‘friction’ also effectively includes cutting forces and simple friction will
vary drastically during a cut from the sawdust, differing amount of
surface that’s been cut away, etc.
in other words, not something we can reasonably calculate.
If we can add chain sag and stretch, that would be a wonderful thing to add.
Any chance of adding it today so that it’s in tomorrow’s 1.0 release?
I haven’t delved into the code, but my first impression is that the calculation of the sprocket position for a particular _TargetX and _TargetYcoordinates depends on the stretched chain length.
But the stretched chain length in turn depends on the unstretched chain length (remember: stretch is given as a % of unstretched chain length).
And the unstretched chain length depends back again on _TargetX and _TargetY… which is where we started.
So it looks like the Arduino will have to perform an iterative calculation for each _TargetX and _TargetY coordinates until it arrives at the correct sprocket positions with stretched chain length.This will slow it down a bit…
I may be misunderstanding the process (like I said, I haven’t delved into the code). If I am misunderstanding, would someone familiar with the code please let me know the details of a simpler way?
I have only just ordered my Maslow, however I am a stunt rigger by trade for feature films and I have extensive experience with vector based systems for flying people. Essentially, we have much larger motors (ranging up to 50hp) with software that creates physical movement similar to the Maslow in a 2 axis application, however we can add as many devices as we like to create 3 dimensional movements for flying people and cameras, etc. After reading (most of) this thread, my first recommendation would be not to complicate the system if it’s not entirely necessary. The next recommendation would be to eliminate as many variables as possible without complicating things to see if the changes offer acceptable accuracy. Because I have not used a Maslow yet I was curious if the catenary deflection of the chains affected accuracy, but thought the sled weight must have been enough to make it a non issue. Reading this thread confirms that it is an issue. This is how I would recommend proceeding:
Eliminate chain and sprocket. Replace with a drum and synthetic line terminated into the drum itself (no friction based drive, too inaccurate). Do not mess with steel cable. Synthetic lines are stronger than steel with less “creep” (term for continued extension in line as static weight is applied and left for extended periods of time). Without researching exactly what is available in small diameters, I would recommend a 12 strand Vectran. Off the top of my head, 1/8” diameter Vectran (when terminated properly) has a minimum breaking strength of around 2,000 lbs, has stretch <1%, but most importantly weighs almost nothing. A smaller diameter may be a better fit, but I know the numbers associated with 1/8”. Eliminating the majority of the catenary line deflection will likely bring accuracy up to tolerable levels, and if a simple software solution can account for stretch, I think we could make Maslow extremely accurate. I will start looking into necessary parts for a conversion, and maybe even have something ready to test when my Maslow shows up in February.
As I said, maybe 1/8” isn’t the best diameter for the application, however the manufacturer states .71% of elongation at 20% of the minimum breaking strength of the line (1900 x .02 = 380 lbs). If the estimations of 33 lbs or even 66 lbs per chain maximum mentioned earlier are correct, we aren’t anywhere near that level of elongation. We are down somewhere around .07% (maybe less with such light loads). If we step up in diameter, things get better for elongation, slight trade for line weight. 1/8” weighs in at .64 lbs per 100’, 3/16” weighs in at 1.3 lbs per 100’. Bottom line- I think dealing with a little more (predictable) elongation is better than trying to compensate for such heavy chain.
Sorry to reply to myself here, lol. I also wanted to mention that different manufacturers rate these lines differently. I looked at 2 different manufacturers for the info I gathered quickly above. I use a company called New England Ropes almost exclusively. They tend to be a bit more conservative in load ratings, etc. I knew that V-12 from New England had a minimum breaking of 1900 lbs. For elongation, I used a different manufacturer (Cortland) because I couldn’t find a chart quickly on my phone from New England. Cortland is rating 1/8” Vectran at 2,800 lbs. Plug that into the 20% Breaking strength (2,800 x .2 = 560lbs), and things get even better. All I can say is that it is defenitely worth a closer look. I know that Aramid fiber rope like vectran will outperform heavy chain in this application.
The strech of a #25 chain (like used in the Maslow) is of the order of 0.044mm/m for every kg of load. That’s 0.004% per kg. From what you wrote above, it would look like Vectran is not even in the same league. Yet even this level of chain stretch is significant enough to affect Maslow’s accuracy.
Terminating the line by attaching to the drum is a good idea. The sprocket and chain does effectively the same thing since every chain link engaged with a sprocket tooth is effectively “terminated” in that sprocket tooth.
If you’re going to modify your Maslow, I’d be interested to see the resulting accuracy. Maybe we could discover something that would help improve all the other machines.Please do post details as you proceed sonwe can provide (hopefully helpful) input if you like.
Thanks again for the ideas! Welcome to the Maslow world!
Replace with a drum and synthetic line terminated into the drum itself (no friction based drive, too inaccurate).
Is there an off the shelf drum that could be used in place of the sprocket? Also, once the rope winds back on itself, that would change the diameter of the drum, which means a rotation of the motor would pull in or let out slightly more line. Would that need to be accounted for?
Absolutely. That will have a significant effect on positional accuracy.
Take a look at this cool video describing a hanging plotter build project (Maslow concept) and discussing sources of innaccuracy (including drum diameter change with cable winding)
Discussion of accuracy starts around 6:00 minute mark if you don’t want to watch the whole thing.
Also, with the sprocket/chain design the stretch of the used length of chain is possible to calculate because we can easily calculate the used length of chain. The rest of the chain doesn’t come into play.
With a cable/drum setup, the full length of the cable is always subject to tension because it’s effectively a rope tied at one end and being pulled. The problem is, stretch would be very difficult to calculate. Yes, you know the length of cable being stretched (namely the full length all the time) but as the cable winds on the drum, the friction forces between increasing amounts of cable and the drum opposing that stretch (and hence affecting the true position if the sled) are very difficult to estimate.
In the end, if you’re not looking for a high level of position precision then all this doesn’t matter, and we could use chain, cable, rope, cotton thread or elastic bands even, depending on what level of accuracy we’re looking for.
I would size the drum so that overwrapping was not necessary. I don’t have my Maslow yet, and struggling to find sprocket diameter (currently out of the country and can’t open files on my phone). Someone suggested garage door cable drums like this:
I like these because they are spiral grooved, ready for termination at the drum, rated high, and accept a keyed shaft. Not sure if any of the dimensions are correct, if the shaft size is correct, and how they would affect things like feed rate, etc.
Thank you! Do you think that inaccuracy is from the slight elongation, or more from the catenary deflection (or sag)? Again, I don’t even have a Maslow yet, but I would guess that most of it is coming from the catenary. If we could use software to compensate for stretch, we could almost eliminate the “sag”.
For 2 m of #25 chain at 30 kg of load, the stretch would be:
0.0444 mm/m.kgf x 30 kg x 2 m = 2.664mm
I haven’t calculated the catenary length to compare. If you could do that it would be great. You’ll need the weight per metre of #25 chain which you can find representative values for in the reference pdf catalog linked earlier in this post.
We can then compare the magnitude of catenary sag to the stretch and see if they are comparable.
currently we are somewhere worse than 1mm accuracy, this is getting close to the original target of 1/64", but quite a ways from 0.1mm
remember that this is a machine made out of wood, which warps and swells with humidity changes, not to mentions the fact that the chain wears over time and has a small, but unknown amount of slop per link, the router is held in place by gravity and can move as the bit hits different wood density
all these sources of errors add up and the 0.4mm is a fair target to be working for. And when you are doing woodworking, 1/64" is very good, if you were to push it to 1/256" (0.1mm) that starts getting to the range where sanding the work has as much effect on the wood.