Layer shifts after 60 minutes of printing, please help!

Hi all!
I have tried everything and still I cannot get rid of layer shifts in my prints. I wonder if I am now missing something obvious in here and would like to hear your expert opinions about how to fix this. The problem I am struggling with is a layer shift always happening half way through the print, but not at any specific point. For some large flat models this has also happened after only 10 mm printing in the Z-direction, so it is not strictly dependent on the height of the part to be printed. It is more like after 60 to 90 minutes of printing. However, all the motors are cool at this stage and the electronics are being constantly cooled by a large 12cm fan running at a full speed. All the stepper drives also have their heat sinks installed, so I would assume they are cooled well enough, although I couldn’t test the temperature because of the fan mounted in front of them.
Although in here the layer shift has always occurred on the X-axis, I have also seen the same problem on the Y-axis. It is also good to note that in some of the parts there are multiple shifts immediately next to each other and then more shifts after many perfectly fine layers.
All the models printed in the attached image were printed using the same G-Code and with the same settings from Pronterface. Print speed for slicing was in the range of 40-60 mm/s, but I always decreased it to 80% from the Pronterface interface.
I can’t exactly remember when this problem started, but it has not always been this bad. This printer has not been used very much during the past two years, mostly because I just haven’t had time to delve into this problem. After I first got it running in 2013, I had dozens of successful prints. At some point I started to have this problem and now it happens every time, at least if the printing it taking long enough time.
Any help would be highly appreciated!

Something about the setup:
- MendelMax 1.5 Printer
- RAMPS 1.4 with Chinese A4988 Stepsticks
- NEMA17 motors (1.8 deg) from Think3DPrint3D
- J-Head hotend, Printing 3mm PLA
- Slicing with the latest Slic3r
- Marlin 1.0 firmware
- Latest Pronterface on Windows 10 for control. Wlan disabled during printing.
- Electronics cooled with a closely placed 12cm fan running always at full speed.

Things I have tried so far:
- Slowing down print speed, does not seem to have an effect.
- Adding a tabletop fan for cooling as I was suspecting the print head is hitting curling top layers.
- Decreasing DEFAULT_MAX_ACCELERATION first to 2000 mm/s2 and then to 1000 mm/s2. The default value is 9000, but this is probably capped by DEFAULT_ACCELERATION set to 3000 mm/s2 by default. In the latest print I decreased both of them to 1000 mm/s2 just to be sure.
- Swapping A4988 drivers and adjusting current to 1A, what seems to be the maximum current for the drivers I have (Vref seems to go back to zero after 1.63 V).
- Adding a drop of oil for all linear bearings on the X and Y axis.

did you notice anywhere might have curl at the shift point, were 3 of them printed together or on their own? if they were printed on their own, its possible a slight curl caught the nozzle it is possible for it to skip steps

They were printed on their own, I just placed them back to the bed for comparison. The prints were generally very clean at the moment of the shift. The only problems with curling I had when the mounting point was being printed (at the front of the model), but this didn't seem to cause layer shifts at that point.

Although people prefer very high acceleration and jerk settings I prefer them to be quite low.
For example my acceleration is limited to just 100mm/s and the jerk to 40mm/s.
I basically reduced these until support structure lines came out really straight and with no wobble at high print speeds.
Low settings here mostly affect short print moves and infill patterns like honeycomb, during long lines or infills the printer still reaches the top speed.
The main benefit for me was much quiter operation and far less vibrations and noise.
Instead of banging form one side to the other the printer now just moves the print head and build table.
But in your pics it looks like you loose quite a few steps at once and you already played around with the current for the motors.
Can you freely move the axis when the motors are off or does require some strenght? If so you might need to check the alignment of your axis and belt tension.

Thanks for your replies! I was actually googling quite a bit and trying to find out why the default acceleration is so high and if there are any reasons not to lower it. For the jerk value, I think the default for X and Y is 20 or 25 mm/s, so it is actually lower than what you have for your setup.
I haven't yet disassembled the X and Y axes to see how smooth they are, but I did this when the problem first occured and they were smooth at that time.
What really puzzles me is how the problem happens only after the printer has been operating for some time. Of course this can be just a matter of probability, but I would expect that binding on the X and Y carriage movement would cause layer shifts already at the beginning of the print. In this case is seems that at some point the shifts start to occur and after that they take place every few layers.
Could this be something related to lag in the USB communication etc? I doubt this, since the problem first happened when I had Win 7 installed and since then I have swapped the HDD for a new SSD and did a clean install of Win 10. The problem was just the same in both cases. During printing I have turned off the Wlan tranceiver and disabled sleep modes and hibernation. I have also been moving the mouse every now and then just to be sure it stays awake.

If there is no stpes in the G-Code than it only leaves problem with the printer.
Do a hand test on the x and y axis: hold your hand in front to block it and let it move 10mm.
It should push your hand away.
If you hear the thing skipping and you only need very little force to prevent the axis from moving you need to check the current settings for the motor(s).

