Random following errors on Kflop/Kanalog/Snapamp DIY CNC

[mckoz]

Since I expect this to be a bit of a process, and have received some replies from Tom K via email, I will start from the beginning and add his replies in as able.

My system:

• Industrial Hobbies Mill CNC Conversion
• Replaced chinojunk motors with Aerotech 1135DC
• Replaced Geckos with Kflop / Snapamp / Kanalog
• Added head counterweight
• Encoders: X/Y US Digital E2-1000-375-HA3, Z axis: Avago HEDS-9100-B00, Shielded wire, drain connected to ground lug in panel only.
• NC = KmotionCNC

Here are a couple of pictures, one of my messy machine and the other of my messy control panel. (Don't bother with comments, I am not going to spend the time to make them pretty until I am sure they work).


Converted to this status back in December 2014, added the counterweight in mid-2015.

I have been using the NC off and on, and get random following errors (as indicated by the error message and console in KmotionCNC and KMotion). Sometimes I can run 3 or four different parts and never have an issue, sometimes I can't get through one part at all. Even when I can't get through the part, the errors won't occur on the same line of gcode. Up until the counterweight, the errors appeared to be slightly weighted towards more happening on the z-axis, but since then they don't have a pattern.

I have a current gcode program that simply will not complete a run, so as a result, I am now starting from scratch to diagnose the problem. The gcode program is attached with the .ngc extension changed to a .txt extension, note that it does not fail in the same place repeatably.

______________________
Step 1: I had been significantly and somewhat randomly detuning the motor acceleration, velocity and jerk since mid-2014, and that hasn't helped, so I yesterday I dusted off the books and retuned them using the step and response screen. After maxing each axis out I detuned by 50% for a safety factor. - this had no impact on the errors. Here is a spreadsheet that shows my calculations:



Note that I had success tuning the motors for 400 IPM, but detuned for 200 IPM as per my spreadsheet.

I've attached my init file, but changed the extension from .c to .txt to support the upload: HellCatInitialize-v3r20.txt.

Here is an image of my trajectory planner:


In the next post, I will attach the output of a test capture program as well as the program.

Stay tuned,
David

Similar Threads:

• Need Help!- Questions about Kmotion (Kflop+SnapAmp)
• Build Thread- Upgrading 1978 Bridgeport CNC Series I to KFlop - SnapAmp control
• Need Help!- Help configuration kflop- snapamp
• Need Help!- Axis stop moving - KFlop - SnapAmp - Mach3
• KFlop/SnapAmp - BDC Servo Beginning Tuning - Oscillations - Tracking Errors

[mckoz]

Ok, here is the test. With Tom Kerekes kind help I built a test program last year and modified it a bit yesterday. Here is what it is supposed to do:

1. Wait for a trigger, then record data for 1000 intervals.
2. The data is x pos, x dest, y pos, y dest, z pos, z dest, and supply voltage.
3. The trigger is based on the difference in x following error exceeding 200 counts.

It seems to work fine, but when the following error faults out, the data collected afterward appears to be garbage numbers. I am not a c programmer, so my understanding of the program is limited. Here are screen caps of the faults:


Attached is the c program (with txt extension), which I ran in thread 6, the results in both excel and csv.

Looking forward to hearing from you folks, especially TK and Bhurts.

Cheers, David

[TomKerekes]

Hi David,

I can't see a reason for the garbage data. Unless the upload is triggered before the buffer has been filled with valid data. When you run the program does it trigger immediately? It should wait filling the buffer with valid data many times until you make a move to trigger the error. When does the program trigger?

Regards

[mckoz]

Hi Tom, it does wait u till the fault apparently. I watched the one I posted and the thread stayed green until the fault triggered it, about 5 minutes of run time. I am running the same CNC Program every time now, and in the screen shot you can see what line it faulted on.

