Re-wiring a 2001 CNT Motion Systems Router on the data side.

[Joe T]

All,

I have made some progress on my "Drifting X" problem. I performed a parametric evaluation and learned a lot. The main thing I learned is that my problem is not one of noise or cross talk. It is apparently a programming issue.

I plan to complete my evaluation tomorrow (Labor Day) and then I will publish my findings.

If things go well tomorrow, I am back in 'business.' If not, eeeerrrrrggggghhhh!

BTW, I plan to post my findings here and on the newly created 'CNT Motion' sub-forum.

Regards,
Joe T.

[Joe T]

All,

An update:

I have been trying to understand why my CNT Motion router 'gets lost in X' for a long time. I decided in early July to take a more structured approach to trying to understand this problem. I have performed a number of simple, parametric tests.

So far, this is what I have learned:

1.) The machine works very well. My problem is not a CNT Motion router issue, except that there is no programming info/training guide available. So, it has been a long learning curve. I don't consider the WinCNC reference manual a programming guide. It provides no syntax, etc., on how to properly use many of the G and M codes. Also, that manual does not have all of the M codes available on my machine. Many have been written by CNT and I can not get a listing or any other info about them. Purely trial and error. Not a good way to run a business - - -

2.) It is a software problem, not hardware. I have eliminated WinCNC and editNC (my G-Code editor) as the root cause of my issues. It is CATIA and how I, and others, have the basic CATIA options set up.

3.) If I run a simple G-code program I have written, including tool changes, work offsets ("HOME" as CNT calls them - how STUPID!), etc., the machine works flawlessly. These programs can be written in a basic text editor or editNC. Doesn't matter. When I try to run a more complex program from CATIA, I have problems.

3.) So, I am trying to identify the reasons the CATIA I use does not give me the tool path output needed to make a part. The WinCNC viewer shows the program tool paths are PERFECT, but the actual part is not correct - sometimes WAY OFF!

So, still working on the problem. I will update again when I have more data.

Regards,
Joe T.

[gfacer]

Joe, are you positive that the hand coded problems are OK? I wonder only because you mentioned thousands of lines of code gave so much movement in X and that's hard to replicate by hand.

The other thought, somewhat related, is some type of rounding error if one system is set to 4 digits and one to 3, and say only the first 3 digits (after the decimal) are picked up and used for motion, then you might, with 1000's of lines of code, get a cumulative error like you've seen. Worth double checking. It might also be worth finding another cam program to demo or otherwise make a similarly complex test program with and see what happens. You'd need a post processor of course, but I have one that would be a good starting point (tool change would be different) for vcarve and aspire of you find someone local with a seat you can use (or not local I guess might work too).

I just can't tell if your really have the problem traced to Catia yet. But, the fact the preview looks OK and the output is not makes me suspect Catia isn't the issue.

Check what the wincnc.INI file.has for decimal places versus the gcode from Catia....although that issue could be internal to Catia too (taking say an internal 10 decimal number to 3 decimal places for output)

Good luck. My focus for next few weeks will be getting a 1992 Cam tech I just bought for 1k up and running....if you want a machine without support, these fit much better than CNT let me tell you!

[Joe T]

All,

I have identified the source of the problem. But, that 'source' could have one of several root causes.

I have continued to run diagnostic tests.

If you remember, I originally thought that the source of the problem was something about the CATIA APT.Source converted to G-Code that caused the 'X-Drift.' Actually, it is an 'X-Shift.' This X-shift only occurred when I ran a CATIA generated tool path. When I ran one of the existing simple programs that were on the machine's computer, everything worked fine.

Well, I started performing parametric tests when I only changed one variable at a time. I would work the X axis. I could not cause an X-Shift using only X motion. Then I would work the Xand Y axes. I could not cause an X-Shift using only X-Y motion. Then I would work only the X-Z axes. I could not cause an X-Shift using only X-Z motion. The simple programs I wrote to test the X/Y/Z xis were all written directly in G-code - no CATIA.

Each of these were 'sawteeth' The X/Y were in a horizontal plane and when I ran these programs I cut traces of the sawteeth in the MDF spoilboard. Each set of sawteeth had six 'teeth.' I would run them down the X axis and back to X=0. I ran 20 sawteeth per test run. They were perfect.

I ran the X/Z sawteeth in the vertical plane. Again, six sawteeth along the X axis. This time, each vertex at the bottom of the sawteeth would touch the MDF and leave a mark (-0.1" cut 'spot'). Again, PERFECT! No X-Shift!!!

