So, as per the discussion in this post: https://plus.google.com/108074515383152206476/posts/eLSbfkybgkz, whosa whatis suggested that I was getting slight platform deformation due the heater plate switching on and off. So I designed and printed a simple platform height monitor as shown. It works fairly well and could be adapted for manual leveling or similar.
Interestingly I did not see any height deviation when the heater was switching on and off even though I could clearly see 0.1 mm height adjustments using the rig.
http://www.youtube.com/watch?v=WGaBcCqUr2M
PID tune the bed and that typically solves the platform deformation without any of this measurement needed.
Sadly I don’t know what PID tuning is. Also, I did not observe any deformation with variation in temp.
I don’t thing the effect on the platform is big, however I think the effect on the filament is much more prevalent because the rate of expansion on the filament is much higher than the rate on the aluminum bed. I’m sure you can find the thermal expansion rate of the aluminum and the filament online.
so, try this, using your fixture here, print one out that only holds the stick. Then print out a tower that has a hole on the top or a U shape which will become the fulcrum. Choose 40mm or so for the height.
now you wouldn’t remove the print, just place the stick in, then ramp and cool the bed.
@Stephanie_A Great suggestion. Will do
.1mm is huge compared to the scale that I would expect to see for this error. It doesn’t take much over/under compression to cause visible artifacts in the Z alignment. A layer height that’s off by 2% should be clearly visible in an otherwise high-quality print (though of course, larger defects can hide the effect).
Really? 2% of my layer height would be .007 mm which would be less than the movement caused by a garbage truck driving by a block away. I realize this is about the distance between the nozzle and the plate but even the deflections of two 8 mm rods, the slop in the threads of the z drive etc. would have more of an effect. I guess if it was sustained for several seconds then it could produce a visible effect but to remove all other variables to verify your claim is beyond my rig.
You’re misunderstanding. I’m talking about the height of the entire layer being off by that amount. This will results in a difference in how much the plastic is compressed as it leaves the nozzle and thus how far it sticks out on the sides. It takes very little error to cause a visible line where one layer sticks out slightly further than the others, or is inset slightly if the layer is too tall.
I recorded this timelapse a while ago, it is a regular Prusa PCB with a 3mm glass plate on top:
You can see it warping by as much as 100µm as is starts heating, it then backs up about 40µm as the PID loop stabilizes. Now if this bed wasn’t PID controlled, you’d constantly have that slight up/down motion during your print.
That the plate expands 0.1mm or so when heating is expected – see http://reprap.org/wiki/Triffid_Hunter's_Calibration_Guide#Z_height
With the same volume of plastic being extruded, and a 2.0 width/thickness ratio, a 2% difference in layer height, would have a 4% difference in width. 2% of a 0.25mm layer is only 0.005mm, or 5% of the 0.1mm worth of thermal expansion witnessed in a bed heating from room temperature to working temperatures.
If you did layers in alternate directions (starting from +y or y-, for example) and had a fan blowing over the part and bed and the part differently over the layer, a 5% temperature difference in the different places of the bed or part does not seem unreasonable. See http://reprap.org/wiki/PCB_Heatbed#Testing for an IR image of image of a bed.
Thank you all for you excellent and insightful help. I will digest all this and post again when I have more to report 