Arrakis 3.0 Updates- Now It's Arrakis 3.1

After living with Arrakis 3.0 for a few months, a few problems have led to some changes. 

1) The cheapo 6A rated power switch welded itself in the "on" position about the fifth time I switched the table on. The table has a 350W power supply - 6A should be plenty of capacity. 

2) I noticed that quite a few of the patterns started with the ball rocketing from the home position to one of the opposite edges of the drawing area, drawing a straight line across the table before the actual pattern started to be drawn. Sometimes the pattern was dense enough to wipe out that line, but very often it wasn't. I found that really annoying.

3) I noticed that the table started making some small noises only a couple months after I finished building it. That suggested parts were wearing.

4) I run the table in random mode most of the time, where the table randomly selects a pattern to draw from over 200 patterns stored in the controller's memory. Sometimes it draws a nice pattern that I'd like to keep on the table for a day or two. The only way to do that is to switch off power, which means the LEDs are also switched off. I'd prefer to be able to have the LEDs on without the table drawing any new patterns.

Changes

1) After checking the LRS-350-24 power supply specs and finding inrush current rated at 60A maximum (!), I replaced the 6A switch with one rated for 10A, and added a 16 Ohm 5A NTC thermistor and 3 second time delay relay (found in a box at the makerspace, with contacts rated for 7A @ 250V) to switch power to the table. Now when power is switched on, the thermistor is in series with the power line going into the power supply, limiting surge current to a maximum of about 10.6A (peak voltage on the power line is 170, so maximum current will be 170V/16 Ohms=10.6A, but only if I happen to flip the switch at the exact peak of the voltage). After 3 seconds, the relay closes and shorts out the thermistor. During that 3 seconds, the thermistor gets about 10C warmer than ambient temperature. This seems to have solved the power switch/surge current problem.

3 second time delay relay (the gray box) and NTC thermistor (the gray disc). If the relay ever fails, the thermistor will remain in the circuit and the table should continue to operate.

This is how the relay and thermistor are wired. Theoretically, I shouldn't need the relay at all as the thermistor resistance drops when it heats up, but I'd prefer not to have it sitting there hot when the table is powered up. If the relay ever fails, the thermistor will still be in the circuit and it should continue to operate with the thermistor sitting at an elevated temperature.

2) I manually edited all 223 pattern files and eliminated the lines that went across the table at the start of many of the patterns. I also eliminated some odd back-and-forth-along-the-edges motion that occurred at the start of some patterns. I'll be more careful to remove such lines before uploading new patterns to the table.

3) Arrakis 3.0 had wheeled carriages running in v-slot aluminum for the Y axis. I suspected that this is where the noise was coming from. When I opened up the table to inspect the mechanism I found that the Y axis carriage wheels on the outside of the mechanism frame were wearing out and the wheels on the inside of the frame weren't even touching the frame. The wear pattern is the result of the tension on the belts pushing the carriages inward while the attachment to the X axis linear guide is loose (by intention, clearly a mistake), allowing the inward motion and some tilting. It was a very bad design!

For the MakerFest in Elkhorn, Wi. on Feb.14th, I installed wheels that were made of some mystery material that was supposed to be "self-lubricating", and it restored the table to quiet operation. When I got the table home I opened it up to look at the mechanism and think about what I could do to improve it, and found this:

Self lubricating?

Arrakis 2.0 uses UHMW blocks sliding in the aluminum slots and has been 100% reliable for years. I decided to try converting the wheeled carriages in Arrakis 3.0 to use UHMW blocks instead of wheels, and to solidly attach the X axis linear guide to the two carriages. 

I found some UHMW in my stash that I was able to carve into sliding blocks to replace the wheels in the Y axis carriages. I tried it and it just didn't work out- the UHMW blocks fit loosely in the v-slots and the result was that the X axis tilted, making a noise each time the Y axis movement reversed direction. Even if they had fit well, as the mechanism wore in, it would have started making noise. Ugh.

