The short motor belt is just terrible and sloppy. It is far too susceptible to error if everything else in the machine isn't perfect. Direct drive with the right coupler works even with fairly poor alignment of things. Here is the mount I am using at this time. It requires a 5-to-6 mm oldham coupler and a new 6mm Y rod (the stock one will be too short).
http://www.thingiverse.com/thing:593006

Yeah your circle is exactly what I want but:
..desperation, I passed all day on this with no results.. especially no improvement.

I tryed to tighten progressively the X from very low to "normal", I tryed to watch when the belt "twanging" and now the X seems to be clear but in the photo are the results, the bulges never changed ....:((

What is wrong..?

knowack wrote:

Here's a print of a 35mm pirate coin.  Even using my crappy white filament (I'm trying to use it all up), I'm only .1mm off from being perfectly round at this size.

Not yet perfect, but certainly good enough for most purposes.

Keep the faith!  You can achieve good circles!

Excellent pictures! Did you put a pen on your extruder or how did you make them?

I have an SD3 but I'm also working to perfect my circles right now and have it nearly dialed in.  Most of it was belt tension - the circles video helped.  some was pulley lubrication on the shoulder screws in the front idlers: 3-in-1 oil helped there for now until I can replace with pulleys fixed to shafts on bearings.  I think the rest is probably vref adjustment on the y-axis combined with speeds and acceleration to overcome the friction/sticktion during the direction change.  I don't think it's missing steps but that it doesn't have enough immediate torque, sticks, and then jerks into the intended position.

I'm sure that my alignment is good and that my belt tension is proper - vref and lubrication seem to be the next best bets.

I had to loosen the 4 screws for the Motor behind the left cover and strengthen the Belt-Tension.

Now I'm quite happy with my circles.

Hoping.......:)

If I were troubleshooting something like that, I'd first print a few layers of the circle, then kill the print.  Next lower the print bed a few mm, and use the arrows on the Manual Control tab of RH to position the extruder to those locations where the deformities occur (looks like 11 o'clock and 5 o'clock to me).  Note the X and Y position of the carriage components at those locations. 

I'd first make a visual check, to ensure nothing is snagging on something at those positions.  Then I'd start checking belts, pulleys and rails at that position (isolating them and then moving them manually if necessary) to see if you can determine where the binding is occuring.

Another possible troubleshooting method would be to print the same circle scaled up or down, to see if the problem occurs at the same X Y coordinates, or print the same-size circle off-center on the print bed (i.e. shifted toward one corner), again to see how the X Y coordinates play out.

It seems to me that if that if the problem is some sort of binding, it would occur along the same X axis or Y axis bisecting the print bed.  If there's no discernable, consistent line or location where this happens, there may be something wrong with a motor itself.

I may not be explaining myself adequately.

I tryed to tighten progressively the X from very low to "normal", I tryed to watch when the belt "twanging" and now the X seems to be clear but in the photo are the results, the bulges never changed ....:((

What is wrong..?

Not yet perfect, but certainly good enough for most purposes.

Keep the faith!  You can achieve good circles!

I printed and installed the pillowblock bearing mod on my Y-rod; I don't think it is necessary on a SD4 (I didn't notice any difference), but it certainly can't hurt, especially if you're going to really crank up the tension on your belts.

Your photo indicates to me that you have some slack in your X and Y belts.  (I had similar-looking circles, but my boxes and other geometric shapes came out fine)  I kept reading posts on here, that said to tighten the belts, but not specifically where or how.  I finally figured it out on my own.

Start with your X belt, the one that moves the extruder left and right.  Using the Manual Control tab of RH, use the arrows to position the extruder in the center of the build plate.  Now use the arrows to move the extruder left and right the maximum single step (50mm if I recall) while watching the belts.  As the extruder moves and stops, any slack that's present will accumulate in the 'return' side of the belt.  When you command the extruder to move the opposite direction, look for a shudder (like the twang of a guitar string) in the belt (the former 'return' side that is now the 'tractor' side).  That 'twang' is slack that needs to be eliminated by tightening the belt's hex screw at the bottom left rear of the extruder.  Make small adjustments to the screw, and repeat the process until the shudder is gone. 

Once you have the shudder removed, loosen the set screws of the sprockets, and run the extruder left and right by hand to even out the tension throughout the three belt segments (upper long segment between sprockets, and the left and right segments between the extruder and sprockets.  You want to remove any slack, have even tension throughout the belt, but not overtighten (which may cause your X stepper motor to stall).  It may take a few attempts to get it right.

If you do this first, the bulges in your prints should move from their present orientation of 10 o'clock - 4 o'clock to 12 o'clock - 6 o'clock.  This will indicate that the remaining slack is in your Y-axis belt.  You are on the right track.

Use the same procedure to eliminate the slack in your Y axis belts, starting with the left one.  Move the extruder forward and back with RH, and observe the belts for slack.  I recall that there is a screw front and back for adjustment.  The added complexity of Y-axis is that you need to tune two belts, and they need to be of similar tension and aligned (not keeping the extruder assembly crooked) for everything to work properly.

To summarize belt tension:  If the bulges are 10-4 o'clock, both belts are loose.  If the bulges are 9-3 o'clock, it's your X axis.  If the bulges are 12-6 o'clock, it's your Y axis.  (the reason both belts will cause a 10-4 and not a 2-8 is because Slic3r runs circles counterclockwise).

This was beginning to frustrate me, especially since my other prints were turning out so well.  My circles aren't quite perfect yet, but I'm very close!  A few more tweaks, and they'll be perfect.

I hope this saves you some butt-pain.