HUMMMM - - - -

A CNC tech I have used to try and diagnose the problem said he thought it was a matter of the acceleration rate (feed rate) used on the X-axis. OK, I'll test that - - - -

Well, I decided try running some slant lines (X0 Y0 to X35 Y35 back to X0 Y0, etc.) at increasing feed rates. For each increase in feed rate, I changed the starting and ending X position by 2 inches. So, each set of slant lines were parallel but offset by 2". I kept the Z axis off the table EXCEPT to mark X5 Y5 at the beginning of each set of 20 lines (feed rate) and at the end of each set of 20. Each run had 20 'slant lines' at different feed rates (accelerations). My program had one set of 20 slant lines run at a feedrate of 500 in/min. The next set of 20 had a feed rate 600 in/min, etc., through 1,000 in/min. All of these were in one the same program - hand written G-code. Each set of 20 was offset in X by 2".

Well DANG!!!

The 'touch mark' at the beginning of the program and the first one after 500/in/min were in the same spot. But, each successive touch marks (after the 600, 700, 800, 900, sand 1,000 slant lines of 20 each) were offset (SHIFTED) by ~3/4" from X=5!!!!!!

So, CATIA is not the problem.

But what is?

By yesterday, Friday 11-6-2015, I had convinced myself that it was either a problem with the x-axis servo motor's encoder or the X-axis servo drive. I went to the shop yesterday (Friday) to try and remove the X-axis servo motor and the servo drive. But, before I did that, I had to try some more 'tests.'

I tried running a set of 20 X only acceleration tests, These started at X0 and ended at X65 then returned to X0, etc. for a total 'back and forth movements' of 20 sets. I ran this at 400 in/min and 1,000 in/min. Again, 1,000 is the maximum my machine will run. Obviously, not an X-axis only acceleration problem.

HUMMM - - - - What's next??

I run an X-Y acceleration test. X from 0 to 65, Y from 0 to 2 (I didn't have much Y to work with since I was near the end (Y+) of my table for the work offset).

Again, sets of 20 with a touch mark at the beginning of each set of 20 and at the end of the run. Like before, i ran two runs - one at 400 in/min, the second at 1,000 in/min.

HUHHHH???? Every touch mark was PERFECT!!! NO X-SHIFT!!

OK, must be the combination of the X-Z axis.

So, I run another pair of test runs varying the X axis from 0 to 65 and the Z axis from 25 to 20. Again, each test run had 20 back and forth movements with a tough mark at the beginning and end.

You guessed it!!

P-E-R-F-E-C-T!!!!!! No X-Shift.

Out of desperation, I decided to make true '3D' moves - changing X, Y and Z simultaneously.

So, this set of runs started at X0, Y0, Z25 and went to X65, Y2, Z20 and then back to the start. Each feed rate had a set of 20 'back and forth' moves with a touch mark at the beginning and then again at the end.

Well, there is a GOD!!!!!

The touch marks are about 4" apart in X!!!! See attached photo.

So, this has been a LONG harangue. What did this tell me?

1.) It is not a CATIA problem
2.) It is not a servo motor / servo drive problem
3.) IT IS A WinCNC PROBLEM!!!!!

I suspected this from the beginning of this quest for truth. But, Kelly and Mark Queen at WinCNC ASSURED me it was not a WinCNC problem.

I will be talking to them again Monday morning.

Thanks for being patient.

Joe T.


First photo: 'Acceleration tests' photo NOTE 'Touch Mrk' points. ALL start at 'X=0' ONLY the X/Y/Z 3-D moves have different 'FINISHING' touch marks!!

Second Photo: Another Acceleration Test photo

Third Photo: Different orientation - sharper image

Forth photo: Multiple 'Diagnostic/Parametric' tests I have run

[Joe T]

Let me clarify the above post:

The 'X-Shift' occurs ONLY when there is a 3-D movement. What I mean by this is that the spindle is moved in the X/Y/Z axes simultaneously. If it is moved in only one axis or two axes at a time, there is NO X-SHIFT. The shift only occurs when there is a 3D move.

This has to be a control software problem. Either the WinCNC executable is corrupted, or the .ini file is corrupted, or the MACROS are corrupted, OR perhaps there is some damage to the two daughter boards - or a DIP switch is incorrect.

