Still More Sand Table Updates

Another Problem

The Spice Must Flow's new, quieter mechanism was great, but I kept having a problem with it.  It would run OK for a while, then the mechanism would bind and the pattern would shift while making a terrible noise.

Then I saw a link to something interesting while looking at the Duet forums:

https://www.machinedesign.com/motion-control/linear-bearings-understanding-21-ratio-and-how-overcome-stick-slip-phenomenon

When you read an article like this, it can be difficult to apply to a real world situation where you don't know any of the forces or the coefficients of friction.  But even without numbers, the article provides useful concepts.  For example, keeping friction as low as possible is best (duh!), and a "longer" bearing is better than a shorter bearing.  Minimizing slop in the bearing will keep the carriage from tilting and binding.

In The Spice Must Flow, the Y axis worked fine, and can even run an erase pattern consisting of mostly Y parallel lines at 1000 mm/sec without any issues.  But the X axis was still problematic.  On certain patterns, the machine would run OK for a little while and then suddenly the ball would abruptly change path accompanied by a horrible noise.  I took the sand box off the table and watched the mechanism run for a while to try to see what was happening.  I noticed that the magnet carriage would wobble as it moved and then tilt a bit right before the noise and path change.

I redesigned the magnet carriage yet again, going back to a printable design that uses eight PTFE inserts as bearings so the printed plastic never contacts the X axis guide tube.  Four of those bearing inserts are fixed and the other four have screws behind them that allow adjustment of contact pressure or play against the X axis tube.  The bearings are spread apart to increase the stability of the whole carriage on the tube.

The new carriage uses 3 mm diameter steel pins to anchor the belts, spaced to match the pulley spacing on the ends of the X axis to ensure the belts remain parallel to the guide tube.

The new magnet carriage design uses 8x PTFE inserts acting as bearings, and uses the same spring and magnet as all the previous designs.  Screws that push on the PTFE bearings allow adjustment of play

I milled a 15 x 8 mm teflon strip, then cut it into 7 mm thick pieces using a Japanese pull saw with very sharp teeth and narrow kerf.  If you've never used a Japanese pull saw, I urge you to try one.  You'll never go back.  I milled the individual blocks to final 5x8x15 mm dimensions.

The new magnet carriage printed in PETG.  Screws hold the cover down, and 

The two halves of the new magnet carriage design.  There are slots for eight 5x8x15 mm PTFE bearings, four of which have screws to adjust their contact pressure or play on the X axis tube.  3mm steel pins to anchor the belts fit into the round holes next to the bearings.

I installed the new magnet carriage and it almost fixed the problem.  The mechanism still occasionally stalled on certain patterns.  I concluded that the mechanism is operating right at the torque limit of the motors.  The 1:5 speed step-up made things run quietly but gave up a little too much torque.  I decided that I needed to get some of the torque back by reducing the speed step-up from 1:5 to 1:4.

The 1:5 speed step-up was accomplished by putting 80 tooth pulleys on the motors driving 16 tooth pulleys on a shaft that spins the 40 tooth drive pulleys.  There are two easy ways to convert a 80/16 1:5 ratio to a 1:4 ratio.  I could either use 80/20 pulleys or 64/16 pulleys.  The 80 tooth pulleys I used originally had 5mm bore and I did a poor job of boring them out to 6.35 mm and they wobbled a bit as they rotated so I decided to replace them with 64 tooth pulleys.  I ordered and waited in vain for them to arrive from China, so I replaced the 16 tooth pulleys with 20 tooth pulleys.  I updated the config file on the controller to reflect the new drive ratio by changing steps/mm from 128 to 160.  That got back enough of the motor torque that the mechanism now works reliably on every pattern I've thrown at it.  The noise level isn't much different than it was at 1:5. 

Another Noise

Once the mechanism was working reliably I went back to attacking noise and one I hadn't previously noticed became apparent.  The machine would make a clunking sound when motion along the X axis reversed direction near the center of the table. 

A little investigation found that the belt tension was deforming the frame members and causing the Y axis rails to bow outward.  The result was excessive slop in the fit of the X axis between the Y rails near the center of the table.  The clunking sound was caused by the entire X axis shifting between the two Y axis rails when the X axis reversed direction.  I cut a strip of plywood the width of the table, drilled and installed a couple screws and t-nuts then mounted it under the table between the two Y axis rails.  Offsetting the ends slightly pulled the Y axis rails back into parallel and allowed me to adjust the X axis bearings so there's no more clunking when the X axis reverses.

More Changes to Come

A future redesign of the X axis may include a spring on one of the bearings to keep the X axis stable, and larger pulleys using flanged ball bearings to get quieter, lower friction operation.