I’m having a recurring issue with CNC jobs on the A350 where at some point it loses it’s position and then continues along with the program with some random offset. Generally speaking, I see this the most when trying to do profile cuts - this is the first time it’s happened on a finishing path.
Do you observe the machine while this is happening? Potentially the load on the machine at some point is too high and the steppers skip steps. This would lead to a loss of correct position. Same might happen if your feedrate is too high, although SM2 firmware limits should prevent this.
As you can see from the example image, there was no particular load (no more than the previous 90% of the job) especially noting that I’m running the 200W toolhead. Also, I’m really conservative about the feed rate on the Snapmaker so I’m not pushing it hard at all.
I have seen this in person where a running profile cut that just forgets its position and drives off as if it missed a bunch of instructions and jumped to another part of the job without going through the movements to get to the correct position.
That seems to be the core of the problem in these cases - it skips some instructions and then continues on, but is not in the correct position with results like this. Is there any way to extract the logs to check that actions in the file against the actions performed by the toolhead?
Feeding your GCode into this GCode-Viewer it looks completely sane and I cannot see anything that would explain the behaviour. There are no coordinate system changes, which sometimes confuse the Snapmaker etc… I have seen the SM2 loosing its coordinates after you pause and resume a job, but it does not sound like you’re doing that.
When it happens: Does any of your linear modules feel particularly hot? It may be that the stepper driver overheats and goes into protection mode. There have been cases where the silicon heat transfer cushion has been badly mounted or not at all, but admittedly I would have expected that the problem occurs faster then.
I hadn’t thought to check the linear module temperatures, but I see this happen with smaller jobs as well (profile cuts of 12 passes of a ~180cm circuit), but even this one had managed two previous passes and 95% of the current pass without issue, plus the A350 is in a fairly cool basement so overheating seems unlikely.
Hmmm - not knowing exactly how the internal communications works between the components, I’m not sure about this, but I think that I may have a working hypothesis. Somewhat related to the stop button thread that got worked in here, I’m thinking that individual controllers (X-Y-Z) have some small buffer that is updated from the control screen.
I also wondered about the sensitivity of the CAN Bus to voltage drops - specifically if I fire up my shop vac to clean up sawdust from time to time (and it’s on the same circuit). My thinking is that in some cases, the voltage drops is enough to interrupt the filling of a controller buffer, but not stop the overall workings and so there’s a gap in the data stored in the buffer of one or more controllers which gives me the “rogue” behaviour since it was missing some move instructions and then keeps on going with the commands that it’s receiving.
I’m currently running some fairly long jobs, but this time I’m using a battery powered vacuum instead of the shop-vac for sawdust collection. It’s not a conclusive test but may be a useful indicator. I’m also thinking that I more frequently had these issues with profile cuts where I would would use the shop-vac to pull sawdust out of relatively deep cuts and that would load the motor (you can hear it ) when I would place the suction end flat on the wood which would draw more power.
Your guess is as good as mine - I admit that I would think voltage drops of that kind would crash the machine, not disturb it this way, but it’s only a gut feeling. And, if your vac sends out a lot of electric noise, it may indeed cause bit shifts in the communication in theory - again, my gut feeling says that this would cause more diverse effects. Your error seems too similar/conclusive to be the result of stray events.
That said, another question: Is always the same happening, i.e. the toolhead always going a bit to the left? Is it always in X direction? Or is it always in the X/Y plane? Did it ever go off in Z direction? This may give additional food for thought if you could spot patterns…
No patterns - this makes sense if it’s a communications issue and dropping stepper motor instructions rather than an innate mechanical issue. I can’t see the system inventing commands to insert which would give a specific pattern either.
