But this will still be wrong if it cable/wire gets wrapped… but it may not cause an issue in practice.
I see what you are saying and I think I see the confusion. It is important to keep in mind that the PWM signal is not telling the motor to “move” a particular distance it is telling the motor to apply an amount of torque. We can tell the motor to “pull harder” and “pull less hard”. With the encoder in place we check to see if the encoder is giving us the reading we want for our target position and then tell the motor to pull harder or less hard until the encoder moves to where we want it to be.
Understood… but it’s calculating that initial torque value based upon a certain mmPerRot… and my initial concern was that if that’s way off, it will result in a large error at start, maybe even to a point it throws the sled not keeping up error…
But, I think now, that if we just adjust the encoder steps correctly to match the encoder’s rotation, all will work… assuming the wrapping of the cable on spool doesn’t cause an issue as mmPerRot will be off. I hope I’m making sense.
1 Like
I might need to be clearer on this point. If, for instance, the mmPerRot of the “spool” is 63.5 mm and the mmPerRot of the pinch roller is 6.35 mm (to make math simple) and it’s a 2K ppr quad encoder (8K steps/rotation) then we need to set the encoderSteps to 8K * (63.5 mm / 6.35 mm) = 80K. That’s probably way too high, so a lower ppr encoder is needed to get into a range the arduino can handle. That’s what seems nice about the capacitive encoders from CMI; their ppr setting can be programmed (or adjusted via dip switches depending upon the model).
1 Like
@bar… interesting concept for using belts and taking up slack…
3 Likes
That seems very very similar to what I have been working on:
This is controlled wireless and seems to work pretty well to adjust the cable length (although I only got it working yesterday).
Before anyone freaks out, yes that is my SPI cable on the left, yes every wire is a different length, yes they run next to the motor. I made a nicely shielded cable first and that works great so I wanted to see how far I could push things before SPI communication broke down, so far I haven’t found the edge which is good.
Edit:
It’s a little bit tough to see what is going on in the video. There is a roller at the front with an AS5048 encoder on the end of the shaft which measures the cable length:
4 Likes
Progress!
1 Like
Nice. Any views of the pinch roller assembly?
I’m working on a sled mounted idea that almost eliminates cable sag, allows cheap direct measurement of angle, and provides optical absolute sanity check against encoder length measurement, and includes integrated slack management with off the shelf commodity parts. It requires a robust pinch roller. It is not a fully formed solution, but maybe this photo will spark some ideas.
1 Like
Here is a ready made wire feed unit, many like it are available. Replacement rollers and copper plated steel wire are commodity parts. I think some come with encoders built in because they are intended for a very predictable feed rate. Or the magnet for the other encoder can be glued to the roller shaft end. They are designed to carry large amounts of current on the wire, enabling an all in one package.
The feeder can mount anywhere on the sled and feed the wire via bowden tube to centrally located rollers, or fairlead.
secondary spindle with extra small bits for cleanup of inside corners?
mig welder feeds do not actually grip the wire that strongly, they are easy to
jam up and not feed, and not that hard to pull wire through. (besides the
problem of the wire not being that flexible)
David Lang
1 Like
so you don’t worry about accurately calculating buildup on the spool and winding
perfectly, you count on the encoder to track things properly.
how strong can you get timing belts? I know that there are some out there with
kevlar and metal reinforcment to help reduce stretch.
1 Like
Timing belts can be very strong. Thin to replace bike chain, They are used behind V10 engines to drive propellers and fans to lift hovercraft. Standard belts have cotton cord reinforcement. I doubt you would need anything stronger. Just go wider if you need more strength. I have some here I can perform tests on if you like.
1 Like
That’s how my design worked (external encoder on a ‘pinch roller’), though I did try to neatly wrap the wire/webbing. My thought was that since the PID controller would perform better with neat wrapping, I strove to achieve that.
My concern with the design I linked to above (not @bar’s) is that most of the weight is supported by the teeth of the belt. For his drawbot, that’s probably not a big deal, but for a four motor self-contained maslow sled, that’s a big issue. I’m less worried about the belt snapping than I’m worried that the teeth break off.
Notice the tension adjustment. When set loose intentionally they can allow you to pull wire by hand, or slip to avoid a rats nest if there is a kink upstream. But the steel V groove pinch rollers can also be adjusted very snugly. Likely more than we can achieve with 3d printed pinch rollers.
There are serrated 4 roller wire feeders that could lift a small person off the ground. They are about $60. Some of the common ones have 4 amp, 24volt motors, more power than maslow. Why would they use larger motors if they couldn’t get grip enough to use it? The jamming comes when the user bends the output welding cable too tight with the wire being fed down the middle. It creates a kink and the feeder pushing wire behind it keeps on pushing. Not a factor for us. We do not require the wire to be any more flexible than it is.
I’ve used a lot of small timing belts driving much larger loads than Maslow and belt teeth failing has never been an issue. The timing belt in your car drives the entire valvetrain against valvespring tension reliably for 60,000 plus miles with barely a sign of wear. The only common failure mode I have ever seen in a timing belt is edge wear and/or longitudinal split, usually due to misalignment. When that occurs, the belt still holds.
Good to know… I just see pictures of 10 mm belts with tiny teeth and think that’s a recipe for disaster. But I have no experience with them personally.
There is a belt size and style for nearly every application. If small teeth don’t work for your application, choose a larger pitch. Several teeth always work in unison creating a very large shear plane. There should be a minimum of six teeth in mesh and at least 60 degrees of belt wrap. The reason I did not suggest them for Maslow is the large potential for misalignment. That does cause belts to fail. We have large spans, uncontrolled tension, a non recirculating, non static belt path. That is not to say belts cannot work with Maslow, just that it is not an ideal application for timing belts. Belts could still turn out to be the ideal solution for Maslow though. Does that make sense?
So currently I am contemplating 4 alternatives to chain and sprockets;Rope, Timing Belts and pulleys, Mig Wire feeder, and tape measure pinch roller.
Rope is being ably tested by Bar so there is no need for me to do the same.
Timing belts are a bolt on swap for chain so anyone can do that.
Our robotics team has already built and used the tape measure pinch roller solution in competition so we know it can be robust and reliable. But I haven’t worked out how to mount them to preserve Maslow geometry for kinematics yet.
Welding Wire has an easy solution to geometry but reliability concerns if kinks can happen. That requires testing, so lets see how long shipping will take.
Are there any others I have missed? Alternatives currently being tested, or that you would like to have tested?
1 Like
As long as you can take up the slack (like with the tape measure spring bdring used above), I don’t think non-recirculating is an issue. I really like what he came up with (except for the use of steppers… but that’s easily changed) and think it has potential as long as the belt is up to the task.
Well, not in the new design which puts everything on the sled. You have to have a take up mechanism for the slack.
Is there a second ‘pinch’ roller directly beneath that we cannot see? Or is that just a single plastic roller?
Is the roller spinning in any bearing? Or is it plastic shaft in plastic hole?
If it is the latter answer to both questions, I am impressed by the alignment and free spinning encoder.
Is the magnet glued to the end of plastic shaft? https://www.quora.com/How-heat-resistant-is-super-glue I don’t think my garage ever reaches boiling temperature, safe.
1 Like