you could have unequal backlash. If so, it could cause your machine to "walk" in that axis.
Good luck in your quest.
I just realized that swapping the motor cables might just reverse directions if a mechanical problem exists in one or both pinion drives. So much for that brainstorm.
CarveOne
CarveOne
http://www.carveonecncwoodcraft.com
I dunno. If the teeth in the rack is cut too shallow the pinion teeth will bottom out before the walls of the rack teeth engage on both sides of the pinion teeth. I have been suspicious of these particular racks and is the reason I have mentioned replacing them with better grade of rack a number of times so far.
CarveOne
CarveOne
http://www.carveonecncwoodcraft.com
Machinists who are checking out gear contact patterns use a type of blue dye on one set of teeth. Then they mesh the gears together and run them for a minute. Then they check how much dye has transferred to the mating teeth to see if the teeth were truly in contact. Maybe you could use some blue chalk?
When I saw your walking cuts, I didn't think it was your gantry either because the cuts do look square. Here's my thinking on your diagonal cuts: (1) X- and Y- are moving by the same amount. (2) X-and Y- hardware are independent of each other. Movement by the same amount caused by hardware problems is too much of a coincidence. (3) X-and Y-controllers are independent of each other. Movement by the same amount caused by controllers is too much of a coincidence. (4) Movement by the same amount caused by wiring is too much of a coincidence. (5) It's either in the parallel port, the software or the computer itself. It's the pulses. (6) The computer is new , so it will have enough power to handle pulses. (7) That leaves the parallel port and software.
That's why I searched the Mach support forum to see if others have had drift problems with software. And they did. If the Sherline does not solve your problem, I've got an extra parallel port that I know is working that I can send you if you want.
Carve One
I have been following this since the start..
I also followed your earlier build and your sander build...Want to do the sander build..
For your current problem I can't offer too much.
I would go to a pencil or pen and paper. I know you are sure there is no load. But you can do X/Y only with out Z and see what happens. Then you can add Z if you have a spring pen holder.
The other thing is I remember some other builder fighting an almost identical problem when trying to mill circuit boards. I can't remember the solution, nor can I remember if it was here on the zone or another site. But it seems as if he was fighting something very similar.
There have been issues where the direction change signal triggers a step signal depending on timing and Bob/driver logic.
It almost seems to repeatable for noise. So I suspect some sort of logic timing issue.
Also I saw you were running 45,000. Art has always said use the 25000 unless you have to have higher step rates...
Good Luck
Garry
Carveone can you try swapping the parallel port cable from your older machine to this one -
My machine never returned to its origin when i was testing, and some one recomended me to change the PP cable and it worked
or may be I missed something???
Been there done that...HTTS
I had this z drift--up that manifested itself when doing the dished v-carvings. I could start at z-0 and end at least 1" higher. I did all of the stuff suggested for several months, swapped drivers, swapped motors, swapped z-screws, play with every combo of settings. I finally gave up and decided to put a nema 34 motor on it so I got one and got a 203v driver. I was running the keling 40xx drivers. Well, I thought that I'd try the motor driver before changing out the motor, and guess what, works perfectly. So I'm not sure exactly why there would be a change, apparently either from the drivers themselves or from the confusor to drivers, not sure.
I thought you were running 203v's, but I don't remember your set up.
I know how frustrating this is...........
BG
After the talk about possible electrical fault/noise/extra pulses I thought some more and really don't think that's it. I'm pretty sure its not mechanical or electrical, those faults just wouldn't be so precise and repeatable...
That first square in the alloy, the amount of XY error looks the same for each new plunge, it also looks the same distance AS the plunge. Hence the "ant olympics" jokes because they look like little steps with the same amount of XY as Z.
What if there's a fault with the gcode, where every time it does the Z+ plunge before repeating the square it also does X+ and Y+ at the same time?
I know you thought gcode error was ruled out, but is it possible there is another setup option separate to the gcode of the job itself? Where your system is interpreting a good Z+ instruction as a Z+X+Y+, or maybe it's a faulty Z+X+Y+ instruction in your code that other people's machines may correct back to Z+ due to different setup options?
You could also look at some other fault with configs, like if a XY compensation or some XY offset is being re-added every iteration of the code. Depending on individual setup this could cause it to have that symptom on your machine but not on other people's machines.
It's just another guess, but personally I'd be looking at gcode and configs at this point.
I've run his code and it works fine
BG
That's a good idea when the rack and the pinion are clean. Unfortunately, the racks and pinions now have white lithium grease on them. Way back when I block sanded the rolled edges off of the teeth of these racks I made sure I didn't go below .500" and I placed a pinion in the racks to insure that I could see light through a small gap between the rack valleys and the pinion teeth. The engagement should be ok. I just have to use them until I can buy longer and better gear racks for the next major upgrade to an 8' working length.
CarveOne
CarveOne
http://www.carveonecncwoodcraft.com
Thanks. I just might take you up on your offer if I can't borrow one locally to try. I don't mind just buying one that is known to work. This one is what TigerDirect had in stock at the time I bought the computer.
First thing I'll try tonight is going back to 25k pulses and leave it there.
CarveOne
CarveOne
http://www.carveonecncwoodcraft.com
It just amazes me no end to see the number of people on this thread trying to help and solve C1's problem. This has got to be one of the best, if not the best, community out there on the net.
Now, if we can only find the real problem, it will be so much better. :-)
CarveOne
http://www.carveonecncwoodcraft.com
I'll certainly "second that emotion".
I do appreciate the help and suggestions. It can get frustrating though. With such limited time to work on the machine I just can't respond here and do the work as fast as suggestions are offered.
This thing will be so much fun to work with when it is working correctly. From past experience once we know what is causing it we will have that "well duh, why didn't we see that right away" feeling.
CarveOne
(Who is on the way to the work shop for a couple of hours)
CarveOne
http://www.carveonecncwoodcraft.com
if you run the code with a spring loaded pen holder (being sure it is aligned as well and that it has enough room to travel the full depth of z axis travel) with out making any other changes wouldnt the "drawings" have the same 45 degree drift ? if so that would rule out "loading " the cutter as there would be no loading on the z axis. or simple running the g-code above the part then going back to some known point and see if they are still the same point.at this step (actually a few "missed steps" ago..) i dont think it has anything to do with loading
if you then have the same problem (the drifting down and left) it would seem to be that the B.O.B. ,or the parallel port, or the wiring inside the cable carrier (maybe the router motor wire??)...are somehow getting some step signals crossed...or bleeding across into the wire next to it
maybe you could take all your wires out of the cable carrier and leave them draped across the back of your gantry ,or taped (apart from each other , isolated from each other) along the back or the gantry and try running the g-code with the router turned off,then on again to see if any change?? this would eliminate the wiring as the bug,..then move on to the port, and then the BOB??
you have alot of support on this carve' ,..just ...we are all in a diffrent state... you would get no sleep at all if you lived near all of us ..as we would be at your house till 3:00AM every night till we get this fixed,
admit it guys you go to work thinking bout carve's router gremlims like me ??
"witty comment"