CNC sometimes has a mind of its’ own.

[JimDav]

Over the past few weeks I’ve experienced a strange situation intermittently on 3 different jobs. I always start with a climb cut then finish with a conventional cut to clean it all up. The run goes fine until the last few inches on the FINAL pass. Sometimes the Bot suddenly takes a turn back towards the X Axis along the Y axis. When this happens. My X zero point is changed by seemingly the distance amount between the original cut and the ‘new’ cut. It seems to be coming a more frequent occurrence. Any ideas on where to look for an issue?

Facts:
1) Older (~2006) PRTAlpha 4x8 Control software 3.6.46 Control board p/n 001003-01 v0.6
2) Windows XP computer - has never been connected to internet, all extraneous stuff shut down. Nothing else is ever run on the box EVER.
3) No comm errors or any other errors or power cliches
4) When jogged home, X home is now changed by the amount of the ‘detour’. Y is fine as is Z.
5) Shut down PC and control box and turned off power at the electric disconnect box, then powered back up normally.
6) Neither the spoil board or plywood is moving. Usually use vacuum hold down but started screwing down to make sure nothing moves.

Details:
2 pics attached.

1) When it occurs, it is ALWAYS in the last few inches of the final conventional pass. (I usually do it in one pass but even changed to 3 passes. Then it ran first 2 fine, opps’ed on last pass, same place.)
2) All 3 jobs have been run several times over months/years with no previous issues
3) Recreated the tool path. Original tool paths were created in VCarve 7.5. Recreated paths Vcarve 9.0. Interestingly file size changed from 205 KB to 72 KB.
4) Changed 0,0 point from actual 0,0 to 0,24. Ran 5 panels and all was fine (almost.) Check image 2. All 5 had a slight deviation as seen withing the green circle though on this job I can live with that.
5) Other day made 3 runs - all fine, next one messed up, next 7 ran fine. Nothing changed.
6) Material - Baltic Birch ply, 18mm with .189 mm DS bit (.189 because I’m creating some .25" bolt holes and don’t want to start a fire again) , 12 mm with .25 DS

7) An interesting observation a couple of weeks ago - was cutting a similar job as this but larger circles. Was cutting 2 pieces - 2x1 matrix. Again this was on final pass. The first one deviated on the X=0 side of the circle by about .45" I didn’t get to the machine instantaneously so the bot had moved over to the second piece for the final pass. It was cutting the final conventional pass there CORRECTLY until it go to the last few inches at the end, then also opps’ed just like the first piece. Very interesting. I would have expected it to cut the final pass on the second one off by .45". Was not so.

Sorry this is so long, but full of details. Any ideas on what is going on and how to fix it?
32061 32062

[srwtlc]

Grab on to the bottom of the spindle and pull/push back and forth in the X direction. Is there a lot of play in the lower rollors allowing the Z axis to be pulled into the cut at that point. Climb passes will pull to the outside of the part and when you come around with a conventional cut at full depth, it may all of a sudden get a full depth bite right there and pull the Z axis into the cut/part. At that point, it goes into Alpha mode and tries to recover, but ends up off track by the time it finishes.

[JimDav]

Grab on to the bottom of the spindle and pull/push back and forth in the X direction. Is there a lot of play in the lower rollors allowing the Z axis to be pulled into the cut at that point. Climb passes will pull to the outside of the part and when you come around with a conventional cut at full depth, it may all of a sudden get a full depth bite right there and pull the Z axis into the cut/part. At that point, it goes into Alpha mode and tries to recover, but ends up off track by the time it finishes.

Opps, I left out a fact I see, I use a Porter-Cable 7518 but the same concept should apply. I'll check it out. My only thought on this - the deviation is ALWAYS at the end of the literal very last pass a few inches from where it started. Even when I changed it to NOT make a full depth pass at once but to make several passes it STILL deviated on the literal last pass perhaps 2 seconds before finish.
Thanks for your valuable input.

Just checked it. Seems solid as a rock in both X and Y.

[jerry_stanek]

Did you use a Vectric product to create the file if so how does it look when you run the preview. It may also be a static problem or even a broken wire.

[JimDav]

Did you use a Vectric product to create the file if so how does it look when you run the preview. It may also be a static problem or even a broken wire.

3) Recreated the tool path. Original tool paths were created in VCarve 7.5. Recreated paths Vcarve 9.0.

Weather was very humid recently - I'm in southeast USA.
When it happens it is predicable to be in the last few inches of the final pass and ONLY there.