Remember, I am the forth owner of this machine. It does not have much use, but it has been moved 5 times!!! Anything could have happened.

AND

WinCNC does not protect its executable code, macros, or .ini files!!! ANYONE can change (corrupt) them.

Joe T.

[Joe T]

One final comment:

When I was running these 'acceleration tests,' I stood near the work offsets 'X0' location. During the single and two axis moves, the spindle came back to the X0 location after every 'back and forth' move. But, when I ran the '3D' moves, regardless of the feed rate, then spindle never made it back to X0. Each time the spindle returned to 'X0', it was just a little further int he X+ direction. Each of those shifts in X was the same increment.

The approximately four inch offset shown int eh photo above was the difference between the start of the run and the ending (finishing) position. Each run was 20 'back and forths.' If I had run 100 'back and forths,' the X-shift would have been 4 x (100/20) = 20 inches!!!!!

Joe T.

[gfacer]

Wow, good work! I agree that it seems like software.....is it exactly repeatable? I could not quite tell from your post but of yes then almost certainly software. Would not be the daughter cards or a dip switch.....could be cables if not 100% repeatable, those are probably the easiest to get damaged in a move too..... Yes, I suppose something electrical could have fried something just right but software does seem the issue.

[gfacer2]

Wow indeed. That is a lot of (great) work. Gfacer - unfortunately I don't think this is your issue at all or might it be? Its interesting though that when I think about it, my controller never does 3 axis moves - though I suspect its more like I never force my cam software to make them. Its pretty much always z down, then x/y cuts, z up, jog, repeat. Though now that I think about it some more, any cut with tabs would do it too...

I know you like to use the ramp in features more (which I assume would force mine too) which would be forcing these moves - at any rate you should definitely try and repeat the issue and Joe, you might want to post this at camaster's forum as see if they can repeat the issue on their wincnc set ups.

[gfacer]

Actually, Joe

One further test you could do is to unplug the y driver inputs and then do the coordinated 3 axis moves again. Of course, it will only move xz but the computer won't know that. If its still off then you are 100% past the drive and wires (data signal wouldn't be enough to mess anything up unless actually "shorting" to another axis' circuit).

But no, I don't think my win CNC install has this problem (exactly) as ramping does not throw off another axis.... I do think that more axis is cutting less smoothly though, which could be related....like maybe the system in either case is choking a bit on 3axis Coordinated moves. Or, still could be mechanical or resonance based in nature as that too could get worse with more axis in motion.

[gfacer]

Hi Joe,

Just tested similar 3 axis moves on my CNT:
100x moves (probably 160 as I ran part way at g1 and 100 at g0), moving from x2 y2 z0 to x50 y4 z2

No obvious change in x y or z at end of those moves.

FYI.

I hope the conversation with software guys is productive.

[Joe T]

gfacer,

I am still working on a reply. Lots of photos, etc.

I'll post in a day or two.

But, yes, the 'Shift' is very repeatable and only happens when there is a '3D' move.

Also, you said something above that I decided to address directly in my summary reply (a PDF with photos, etc.). A hint: The Backplots will show the 3D moves clearly. No 3D moves on the programs which were on the machine when I bought it. But, ALL of the CATIA programs I wrote have 3D moves. HOWEVER, I stumbled onto some simple G-code programs I wrote which did have 3D moves and they have the SHIFT.

Regards,
Joe T.

BTW,

My machine was a special build for Mr. Amit in 2004. He specified the machine to do 3D carving of rigid urethane foam to make moulds/patterns for the composite industry. That is exactly why I bought the machine. So, I have hope - - - - -

[Joe T]

All,

I am still working on a set of charts/slides that illustrate what I have done to date and hopefully show that the problem SEEMS to be WinCNC control system problem. If it is not a WinCNC control problem, then I am totally lost.

This machine has had many owners. I am the fourth owner. The third owner never hooked the machine up and decided to sell not long after they purchased the machine from the second owner. The second owner had the machine shipped from the first owner back to CNT Motion for inspection, repairs (if needed), and an updated WinCNC control software and maybe hardware - not sure about the hardware. Then the second owner shipped the machine from CNT Motion to their facility. It was properly prepared for shipment when I left CNT Motion.

I have no idea how the equipment was shipped from owner #2 to owner #3. But, I can tell you that EVERYTHING in the computer cabinet was installed as if it was ready to run. Like I stated above, the third owner never hooked up the machine. It was stored in a warehouse. So, I strongly suspect the machine, including computer cabinet, was shipped from #2 to #3 without a lot of concern for anything that might 'shake loose' in the computer cabinet.