The real problem with Arrakis 3.0 was that the X axis wasn't rigid enough. Each of the wheeled carriages could tilt due to the belt tension and that caused uneven pressure on the wheels resulting in uneven and excessive wear. Another problem is the limited accuracy of the construction of the mechanism and the use of just two eccentric adjusters at each of the wheeled carriages.

The solution to the first problem was to make the entire X axis more rigid, which I did by mounting the X axis guide rail on a piece of 1/4" aluminum tooling plate that spanned the width of the X axis. The 1/4" plate is flat and rigid, so it will flex much less (especially when screwed tightly to the X axis guide rail) than the two original 1/8" carriage plates that were loosely screwed to the ends of the guide rail.

Set up for drilling and milling the tooling plate for the X axis. Hole positions were critical, the milling was for weight reduction and cosmetic reasons (though it's inside the table and most people will never see it).

The X axis linear guide mounted on the tooling plate. 

The second problem was solved by using eccentric spacers for all the wheels. The first end is adjusted so all four wheels fit the V-slot rail, then, at the other end of the X axis, the adjusters for all four wheels are positioned to comfortably hold both sides of the V-slot rail. The holes in the eccentric spacers are offset from center by about 0.79 mm, so allow for about 1.58mm adjustment, which is more than sufficient. I would really prefer to spring load some of the wheels so they are always held in contact with the rails, regardless of imperfect spacing or flexing that may occur when the table is operating. 

I'm not sure about using 4 wheels at both ends of the X axis. If I just put 4 wheels (3?) at one end, the X axis assembly will follow the rail that it is clamped to by those wheels. Adding wheels at the other end over-constrains the motion. If the two Y axis rails aren't absolutely parallel, the second set of wheels may cause the mechanism to bind. This is why it would be good to have one set of wheels spring-loaded so they can allow for the rails to be out of parallel. Arrakis 2.0 has a spring loaded UHMW block at one end of the X axis for this reason and it works perfectly. In the end, I used 4 wheels at each end of the X axis. I will reexamine it when I see how well the wheels hold up in the new configuration.

Arrakis 3.1 left side of x axis. Yes, I know the screws holding the wheels are short- they will be replaced with longer screws when new wheels arrive and get installed.

Right end of X axis.

The tooling plate raised the X axis guide rail, belt clamps on the magnet carriage, and pulleys by 3.175 mm. That meant I also needed to raise all the corner pulleys and motors/drive pulleys, and endstop sensors the same amount. I had to modify and reprint all the motor mounts, end-stop sensor mounts, and corner pulley stand-offs. Raising the X axis guide rail also raised the magnet, so I reprinted the upper belt clamp part of the magnet carriage to lower the magnet so it stays at the its original vertical position and doesn't scrape the bottom of the sandbox.

Magnet carriage. The green parts is the new, thinner version that was needed to lower the magnet and prevent it from scraping the bottom of the sandbox. The gaffer's tape wrapped around the magnet keeps it from lifting up 

Note: the 3D printed PETG concave pulley flanges and motor mounts appear to be holding up just fine, as are the belts. There are some rubber crumbs around some of the pulleys, but that's to be expected. Eventually the belts will wear out and have to be replaced, but it won't be any time soon. I tried moving the twists in the belts from the short segments between the drive pulleys and the corner pulley blocks to the very long segments between the corner pulley blocks and opposite ends of the table. I was originally concerned about noise from the belt teeth hitting the pulleys at the corner blocks and with belt clearance along the long sides of the mechanism. Neither seems to be a problem, so I left the twists in the new positions. 

I may take another crack at using sliding UHMW blocks instead of wheels, depending on how well the wheels hold up with the new X axis configuration.

4) The solution to the LED problem was to install a switch that allows me to cut 24V power to the controller board. There are now three switches located on the underside of the table, one for main power, one to control power to the controller board, and one to switch power to the floor lights on the underside of the table. 

Here's the mechanism running: