Marln/Sfact/Pronterface wiered callibration problems

Hi Everyone,
I finished my Prusa SAE last week and am having a hell of a time calibrating it.
I am using RAMPS 1.4/Mega 2560 running Merlin firmware with 1/16 pololu stepper drivers.
The default axis_steps_per_unit in the configuration.h file was not working. So, I took out my dial calipers and after leveling the bed via the spring loaded screws, I dialed the x,y and z axis so that when I press x/y/z +/-1 jogging buttons in pronterface, I move 1mm +/- (0.05 worst case on z axis and better resolution on x and y).
With those settings, when I try to print, it starts out ok but after 3 or 4 layers, it starts printing in the air!!
I then reduced the numbers for the z-axis in configuration.h file, until I got descent print layers. After using the sfact calibration routine, when I printed the 20mm box walls, it came out to 22mm on each side +/- 0.02mm but the right thickness of 0.5mm. The box is bigger by 2mm!
My dial caliper is accurate to 0.001" (it's on inches which requires the usual multiply by 25.4 to convert it to mm.) my caliper is digital so it has a mm setting which is nice.
The original axis_steps_per_unit in Marlin was {104.987, 104.987, 2272.7272, 1487} for Makergear Hybrid Prusa
my new settings are {62.4983, 62.284, 1353, 585.878}

What am I doing wrong? I am baffled.
Thanks for any help.

Edited 1 time(s). Last edit at 12/19/2011 02:10AM by pic1832.

I'm not familiar with the Makergear hybrid Prusa so I don't know what pulleys, belts and lead screw it uses. It's better to set the steps per mm for the x, y and z to calclulated values. For the x and y, it's simply the (number of steps per revolution)/(belt mm per teeth)/(number of teeth in pulley). For the z-axis, it just (number of steps per revolution)/(lead screw mm per thread). This might need to be adjusted for shrinkage but as Nophead as pointed out, it's better to adjust using large parts since the dimensions of small parts would depend more heavily on the extruder calibration. If we know the thermal expansion coefficient of the plastic (ABS or PLA) being used, then this shrinkage corrections can be calculated.

Edited 3 time(s). Last edit at 12/18/2011 01:34AM by brnrd.

If I'm not mistaken, Rick sometimes changes components shipped with the kits. I also seem to remember that he likes to run at 1/8 step instead of 1/16. I have also heard that he has sold some kits with 0.9° steppers instead of the more common 1.8°.

The axis steps you find are all very roughly a factor of two from Marlin's MakerGear defaults. My guess is you have different microstepping settings, different pulleys, and different filament diameter from the machine that was used for the Marlin defaults. Go with what works on your machine.

When calibrating the steps, measure a large real movement of each axis, instead of a printed part. I used 100mm moves to keep the numbers simple. ±0.05 on a 1mm move is a very large error (5%) compared to ±0.05 on a 100mm move (0.05%). Even though the caliper reads in 0.05 increment, your measurements are probably not that accurate due to a number of things having to do with using calipers. The large move spreads the errors out, reducing their effect on the calibration.

Once all three axes are moving accurately, size deviations in printed parts should be corrected with print settings.

measure 4 times cut once... no really tell your x axis to move 100mm from home measure how far it really went
do it again three more times... now math it up... look at your notes one of the measurements will be out a bit more than the rest,
a flyer one might say not close to the shot group throw it out... .
with the remaining three add them together thence divide buy three plug the result into your calibration
math .then save in your firmware.
home the X axis and tell it to move 100 mm again measure again at least three times repeat as needed.
do the same for each axis but do the math for your extruder and only check that you input the ratios correctly.
by measuring with cut straw and mark on filament .
so wiggle in with your measurements .

This all seems like a waste of time since you already know the pitch of the teeth in the belt and the number of teeth in the pulley. I'm curious what the difference between the calculated versus the measured values is.

First off, this is not a makergear kit. MG was the closest to what I wanted in configuration.h
Second, I used a dial caliper which is the most accurate way of moving in a milling process which can compensates for backlash. (not to confuse this with calipers). did it multiple times and average them out.
It turns out, that the square block were indeed 22mm with 5mm perimeter walls. I imported the .stl file in to solidworks and used the measuring tool to see the dimensions, and long and behold, it was 22mm!
The calculation as brnrd suggested, was 64 and my measurements yielded the value of 62.4983 and 82.834 for x and y.
These printed pulleys and XL 5mm belts are totally wrong parts to use for the kind of precision required. So, I ordered manufactured plastic pulleys and GT2 belts from Stock drive products and the part numbers are;
A 6Z51M036DF0605 {GT2 36 tooth timing pulley $5.84 each}
A 6R51MC060 {GT2 Drive belts (2 meters) @ $9.71/meter}, in case others would want to upgrade.
At the time of this writing, they list them as not in stock, but when I called their customer service, they actually have them in stock but not showing on their site. it was shipped the next day. Why list them as non stock? I have no clue neither did the rep!
I know the pulleys will fit the y-axis but the x-axis might need some filing of the plastic.
Thanks for all your input.
With these crappy pulleys and belts, I get a nice output and am looking forward to see what the upgrade will bring.
Will post some pictures this week sometimes.

Edited 2 time(s). Last edit at 12/19/2011 02:16AM by pic1832.

I'm glad things are going well, but I wanted to clear something up.

pic1832 Wrote:
-------------------------------------------------------
...dial caliper which is the most
> accurate way of moving in a milling process which
> can compensates for backlash. (not to confuse this
> with calipers). ...

Do you mean a dial indicator?

Yes...Dial indicator.. what was I thinking?