I personally moved the machine from the third owner to my shop. I supervised all loading, securing of components, at the warehouse in Geneva, New York. I supervised all of the unloading at my shop. I removed EVERYTHING from the computer cabinet that was not securely bolted to the cabinet. This included the computer, monitor, uninterruptable power supply, master switch, etc. Anything that could be taken out was put in my wife's Dodge Durango for the trip from Geneva, New York to Winfield, KS.

Once I got the machine to my shop and re-installed the computer system, etc., got the 'European' 380V/3 phase transformer purchased and installed (I was not told about that additions ~$2,800 dollar expenditure that is unique to the HSD spindles of the time period), the machine would not work properly. It kept giving me 'Limit Z' errors.

My local CNC technician came to work on the machine and was unable to find the source of the problem. Kelly Davis of WinCNC was generous enough to discuss the problem with him and very soon he discovered the problem: A loose electrical connection (multi-pin plug) in a box. One side of the loose connection came from the PCI card. The other side of the connection went to the Teknics servo system.

I state the above to emphasis that this machine may not have been used much, but it has been moved a lot and at least once I don't think it was properly prepared for shipment. Heck, I ain't sure I prepared it for moving PROPERLY! I just did the best I could and at least removed the fragile electronics.

OK, all of hte above is a set-up for the following:

I spoke to Kelly this morning explaining my latest findings (shift only occurs when there is a '3D' move). He made some suggestions. Both may be a result of a move. He said he thought that when 3 or more axes are moving at once, the daughter board required more current. That is why there is a separate power connection directly on the board. He said he can not explain why it would just impact one axis. But, this is the first information that makes sense. Remember, I don't know anything about electronics - it bores me. Also, he said there there are three chips (two or three, already forgotten what he said) that could be faulty. He thinks if they were bad there would be no motion.

So, I have some things to do tomorrow. I will pull out and reset the PCI card and the daughter board plus all connections to them. I will also pull off and reset the power supply connector to the daughter board. If that does not make any difference, then I will order a new chip set.

WISH ME LUCK!!!!

Regards,
Joe T.

[gfacer]

Just to close off this thread....
Never did get to the re-wiring project, mostly.

I found a good deal on a chinese tool change router of a newer vintage and bought that one as a bit of an upgrade and sold the CNT locally. As part of the sale agreement, I made sure to re-wire the spindle fan that was the cause of my issues with a proper wire from the cabinet and get rid of my "extra fan on top" setup I had been working with. A small (large!) hiccup was that the HSD spindle on the machine died after the deal but while the machine was still in my shop in use by me! Ceramic bearings lasted 15 years for me, medium usage, if anyone wants to keep track. So, buyer is getting a toolchange HSD we picked up a while back just because it was a good deal and might work well with the CNT (and it does, the electrical was set up for the HP of the new spindle).

Before I post a few questions on the new machine, just thought I would wrap up this thread on the old.

[gfacer]

Just to close off this thread....
Never did get to the re-wiring project, mostly.

I found a good deal on a chinese tool change router of a newer vintage and bought that one as a bit of an upgrade and sold the CNT locally. As part of the sale agreement, I made sure to re-wire the spindle fan that was the cause of my issues with a proper wire from the cabinet and get rid of my "extra fan on top" setup I had been working with. A small (large!) hiccup was that the HSD spindle on the machine died after the deal but while the machine was still in my shop in use by me! Ceramic bearings lasted 15 years for me, medium usage, if anyone wants to keep track. So, buyer is getting a toolchange HSD we picked up a while back just because it was a good deal and might work well with the CNT (and it does, the electrical was set up for the HP of the new spindle).

Before I post a few questions on the new machine, just thought I would wrap up this thread on the old.

LOL,

Now I am really reviving an old thread.

As I mentioned here (and happened upon again), I sold the CNT. That buyer had some issues early in Covid and essentially went / got out of business. Sold the CNT and I think signed over his lease, to a buddy of his (or perhaps the machine moved one door over, hardly matters).

Regardless, I got the opportunity to see the machine up and running with the 3rd owner, who had reached out to CNT and gotten a new control PC and attached bits. Well...that seems to have helped the overall operation as he had some carving work that I am sure the machine was not capable of when I owned it.

Hurray for companies still in business and able to support them! It's probably good for another 20 years of use.