Z-Axis getting higher and higher during prints?

Mine is a repstrap but I think someone might have encounters this already too...
It is converted from a homemade CNC. The Z-axis assembly is mounted on the X-carriage with ballscrew / smooth rails.

My problem is - during printing, the Z-axis is getting higher and higher. Resulted in a object which is higher than specified, and poor print quality (gaps between layers)
I checked mm/steps settings, it's correct.
In fact I am sure the Z position changes even during the same layer, as there is set screw in motor-shift mount which makes the position changes obvious.
If I pause the print, homing and go back with G1, it will goes back to the correct level. (My endstop are mounted at the top / max position)
I am using lift Z for retraction. I bet sometimes after lift it doesn't come back to the original position.

I have 320steps/mm (ball screw + 1/8 micro steps), layer height 0.4mm, lift Z 0.5mm, so it's all divide cleanly even if I don't use microstepping. (1 full step == 0.025mm)

Hm. Maybe I should try disabling lift Z before asking? Couldn't think of this before writing this up.

I hope I could start printing functional objects anytime soon. I need to print myself a filament spool before going to print seriously, haha. And then printing a reprap. Only one more axis!

Are you saying the Z axis moves when it should be stationary? If so it seems like noise on the step signal. What sort of electronic set up do you have?

---

[www.hydraraptor.blogspot.com]

Ah no.
Not that it moves when it should be stationary, but when it does (like after Z-Lift), sometimes it doesn't go back to the original layer height X but a little bit higher (X + unknown delta).

Stepper skipping steps? It should sound obvious but I don't hear that happening.

I am adding re-check motor-shaft coupler, and double verify max speed / acceleration / jerk to my todo list.

There are obviously a limited number of possibilities.
Missing steps during the retract being the most obvious, I'd try running some simple move the head up and down tests, and see if it's accumulating error.
You should also measure the Z backlash, it it's large it can cause issues with Z retract, though It wouldn't cause a significant error I overall height.

You can always try printing without the Z Retract and see if the issue goes away

Backslash is minimal and yes the overall height in my case does changed.
Printing without retraction is nice enough.

I did thing like

G91
G1 Z0.5 F9999
G1 Z-0.5 F9999

for hundreds time. Position is accurate.

No idea what went wrong yet.

Comparison - Left with Z retract. Right: without.

After a week or two at least it's printing something functional - based on a few years of community's effort.
I have been following the reprap info early on. So much improvement since Darwin. Kudos everyone!

Edited 1 time(s). Last edit at 07/29/2012 02:32PM by sam0737.

Given the large difference in overall height, I'd say that the stepper is stalling as it tries to move back down.
It's interesting that the test on Z didn't show it though, might be noise related during the print other things are happening as the retract ends, notably the extruder fast forwarding at the same time.

Ya-let me try to combine other axis in repeated-movement test.

I wonder would that be the blub of plastics that prevent the printhead from moving down?

I can't resolve this yet. It doesn't seem to be the signal noise though.

Well, it doesn't hurt much without retracting Z.
At least I got some working prints from it already - With a working slug fit dovetail joint I bet it could print a new extruder, and a Mendel 90 from it.

> What sort of electronic set up do you have?
I think I missed your question, nophead =) Mine is a Arduino Mega + handmade fan out board + MOSFET for heaters, connecting to a 3-Axis stepper driver designed for CNC with parallel port interface, and another standalone TB6560 driver board for the Extruder.

Some CNC stepper drivers don't like the short pulse that the reprap firmwares tend to provide.
You might want to try adding a delay at the point in the firmware where up it sends the pulse to make it longer.

Not only do they need longer pulses, they also need more time between changing the direction signal and the first pulse (setup time).

---

[www.hydraraptor.blogspot.com]

I found that the same thing is happening on X and Y too. A particular pattern will trigger a particular amount of shift. And the shift is quiet consistent. (printing the same thing will get the same distortion)

And I think found the root cause and nailed it.

The problem is the opto used, it's TLP521-2. The stated turn-on time is 2us, and stated turn-off time could be as long as 25us. (BTW the driver chip TA8435H itself is min-1us for CLK/DIR pin)
So hack to the stepper.cpp is needed. Considering after a H to L direction change, I modified added 30us after DIR pin changes and before STEP pin changes. Also 3us delay between switching of the STEP pin.

Hm. I don't feel good to add that many delay in the interrupt service routine, but it does the job so far.