Tomorrow if it happens I will note the line number to see if it is consistent.
Thanks for the input.

[srwtlc]

Run an air cut, no material.

[JimDav]

Update - mostly good news. Cut 22 circles today - all perfect except one with very slight deviation in the expected place. 22 more tomorrow of this, then 50 more of a similar part. :(

However - interestingly, between the 3rd from last and 2nd to last, somehow the Bot decided my X zero point changed about .45 inches. Lower left corner WAS 0,0. That point changed to -.45,0. I did nothing but change the plywood blank. Gremlins at work.

Only deference between today and Friday/Saturday - I cleaned up the saw dust that the cable was dragging over, then started cutting after zeroing. That potentially could have created some static but really weird the deviation was ALWAYS at the same place.

[coryatjohn]

I'm curious about how you've grounded your machine? Do you have a dedicated grounding rod? Are you grounding every moving part separately? Is your electrical source grounded to the same rod as your machine uses? These machines and their dust collectors can generate a large amount of static, especially if the weather is warm and dry. Careful grounding can eliminate a lot of "gremlins".

[Gary Campbell]

From your pictures and descriptions you are losing or gaining steps. You say it always happens in one location. What is happening at that location? Is there a mechanical bind? Does the dust hose come in contact with the machine? Does a connector or wire loose connection at that point? Since it ALWAYS happens at the same position, it should be easy to diagnose.

[JimDav]

I'm curious about how you've grounded your machine? Do you have a dedicated grounding rod? Are you grounding every moving part separately? Is your electrical source grounded to the same rod as your machine uses? These machines and their dust collectors can generate a large amount of static, especially if the weather is warm and dry. Careful grounding can eliminate a lot of "gremlins".

Sadly (and embarrassingly) I have not done anything along those lines. Bought it used 5 years ago and set it up as the previous and original owner (Built cabinets) had it set up. Neither of us have put a dust collector it. I've seriously considered it but it has the 2 router heads and no room for a hose. Also it is located in a tight space and ceiling is barely 8' Early on I read about grounding and started looking into it. As time went on, surprisingly, never encountered issues. All excuses I know. Also considered getting echain but haven't.

[JimDav]

From your pictures and descriptions you are losing or gaining steps. You say it always happens in one location. What is happening at that location? Is there a mechanical bind? Does the dust hose come in contact with the machine? Does a connector or wire loose connection at that point? Since it ALWAYS happens at the same position, it should be easy to diagnose.

By 'same location' I mean the same relative position of the cut on each piece as in the images. No dust hose on the machine. :( Dust is blown off with air hose after each run, then (eventually) swept up and dumped.

Let me clarify what I mean by "ALWAYS happens at the SAME location". It DOES NOT happen on every run, its intermittent. WHEN it happens it is ALWAYS in the same relative location of the cut piece - end of the run on final cleanup pass just as it is approaching the 'start' point and ramping out to stop. Happened once a month ago. Happened again 2 weeks ago. End of last week it was happening about 3 times out of 7 runs - all cutting 1 unit of the the same thing as in the pictures. This was 3 different jobs. Today I cut 22, no issues other than one had a small divination on the 8th run. WHEN it happens, it is on the very last CG command then ramps out and back to a 'Home'.

Tomorrow I will be cutting the same pieces except rather than cutting one at a time (using 24x24" cut offs from another job) I will be cutting a matrix of 3. It will be interesting to see what happens.

I suspect the cable dragging though so much saw dust was most of my problem - there was A LOT of saw dust. Static was probably created. Yet, with my math/physics degree I find it almost impossible to think static would cause the same problem in the same relative location of all pics (WHEN it happened). I would expect it to be in random locations and different times within the run.

Thanks to all for the very welcome input. Yes, I need to work on grounding and dust collection. I do have a 3HP Laguna cyclone collector. It is just not piped to the Bot. Home basements leave a lot to be desired sometimes but it sure cuts down on overhead. :)

[JimDav]

I'm curious about how you've grounded your machine? Do you have a dedicated grounding rod? Are you grounding every moving part separately? Is your electrical source grounded to the same rod as your machine uses? These machines and their dust collectors can generate a large amount of static, especially if the weather is warm and dry. Careful grounding can eliminate a lot of "gremlins".

I should have stated that at install, a new 220 line was run by an electrician from the shop sub-panel. The sub-panel is a drop off the main house panel. It is a dedicated 220 circuit breaker and the ground bar used there. Essentially it is all grounded to the house ground which DOES have a ground rod. The PC is plugged into a regular 20 amp circuit. Hope that makes sense.

