So I’m really trying my hardest to take a proper break to rest and recuperate, but I can’t stop thinking about Maslow stuff so I want to write out what I’ve been thinking so that I can 1) not have to remember it all 2) get community feedback 3) put my ideas in the public domain 4) maybe someone else wants to make them so I can just but the kit
There are four main weakness with the current design that I would like to improve. They are:
Ease of calibration - this is a big stumbling block for a lot of folks and needs to be made easier to get widespread adoption
Accuracy in the extremities - Especially in the lower corners we loose accuracy
Cutting speed - To compete head to head with gantry machienes we need to be faster
Productizable - I would love to see companies take the idea and make a product out of it, but for that to happen it would need to be small enough to ship and fit on a store shelf
What I am thinking to meet these goals is a design which uses four kevlar (or some other low stretch material) cables instead if the chains and which has all of the electronics and motors mounted to the sled. Cables need to be used instead of chains because if the motors are on the sled we need to store the extra in a compact area. Also the chain has proved to be a hassle.
This type of design would simplify calibration by making it possible to calibrate with a single button press. By pulling each of the pairs of cables taught one after the other we could measure the distance between each of the four attachment points. This would give us enough information to fully calibrate. It would be nice to have current readings from the motors rather than stalling the motor as we do now.
This type of design would improve the accuracy in the extremities because in the lower corners we could provide the extra tensioning needed to pull the long cable tight and the sled would be lighter without the bricks allowing us to go higher between the motors.
This type of design could cut faster because we are currently limited by the force provided by the weight of the bricks and gravity. If we control all of the forces we can apply more force in the direction we need to cut.
This type of design would also be more productizable because all of the parts come attached to the sled and could be purchased as a built unit. The only external thing which would be needed would be four solid attachment points. This design would also work in fully vertical or horizontal modes.
The first big challenge to building this design would be how to accurately measure the length of the cables. Chain on a spocket moves a fixed distance per rotation while cable going onto a spool winds in more or less cable with each rotation depending on how the cable is stacking. This rules out our current method of using an encoder on the back side of the gearbox which is a significant loss.
To make up for that we would need a very high resolution encoder on the other side (8000+ steps/revolution) and we need a way to relate it’s rotations to the cable movements. I would like to explore using pinch rollers which roll as the cable moves in and out. We would need to verify that they won’t slip; ie if the cable moves in and out a lot of times a mark made on the roller and the cable will still line up.
If that looks ok we need to find a way to read the position of the roller. A larger roller will be less likely to slip and more reliable in general, but a larger roller would require a higher resolution encoder. I’ve played around with the magnetic AMS encoders and they aren’t linear enough for this type of work, the other ones I’ve seen are too expensive.
What I would like to test first is a system to use the sensor out of an optical mouse to read the movement of the roller. Optical mouse sensors are cheap and easily available and can have very high resolution. Unfortunately they have a problem with drifting over time. To solve that I would like to color code the roller with black and white striped to give us a concrete reference for position. Reading the black and white stripes gives us a low resolution position (~10 steps/inch) which won’t drift while the optical sensor gives a very high resolution position (~8000+ steps/inch) which drifts, but which can be corrected by the measurements from the stripes.
Other than that some more thoughts I have are the possibility of using an angled track to have the bearings run on like this:
And standardizing on the DeWalt palm router which is cheap and available in 120v and 240v versions.
With this type of z-axis:
I think that the zaxis motor could potentially be mounted to the same ring that the bearings run along.
There would also need to be stops on the cables such that when the cables are pulled all the way in for storage they can be “zeroed”.
What does everyone think? Glaring flaws? Things we should test early to see if the idea is viable? Potential materials for the cables? Thanks for reading all the way to the bottom!