Odd circle test results (Page 19 of 24)

seems most SD4s and some SD3s didnt have them glued causing this issue. Solidoodle quality control has dropped drasticly the last 2 years since I got my SD2 by all the topics I have seen this year.

Did your parts make it yet?

I just checked mail yesterday morning but didn't see them.  Do you have the tracking ID #?

Those circles look pretty good to me.  Since it's easy, I suggest now trying to see software backlash compensation can eliminate the rest:

http://www.soliforum.com/topic/313/hysteresis-fix/

Hmm.  Here's how wikipedia defines "hysterisis":

"Hysteresis is the dependence of the output of a system not only on its current input, but also on its history of past inputs. The dependence arises because the history affects the value of an internal state."

I guess in this context the input is the y position specified in the gcode and the output is the actual y-position of the print head.  If you have backlash, the output depends on whether previously the head was higher or lower (which direction it's moving) so in that sense it can be called a form of hysterysis... but yes, "backlash compensation" seems much more specific and descriptive.

An it looks like you can modify the value with gcode (m99) and presumably save it to flash (m500) without reflashing the firmware.   Too easy not to try.

The value is in mm, so you can calculate the backlash from the length of the flattened section and the diameter of the circle it's on.  I think it's something like (I might make a math mistake):

R = radius of circle (mm)
L = length of flattened section (mm)
E = backlash (mm)

(R-E)^2 + L^2 = R^2   (from pythagorean theorem)

Solving for E

E^2 - 2 RE + L^2 = 0

a = 1
b = -2R
c = L^2

E = (- b - sqrt( b^2 - 4ac )) / 2a   (quadratic formula)

E = (2R - sqrt( 4R^2 - 4L^2 )) / 2 (plugging in values)

E = R - sqrt( R^2 - L^2)   (simplifying)

So if, for example, for  R = 80mm, and L = 20mm

E = 80 - sqrt( 6400 - 400 ) = 2.54mm

Correction: rethinking the geometry I think I was off by a factor of 2 on L.  It should now be correct.

Very interesting. Thanks both of you.
So, I need to edit the gcode manually for each print?

When you have the belts tightened, and you press down on the belt, it deflects, right?  where do you think that deflection comes from?  t's not from the end points moving, it's from belt stretch.

I think M99 hysteresis settings save to eeprom... Should check that though. Otherwise, you can add the command to your "start code" in either R-H or slic3r (I prefer the latter - it's stuck in the file for things like octoprint that way!).

@Tealvince
I don't mean to dispute your theory, I anticipate the results highly, but I have a hypothesis to add:

If you only fix the location of direction change, the 12-2 O'clock side will still be scewed.  Is this thought valid?

not sure what the numbers starting from line 366 are though

Why? Isn't there some slack on the pulleys when reaching both sides of the circle, just the opposite "side" of the pulley?

I'm guessing the numbers coming after the 365th line tell how many mm of backlash you have for the given angle or relative position of your 'dents' or something. Someone just linked to that file in another forum, I couldn't find an article or something where the data in that file is used and explained. But the circle illustrations are helpful for newcomers to determine if they have backlash in x or y.

------------------------------
OK, so as requested:
How I replaced bushings with bearings.

1) Okay first of all, it's not as simple as getting the bushings out, then putting the bearings in. The bushing diameter is around 12.8 mm, the bearing diameter for me was 12.94 (advertised as 13). The holes the bushings go through are even smaller than the bushings (I think it was 12.3mm), so getting a 13mm flanged bearing in place is not easy. The fact that the one end of the bearing doesn't get thinner which is the case with bushings also doesn't help.

2) In some wiki videos you can hear someone hammering something in the background in the Solidoodle factory. After taking everything apart on my printer the only thing I can think was hammered were those 12.8mm stock bushings in those smaller holes on the frame.
To not damage the stock bushings, I put a soft wooden stick on the bushings from the inside of the frame and started hammering it until the bushing fell off. The same on the other side. Now I had around 13mm bearings I had to get in a 12.3 mm hole. I decided to make the holes larger with sandpaper instead of hammering it in and risking to damage my bearings and the frame.
I was really careful to sand the holes evenly on all sides.

Same for the hole in the pillow block, which for me at least could be removed by hand.

3) I put the 3 flanged bearings in place, slid the rod in them. Then removed the screws of the pillow block and started to rotate the rod so everything would "fall in place" and I could make sure the rod is rotating smoothly.

4) After that, I applied 2 drops of liquid superglue to the top part of the walls where the lipped part of the bearings were. The glue leaked down, coating the intersecting walls of the lip of the bearing and the wall. Just two drops was enough to secure them in place for good and enough to not make it impossible to remove afterwards if needed. Make sure you don't get any glue inside the bearings, just the sides of the lips!

This way I could get a printer I could tighten the Y motor belt as much as I could without getting any binding as before.
You'll end up with larger holes but if for whatever reasons you'll ever need to put the stock bushings back in, you can probably just superglue it in place, it sticks really well, and superglue can be removed with acetone.

Hope this info was helpful. I'm not sure this is the best way to do it, just telling how I did it.

Photos:

left frame:

right frame:

stock SD4 pillow block:

Stock SD4 pillow block seemed fine to me but I think someone was complaining about it. You can always 3d print one.

How the printer looks now:

And an example (though poor quality blurry) photo of an issue arising in real life prints from backlash:

Holes between infills and shells. It's a lot thinner for me after this mod and tightening the Y motor belt some more, but they are still visible holes. Just one example.

Haven't had time to try the possible software solution yet.