[coryatjohn]

I should have stated that at install, a new 220 line was run by an electrician from the shop sub-panel. The sub-panel is a drop off the main house panel. It is a dedicated 220 circuit breaker and the ground bar used there. Essentially it is all grounded to the house ground which DOES have a ground rod. The PC is plugged into a regular 20 amp circuit. Hope that makes sense.

I'm guessing this is probably the root of your problem. The amount of static electricity generated by the machine is too much to rely on the ground from the electrical outlet. The machine is also highly sensitive to static shock and can go bonkers easily.

You should do the following:

1. Drive a standard grounding rod into the ground at the closest possible point to the machine.
2. Run a heavy (6 gauge or greater) copper line to EVERY moving component on the machine and ground them to that rod. These lines can be daisy chained.
3. Run a separate line from the electrical box ground line to the grounding rod. DO NOT daisy chain this line. Run it direct from the box to the rod.
4. If you have a dust collector or vacuum hold down, ground the hoses, motors and housings.

Be sure to attach the grounding lines to your machine securely. I used copper eyes soldered to the lines and then bolted them to the machine. You may think this is all silly and overkill but it is not. These machines aren't reliable without absolute and perfect grounding. Any moving part that isn't individually grounded isn't grounded at all.

[JimDav]

Nah, With my science/computer background I know it's not silly and overkill. Actually I am a bit surprised that, in 5 years, there has not been the dreaded comm issues and that sort of thing. That said, I'm not a big production shop. Just me and a high school guy. The Bot may run 15/20 hours a week max.

[coryatjohn]

I have a 2013 PRS Alpha and before I went insane with the grounding, I would get occasional weirdness out of the machine. Typically on dry days or when cutting something really dust intensive. Since grounding the heck out of the machine (several years ago), no issues at all except when the power fails. Can't stop that.

It's an inexpensive and relatively simple fix. Maybe $40 for wire and the grounding rod. You can buy it all from Amazon or Home Depot.

[ron_moorehead]

Just one word of warning, to may ground wires can cause a ground loop which will cause a whole lot of problems for you. Best to check the ShopBot documentation for recommended grounding.

[skintigh]

Update - mostly good news. Cut 22 circles today - all perfect except one with very slight deviation in the expected place. 22 more tomorrow of this, then 50 more of a similar part. :(

However - interestingly, between the 3rd from last and 2nd to last, somehow the Bot decided my X zero point changed about .45 inches. Lower left corner WAS 0,0. That point changed to -.45,0. I did nothing but change the plywood blank. Gremlins at work.

Only deference between today and Friday/Saturday - I cleaned up the saw dust that the cable was dragging over, then started cutting after zeroing. That potentially could have created some static but really weird the deviation was ALWAYS at the same place.

That's a lot of circles. Naturally the CnC machine is getting bored and drawing outside the lines; give it a more complex shape now and then to keep it happy.

My initial thought was dust buildup is causing drag/slipping/catching or confusing a proximity sensor and you should try adding more circle loops to a cut to see what happens. But the randomness of the miss-cut suggests a communication issue, so the static comments are more likely. And good point on avoiding ground loops!

Also, I don't know anything about your machine, but is it using a USB connection? If so, you could try a different USB port on your PC -- if you are using a port on the back, try one on the front. I know that sounds silly, but I had a scanner that would work great but sometimes screw up right at the end. After trying literally everything else, it turned out the the USB ports on one side of the machine didn't quite meet the USB spec for current and the voltage would drop during communication and sometimes glitch, but the ports on the other side had a different USB controller and were to spec. I swapped the ports and never had a problem again. Kind of a long shot, but a simple test.

Perhaps even a longer shot is the power supply in your computer is dying. It sounds like an old machine. Does it sometimes not boot up the first time you turn it on, and you have to press the reset button to get it to boot?

You say you don't have dust collection, dare I even ask what the inside of the computer looks like...? Dust buildup could cause it to overheat towards the end of a job.

[JimDav]

That's a lot of circles. Naturally the CnC machine is getting bored and drawing outside the lines; give it a more complex shape now and then to keep it happy.Love your comment. Don't know about the machine but I sure get bored. It does get other things to cut besides these.

I come from a career in the IT field and understand what is going on inside the box probably better than modern day 'computer' people from a bits & bytes point of view. Yep, I've flipped around some dials and changed program code on the fly in both octal and hexidecimal. Those were the really fun days. I've been around for a looooong time. Seen the industry grow up. Yes, sometimes seemingly silly things can, and do, make a difference.

