Poor Man's Anti-Backlash Nut - No Assembly Required (Page 2 of 3)

Yup... but a float for a step value is sorta-redundant since you can't have a partial (micro)step - it will just fall back to the nearest full step anyway of a random forward-or-reverse nature...

Combine that with the fact that Microstepping is arbitrary to the specific motor - you can't tell the exact position of the rotor between steps, it all comes down to voodoo magic in your specific motor.. so there is a margin of error for every step in between full-steps that is not linear - means I'm not sure which is the greater evil here... Margins-of-positional-error based on Microstepping or the fractional loss of fidelity by converting to a whole integer..

Interesting...

And, for completeness, here is the same plot for steps = steps_true  = 3200/.5 (.5mm pitch, which has a finite decimal representation in the chosen unit of measurement):

Note that the 10^-4 became 10^-16 in the plot's y-axis (probably the limits of IEEE precision).

Anyway...

Thanks but, actually, patience in doing precise experimental measurements isn't really my strength, to put it mildly (I still have traumatic experiences from the Cavendish pendulum in undergrad physics... a bruised head from bumping into table, pissed off classmates in other sections who couldn't do anything for the rest of the day because the oscillation took forever to dampen and, of course, random results... but I digress) and I do often get carried away with "academic details" that turn out to be irrelevant in practice, so pls double check everything I say/write (which is why I post ).

Off-axis rod, that makes sense, thanks!  So... did I just get lucky, offsetting the well nut by close to 1/2 revolution by chance (provided thread_depth*sin(tilt_angle) < backlash, or thread_pitch*cos(tilt_angle) < backlash, or something like that)?  Hm... just off the top of my head, not sure...

Edit: paying close attention, the rod seems to trace a conical surface -- I think.  Which would imply that banding should improve at the top of (really tall) prints (all else equal: perhaps a big "if")..?

Edit2: Just noticed that my well nut holder has developed a stress crack.  Perhaps I was not imagining that banding has recently changed slightly, after all.  Should have marked its angular position, dammit... (Rubber is still slightly compressed, so it should still be doing it's job with backlash, I think).  Anyway.

Interesting and informative, thanks! 

And, yes, not sure what I was thinking about angular offset of nut pair (probably wasn't ). First, it would seem that only the nut attached to platform is relevant to its motion. Even if not, a linear combination of same-period(=thread pitch) functions (z = f(rod rotation), with phase=initial nut rotation) can change phase but not period (plenty of handwaving and circuitous thinking here...). Anyway..

This might help: http://www.thingiverse.com/thing:100053
(After several months, its developed tiny stress cracks -- hole distance parallel to X axis might be a fraction off? -- but still does job, so haven't bothered replacing)

I just did this to my printer, and installed it upside down from how you pictured it. However I think this causes a problem at the beginning of the print since there aren't enough threads in contact with the rod. I'm going to do a test print now.

No, that's how it's installed on mine as well (brass nut at top, which is opposite than with the older platform), but my rod is long enough so the nut doesn't come off when Z is homed. Guess I was lucky...?

Also, seeing your other pic on Thingiverse, the plate should be tightened flush onto aluminum platform (file the inset for rubber rim, if you have to, it's intentionally tight) -- but if the rod is shorter, doubt that will help.

from looking at your picture, I'm not sure you got this built right.

Is the original nut still on, under the well nut?  If not you got it all wrong.

The idea is to have 2 nuts, the one already attached to the print bed, and the well nut which has the threaded insert.  When you install the well nut, and screw it on tight, the rubber compresses and pushes against the original nut, so you now have an assembly with two nuts pushing away from each other and pushing against the threads of the rod to take up all the slack, thereby reducing backlash.

What is your layer height set to? That looks like the periodical round off error that occurs if you don't set the layer height to a a number equal to a step/micro-step.

Because you're still using the 5/16" rod, your layer height should be:

.1   =    0.09878   or   0.1058
.2   =    0.1976     or   0.2046
.3   =    0.2963     or   0.3034

you can use the RepRap calculator for other layer heights:
http://calculator.josefprusa.cz/#steppers

Hummm well I got a regular spring that fit the rod, and tighten the nut onto it when the Z was homed. Looks like about half the nut is threaded when homed. I'm not sure why my Z rod seems shorter than other peoples. The spring is crazy compressed too. See here:

Then I used a nut hugger https://dl.dropboxusercontent.com/u/841 … G_6658.JPG to hold it in place. See here:

so I'm going to see if that will work. If not then I'll try changing the layer height to try to accommodate for the potential rounding problem. I'm gonna start a test print now and then I'll post back.

The collective "we" have been using this type of a solution for a long time. It has been thoroughly tested. Generally though, add grease to the system to keep friction down.