by cvoinescu » Sun Mar 10, 2013 5:44 pm
I have a feeling that printers do not differ substantially in the basic, static accuracy -- and with many of them being kits, these depend a lot on the skill and patience of the builder, as well as on the design. You can rate them, subjectively, on the ease of squaring and levelling -- useful information, but not easy to devise a good test for.
Some criteria have a habit of becoming meaningless (can't help thinking "capture resolution after software processing" for cheap webcams). For instance, the theoretical Z resolution of a leadscrew design is very fine -- but does it make sense to say the ORD Bot has a layer resolution of 3.125 microns? That's the smallest vertical displacement possible with a 0.9 degree motor in full step mode (with 16x microstepping, it's about 0.2 microns), but can you actually print anything with 0.003mm layers (let alone 0.0002mm), and is the result any better than 0.2mm layers? We know that the quality of the print begins to degrade below a certain layer size (the threshold being somewhere around 0.1..0.2mm in the test I read about). My point is that the parameter you can determine objectively (Z axis resolution) is meaningless, and the one you really care about is subjective and depends on too many factors.
Layer consistency touches a sensitive spot -- that's good. The ORD Bot can suffer from at least two modes of "Z wobble" (periodic vertical unevenness, and periodic horizontal displacement coupled to Z movement). An interesting metric for printer designs with allthread as leadscrew would be how much of an error is caused by a given curvature of the rod, or by a given misalignment of the coupling to the motor. Again, not an easy thing to standardize and measure, and it would not say much about a particular instance of printer.
My main point, however, is that many errors are of dynamic nature, and difficult to quantify. Sharp cornering can cause parts of the printer to oscillate, imprinting a wavy pattern into the printed object. Backlash can cause problems too. An extruder can ooze, or it can have a slow response of actual extrusion vs. motor movement. One extruder may deposit a nice, smooth line of filament, while another, although driven in a straight line, may deposit a more variable shape or quantity. All are things you see in the print, yet it's often hard to tell what the cause of a problem is, let alone measure it to compare printers.
I agree that it would be useful to have a small suite of objects that could be used to evaluate printer performance. However, simply measuring them would not be enough; evaluation will have to be at least somewhat subjective. And you'd have to study several printer designs carefully, think of what inaccuracies would affect the print and how they would affect it, and design objects that are likely to catch them, and to help distinguish between them. It's not easy!