Per your comment about a dying computer - nah, it's fine. Boots up fine. It sometimes gets a good cleaning.

So, 50 more circles tomorrow same thing but a bit smaller and no hole for a plug. Next weeks job is another job like this - 60 circles but 30" in diameter. After that is another job that is not a bunch of circles so may be the machine won't be bored. :)

I think everything is under control now. Even let my mid-teen helper run the Bot this afternoon.

Thanks for your comments.

[coryatjohn]

Just one word of warning, to may ground wires can cause a ground loop which will cause a whole lot of problems for you. Best to check the ShopBot documentation for recommended grounding.

The key point in avoiding issues with grounding is to ground everything to a single point, like in one rod.

[Gary Campbell]

If you add a ground rod, doesn't that introduce a second ground point? Is not the building electrical system ground functional? Adding one is asking for trouble.

[coryatjohn]

If different points of the machine are grounded to different rods, then you have a problem. If all the points on the machine and the electrical box are grounded to the same rod, even if there is another rod located elsewhere for the building's electrical system, then you have a good zero point.

[robtown]

If different points of the machine are grounded to different rods, then you have a problem. If all the points on the machine and the electrical box are grounded to the same rod, even if there is another rod located elsewhere for the building's electrical system, then you have a good zero point.
I just completed building a house last year.
I failed the final inspection first time around because of improper grounding.
I needed to bond both the gas and waterlines back to the main, "married to the main" the inspector called it, as well as add a second ground rod at least 60" away from the first, connected to the first.
So Basically my switch panel, water and gas lines are all grounded to the meter box which is then grounded to the two ground rods.

[skintigh]

If different points of the machine are grounded to different rods, then you have a problem. If all the points on the machine and the electrical box are grounded to the same rod, even if there is another rod located elsewhere for the building's electrical system, then you have a good zero point.

If the PC is grounded to 1 rod, and the CNC machine is grounded to a different rod, current can run through the commutation cable between the PC and CNC, and neither machine may be designed to handle current on the ground wire of the comms cable.

Just being plugged into different outlets can sometimes also cause an issue like that. That's one of the reasons a lot of audio equipment uses fiber optics for connections between distant components. But I think it should be easy enough to test with a volt meter... unless it changes with the weather...

[CNCDave]

Check your bearings on each axis. I had an issue with my 2012 Desktop (circles in particular) and found that one of the y-axis bearings was going bad. This drove me crazy for a couple of weeks since it appeared to be random, but in retrospect, was during movements that pushed against that bearing.

The fact that you're doing a conventional cut on the final pass (thereby increasing the load on the bit) makes me think you might be running into a moment when the bad bearing pushes back just enough against the x-axis to cause it to stall -- think

"Oh goodness", I forgot to attach the clip to the router bit when zeroing my z-axis, and as soon as the bit hits the plate, it tries to drive through the plate, thereby mucking up the axis zero values
only the x-axis is able to recover after the stall without you really noticing it (the x-axis motor may have complained a bit momentarily, but the spindle/router racket would likely have covered up any noise).

[bob_reda]

Has the same problem at the same point when doing circles. Brady told me to slow it down a little and I haven't had the problem again

[JimDav]

Has the same problem at the same point when doing circles. Brady told me to slow it down a little and I haven't had the problem again

I'm the guy that posted the original post. In my case when it happened it was ALWAYS the last inch or so of the final pass. It was ALWAYS on the last CG command. I too climb cut, then finish conventional. Didn't matter if I made the last cut in one pass or several - it was ALWAYS the last iteration around the perimeter. I too was cutting a ton of circles though did have it happen once or twice on rectangles. After those 'circle' jobs of 200 or so circles, cut 30/40 sheets of other stuff with no issue. Last job was a narrow rectangle with a round nose. Problem popped up again. I did several things and got around the problem: 1) Started using spiral ramp, 2) Used a new 1/4" bit, 3) made multiple passes on last conventional cut, 4) watched it like a hawk with finger near the space bar to stop the machine if necessary. May have been lucky but in the first runs of circles got the jobs done. This last time cut 60 narrow 5.5" x 26" rectangles with round nose with no issue. Many suggested a static problem. It is not. A static problem would be random. This was predictable down to the exact spot, exact command where it would happen if it did. Someone suggested a bit can sometimes grab the wood and get pulled in. That MAY be what was happening. I'm claiming it. Or maybe I'm just lucky :) At any rate, this is an update on my situation that I should have posted a while back.