Thanks to all for posting the resources here and on the wiki.

Rincewind wrote:

I created a wiki page on this: http://www.soliwiki.com/Motor_skipping_steps

Excellent!  And should the procedure and circumstances where one might want to isolate the Sanginololu from the USB power be mentioned here as well?

Good point! I will add it (as soon as I finish compiling the list of compatible hardware I'm working on ATM )

I created a wiki page on this: http://www.soliwiki.com/Motor_skipping_steps

Excellent!  And should the procedure and circumstances where one might want to isolate the Sanginololu from the USB power be mentioned here as well?

That's true, I guess it shouldn't skip from that error.  However, I have accidentally interrupted power to the machine once, and the print was shifted by a distance proportional to the time in which it lost power.  This is a different problem, but it may be some use in rare occurrences.

Ah, actually this situation is very interesting because there is a very logical explanation for it: the electronics stays powered by the USB but the motors don't move because they need the 12V. So from a firmware point of view, the machine is still printing fine, except it's not.

This I'm not sure. A baud rate mismatch would just make the printer fail connecting, and an error in the serial communication would be detected by the firmware with a line number of checksum mismatch.
I don't think it can cause skipping steps. Or do you have a different experience?

That's true, I guess it shouldn't skip from that error.  However, I have accidentally interrupted power to the machine once, and the print was shifted by a distance proportional to the time in which it lost power.  This is a different problem, but it may be some use in rare occurrences.

This is very good to know.  We should compile a list of problems that cause layer shifting / skipping.

Thanks for the list. I think it is very important. I was also thinking about something along these lines

This I'm not sure. A baud rate mismatch would just make the printer fail connecting, and an error in the serial communication would be detected by the firmware with a line number of checksum mismatch.

I don't think it can cause skipping steps. Or do you have a different experience?

I also noticed you have to set the trim pots with the PSU power plugged in NOT JUST the USB power I found out it makes a big difference.

And I also noticed after all my fixes above, I still had shifts. I realized I might have a small clog in the nozzle. I heated up retracted the filament then pushed out burned plastic with a guitar string. And the next print when fine. I think the clog was causing the filament to come out at odd angles and either shift or drag across layers especially small ones

Simple fixed but immediate difference.

So far, I know of:
-Belt tension too tight or too loose, or uneven
-Inadequate lubrication for guide rods
-Set screws on pulley collars loose
-Stepper drivers overheating; require cooling
-Trimpot voltages incorrect (either stock, or caused by USB voltage variations)
-USB communication dropouts due to physical connection, software, or baud rate mismatch

I wanted to share the experience I recently had with a motor skipping steps/layers shifting.

I have a solidoodle 2 that worked perfectly when attached to a computer, everything calibrated, smooth action. When attached to a different computer (a tablet, through a USB OTG interface), the print would work fine for a while, and then start skipping steps (in the Y direction, in my case). And they were not one or two steps either, sometimes a complete movement in the Y would be ignored and for a diagonal move, only the X component would be executed.

It took me some time to track down the problem: it was not mechanical, it was not the serial communication... In the end it turned out that the problem was with the USB voltage.

Apparently, the OTG interface had a voltage supply a bit higher than 5V, and this was shifting the references throughout the Sanguinololu board. In fact, the Y trimpot voltage was 40mV higher while connected to the tablet than when connected to the PC.

In the end, I solved the problem at the root by cutting the 5V supply from the USB as suggested here.

I think that adjusting the trimpots would have also worked, but that would have meant a readjustment every time I changed the computer driving the SD, which I didn't want.

This is NOT the only and not even the most common reason for layer shifting, but I think it was worth sharing because it was a tricky problem. And I guess the take-home message is: the trimpot voltages can change when connecting the Solidoodle to a different computer.