When building your printer, don’t adjust X/Y/Z esteps to change distance traveled. Instead, do the proper calculations here: http://calculator.josefprusa.cz/ based on your setup, use those. – and then fix the mechanical portions of your printer until it’s correct. So many times people calculate their own number, and it’s 5%+ off, and all it serves to do is mask a hardware issue. Obviously this does not apply to your extruder.
It doesn’t apply to all XY drives either. I’m using Spectra fishing line on motor shafts on my Printrbot Simple.
Also applies to spectra lines. It’s just that Josef Prusas calculator uses belt pitch and tooth count to determine pulley circumference.
There is too much variation with line thickness & shaft diameter v line tension, and the effects of the line moving side to side on the pulley. It’s never going to be mathematically perfect.
Checking my values, I have about 2-3% difference between two axes using identical motors and line. It was a far bigger difference when I still had the nylon tubing on the shafts that it comes with.
That makes me want to not use Spectra.
@Dale_Dunn Well, it’s obviously a cheaper and inferior alternative to belts.
I dont understand why it obviously doesnt apply to extruders. I have had very good success measuring the smallest point of the drive gear pitch and calculating from there.
Is there a similar thing for Deltas ?
@Brian_Evans it does, somewhat, but extruders vary a bit more. Tooth bite, idler pressure, material, all change your proper esteps enough that I can’t recommend it for extruders.
@Shachar_Weis , Z axis for deltas would be what you test against.
Technically, the E steps should be calculated from the effective circumference of the drive gear. The problem is that the factors @ThantiK mentions, especially tooth bite for a given profile, make the old style empirical measurement the best way of measuring the gear’s effective circumference. If the gear was precision machined as some are these days (ours at Deezmaker are individually cut at a regular interval and to the same depth with a thread mill on a 5-axis machine), you should be able to count on the whole batch having the same effective diameter, but if they are made with a hob (or worse, by using a tap as if it were a real hob), there might be enough variation from one part to the next to make it worth re-measuring.
Spectra drives suffer similar issues. The big difference between using spectra and using something like a toothed belt is that the teeth (in addition to preventing slipping) essentially quantizes the position to a regular set of possible values. If the pulley has 10 teeth, and those teeth properly mesh (or even just come close enough for the teeth to re-settle before it can jump a tooth), then you know that every full rotation of the pulley will move the belt by 10 teeth. If you know the pitch of the teeth, then you also know how much linear motion that 10-tooth movement translates into. Threads work the same way. Spectra doesn’t have teeth, so it lies against the shaft wherever it pleases, and if it’s stretched or loose, or if the shaft has a diameter slightly different from the nominal value (or worse, one that is compressible as on the printrbot simple) or the ratio of the thickness of the line to its bend radius around the pulley causes the surface in contact with the shaft to have a length different from that of the straight line resulting in a change in the shaft’s effective circumference, there’s nothing to force it back into the expected value as there is with a toothed belt and ensure that any error in position below a certain threshold is of the non-accumulating variety (btw, the internal physics of stepper motors have the same effect when translating electrical signals into shaft rotation).
So spectra drives do need to be measured empirically. I will disagree with @Jasper_Janssen that the walking of the filament along the shaft is something that should be adjusted for. This effect is not constant across the linear axis’s range of motion, so it cannot be compensated for with the adjustment of a multiplier, and it falls squarely into the category of mechanical problems that need to be fixed to make the machine’s actual motion match the math, rather than vice versa. The only good solution I’ve seen to this particular problem is the idea of using a multi-rimmed drive pulley with an extra idler to get the multiple wrappings required to produce enough friction to prevent slipping without causing the line to walk along the pulley, though it seems that it should also be possible to fix it by allowing the pulley to float along its axis of rotation, or by using an idler pulley that walks the filament same way the motor pulley does, and using a floating attachment point on the carriage.
@Shachar_Weis , all of this also applies to a Delta, though it has to be applied before the delta coordinate transformation. The dependency of this coordinate transformation on the exact length of the effector arms (which are usually homemade to levels of precision that, while they may be impressive given the tools used, wouldn’t cut it in a precision machine shop) adds an additional layer of uncertainty, and cannot be adjusted-for with a simple multiplier due to the non-linearity of the effect.
The problem with measuring anything, like how how far the filament feeds through the extruder, is in the accuracy and repeatability of the measurement. If you are marking the filament every 10 or 30mm your mark is often a mm or more in thickness so how accurate is your measurement? That variation is going to be more than the variation I get from the hardnesses of various plastics and how much the drive gear bites into the filament. From tapped bolts to machined drive gears Ive had much better luck measuring the OD of the drive gear with calipers than I have measuring the feed lengths going into the extruder. I remember doing that with 15 students and a MakerBot Mk5 back in the day - it seems arcane now. Just seems like if you are properly calculating steps/mm you should do that across the board.
And with this Spectraline will never occupy my mind again. (Unless I pick up fishing) 
@Brian_Evans You are correct about the importance of accuracy and repeatability. That’s why I would always do the filament measurement by measuring from the edge of something, like a piece of tape around the filament, rather than from a line, and also why I would never do such a measurement with less than 100mm or do the test only once.
Trying to measure the diameter of the drive gear directly has the same problem. If you’re using calipers to measure the drive gear, it’s only going to be about 10-11mm for a common drive gear, and less than that for a hobbed bolt. How accurately can you measure such a small diameter? I would wager that I can measure the length of a ~100mm piece of filament with a smaller percent error. Furthermore, I don’t see how you would account for how much the teeth sink into the filament with your method, whereas it is an inherent part of what is being measured with mine.
@Tom_Oyvind_Hogstad As I said, it is not in any way a better solution than belts (even the crappy ones). What it is is cheaper.
I see no point in saving $10 on my DIY printers. If you sell printers and really need to push the price down, then maybe …
@Tom_Oyvind_Hogstad it’s quite a bit more than that depending on the printer. Pulleys can run $10-$15 each. So it’s probably closer to $30, more if you have a belt-driven Z.
RobotDigg.com sell GT2 pulleys for $2 each, and belts for $2 per meter.
@Shachar_Weis holy crap, nice.