Disable steppers right after the shift. See if there is binding on your axis. Happen to me when the printer got hot and frame expanded creating the belt becoming too tight.

I'm not sure I follow this part of your post:

Quote
Rolleros
- Swapping A4988 drivers and adjusting current to 1A, what seems to be the maximum current for the drivers I have (Vref seems to go back to zero after 1.63 V).

What voltage are you setting on the drivers? From Polulu:

Current Limit = VREF × 2.5

So a Vref of 1.63V would command a current of over 4A, way in excess of what the boards can deliver. For 1A current limit you would need 0.4V Vref, The symptoms sound like the drivers hitting thermal cut off, so I'd double check those voltages and get a spare thermistor hooked up to measure the temperatures.

Thanks for all the comments.
The drivers I have are stepsticks, so they are a bit different from Pololus. To be honest I have not checked the sense resistor value from the board, but from the StepStick Wiki I found the following:

"To calculate the current, A = VREF / (8 * RS). For a standard stepstick, RS is the rating of the Sense Resistor = 0.2ohm. So A = VREF / 1.6 "

So the maximum Vref of ~1.6 V would give me a current of around 1A. If this is correct, it sure is a bit strange, as the maximum current should probably be around 2A. If the maximum current is indeed around 1A, do you think this should be enough for normal printer operation in most cases?

I now detached the belt from the X-axis and the carriage was moving really smoothly. A gentle push made it slide all the way to the opposite end, so it should be fine. I also tried to resist the movement of the X carriage by and and it seems the entire printer would sooner slide on the table than the X motor would stall.

I also lowered the DEFAULT_MAX_ACCELERATION value and DEFAULT_ACCELERATION to 200 mm/s2 for both X and Y. This really made a huge difference in the sound of the printer and it is now operating really smoothly and without any vibrations. The estimated print time for the model increased from 2h 30 min to 3 h, but it doesn't matter as long as it helps. The printer is currently printing, so I will know more soon.

I am not completely sure I understand the acceleration settings in the Marlin config. Is the default-acceleration always limited by the default_max_acceleration and why there are separate values for the maximum accelerations but only one general value for the default acceleration? In the following situation the acceleration value for the X and Y moves would be 200 mm/s2, but how would it go for Z and E? I am also puzzled by the comments for the retract values. I cannot see how the filament retraction settings would affect X, Y or Z movements.


"#define DEFAULT_MAX_ACCELERATION {200,200,100,10000}

#define DEFAULT_ACCELERATION 200 // X, Y, Z and E max acceleration in mm/s^2 for printing moves
#define DEFAULT_RETRACT_ACCELERATION 1000 // X, Y, Z and E max acceleration in mm/s^2 for r retracts"

Ok, I won't be perfect but I will try to explain...
The default max acceleration limits the general movement, that is for all non print moves and normal moves.
The the default acceleration is for all printing moves, e.g. when filament is extruded.
The default retract acceleration regulates how fast the extruder speeds up, going a bit low here can help if your filament get's chewed often.
The default travel acceleration is for all non printing moves.

So basically the first value is a genearl setup while the others do the "finetuning" for the different types of moves.
You can set the first and leave the rest untouched to override them through the slicer but I prefer to set them all to a value that allows for really good results.
As you noticed the print times go down a bit, but the noise and print quality make up for it.

Thanks for the explanation!
It seems that lowering the acceleration values to 200 mm/s2 solved the problem, as I got the part printed out without layer shifts. I am just wondering if this behavior is really a sign of some kind of problem and going to the low acceleration values just overcomes the problem but does not fix it. I have to see how other models would print out and double check everything once more.

Both I would say
If the hardware is a 100% up for the job and speed you can put the jerk and acceleration to the max values.
Sadly in our real world we have to fight with inertia due to the mass of moving parts, flex, tolerances ,vibrations and more...
In your case the motors are not capable of stopping and starting the axis without loosing a few steps.
This can be due to current limitations, friction, type of motor and so on.
But keep in mind that in part the acceleration can compensate for this.
Instead of forcing the motor into extreme torque situations requiring a lot of current the movement is reduced to a speed where the motor can handle the mass and movement changes.

Crank your print speed up to the point where your print loses a bit of quality or the layers no longer stick properly.
Use something that requires support and keep the support between 1 and 2mm spacing.
Watch the support lines!
If they are still straight increase the acceleration a bit until you see a slight wave pattern forming after the 90° turn.
Do the same for the jerk but if in doubt use a honeycomb infill for this.
You want (at highest possible print speed) the support and infill lines to be just completely straight without a wobble or overshooting corners.
Once you have found the right values that work good for you reduce them by 10-15% (whatever gives a nice round number) to be on the save side.

Sure your print times go up a bit but I think quality comes over quantity when it comes to printing.
After all a lost print wastes much more time than a good print that takes a bit longer.
I also noticed that the longer print times are ofen only happening with complex parts.
For example a simple, straight box still prints really fast but an air diffusor or impeller with a lot of details might take twice as long as predicted in the print program.