[bhurts]

The first thing I would try is to lower the acceleration in the trajectory to 6 or 7 ips and see if that helps cut down on faults. I would also like to see step response plots for each axis. It would be best to have two plots for each one. One at .5 inch moves and the other with 5 inch moves. Also what is your power supply volt and amp ratings. I suspect from where your at in the code that it's either acceleration or noise since your only moving 40ipm it seems. Best to look at everything to be safe. I hope it gets sorted quickly so you can enjoy making parts without worrying what's going to happen. As a after thought what version of k-motion are you running?

Ben

[TomKerekes]

Hi David,

It seems to work fine, but when the following error faults out, the data collected afterward appears to be garbage numbers. I am not a c programmer, so my understanding of the program is limited.

I see a bug causing garbage data to be printed. The data is put into the buffer 7 values at a time and then printed 7 values at a time. So the line of code that checks if it is at the point to wrap back to the beginning of the buffer:

if (p == gather_buffer + N * 4)

should be

if (p == gather_buffer + N * 7)

Regards

[mckoz]

I will try and answer in order:

The first thing I would try is to lower the acceleration in the trajectory to 6 or 7 ips and see if that helps cut down on faults

I was running the trajectory at 1 IPM acceleration up until I retuned on Sunday, with exactly the same results, so I think we can table this one?

I would also like to see step response plots for each axis. It would be best to have two plots for each one. One at .5 inch moves and the other with 5 inch moves.

Ok, I will attempt to get those done and posted tonight.

what is your power supply volt and amp ratings.

Supposedly it's 85Volts at 20 amps. Based on my experience with the "fake" original motor ratings I am not going to stake the farm on this claim however. The Snapamp is reporting 79-81VDC.

I will also make the change Tom is recommending in the code and repost the results.

Thanks for the reply, and for sure when this issue is cleared, I will be two years worth of happy....

Cheers, David

[mckoz]

Guys, had an odd evening while trying to get the step response screens for Ben. I did two things:

1. Fixed the c program file.
2. Removed the old KMotion432 from the root drive into a new sub folder called old (c:\old\KMotion432). So it was this:



It is now this:



After I did those two things, I loaded the c program into thread 6 and ran the g-code program. And ran it. And ran it. 8 times, and I couldn't get it to fail. Finally I actually ran the real part 3 times.

So, could there be some sort of environment variable or locked path in 433 that is trying to call 432? This doesn't make complete sense, but those are the only two things I changed?!?

Confused - David

[bhurts]

While retrofitting a machine for a friend I did run into a similar issue. I had multiple k-motion cnc icons. I noticed the machine would fault randomly. I got rid of all the icons and added back the ones for the version I was currently flashed for. No more issues after that. Not sure what caused it but assumed it was making calls that didn't match up with what was flashed causing the fault. Glad you got it sorted and hope you don't run into any more issues.

Ben

Sent from my HTC6525LVW using Tapatalk

[mckoz]

Thanks Ben, although after fighting this for near two years, it doesn't feel significant enough to call this a solution... Tom, what gives here, anything make sense to you?

[TomKerekes]

Hi David,

I wouldn't think that would make a difference. I assume you were having the issue before when you were not running the diagnostic program that printed all the garbage?

You might put that directory back to see if the problem comes back.

Mach3 links to the KMotion version specified in the registry HKEY_LOCAL_MACHINE\SOFTWARE\Microsoft\Windows\Curr entVersion\App Paths\Mach3.exe\Path. You might look to see what is there. If Mach3 was referencing the old KMotion Version I would expect an error trying to run after you moved the old directory.

Regards

[mckoz]

Hi Tom, further details:

I went into the registry and found multiple references to files within the KMotion432 directory. As a result, I uninstalled KMotion432 entirely, rebooted and reran the install of KMotion433. I believe I had installed 433 in addition to 432 in case I needed any files. I was calling KMotion433 only, but the registry still had entries for 432.

I believe the lesson here is you can't leave 432 installed even if you are pointing to 433.

I also strongly recommend that your install file check for previous versions and require an uninstall, as this has cost me hundreds of hours of troubleshooting and lots of wasted parts.

Cheers, David