I haven’t been nearly as active the last week but this is really some great progress.
I’ll say, I feel a bit less insane now. Thanks @bar and the many others who have been so active in this thread!
3 Likes
Glad to see that progress is being made. Looking forward to seeing what happens for you with the chain corrections. FYI, I built a workbench that had 2 components that were 72" wide x 36" high. Widths were pretty perfect, but the heights were off by about 3/8"’ which I expected. What I did not expect were top cuts that were not straight… The boards were 909 mm high on the left, 902 on the the right, and 899 in the middle. Boards that were cut at the bottom of the machine that were 9" high in file were maybe 1/8" less, but the lines were straight top and bottom.
I will post the work bench project in the “Projects” section later tonight.
Thanks!
2 Likes
That is what will happen if it is calculating the width of the machine
incorrectly.
It would be a good idea to have someone go through the math and make sure we
didn’t let a bug creep in with the recent enhancements.
are you top feed or bottom feed? can you switch and calibrate/test again?
I’m using the bottom feed with the slack chain at the top. I guess I could try switching to the top feed, but I would have to re-engineer the slack chin handling. Sadly, I will be away from the machine for about a week. If I am told that switching to top feed will correct the problems, I’m in… Just can’t do it for awhile.
Thanks David!
1 Like
I will have to try this again tonight, but last night I was trying this, and was not getting such repeatable numbers. They were more than a couple of mm different between runs, using same chain. What does 1degree on sprocket translate to distance of chain?
3 Likes
@bar and @LakeWorthB. I bet we all have a few scraps of plywood around. What about a calibration jig for this?
Something as simple as a a long peice of plywood that can be clamped to the top of the motor extending out horizontally, and a second one that is screwed facing vertically down providing you a vertical reference line to match the sprocket to.
You might need a small groove cut in the first peice so you can sit flat against the gearbox because the mount sticks up a little. But this would be good, you would have a reference feature you could use each time you clamped it on.
This may not be perfect but it certainly affords a tad more accuracy than just an eye ball.
2 Likes
For setting 1 tooth to the 12 o’clock position; I really liked the ideas in this thread.
- Just drape the chain over the sprocket and line up 2 links the same distance from the sprocket
- Drape the chain and put a line level between the chains on either side and level it. (Line levels can be bought at any hardware store for about a $1-3 normally in a 2 pack or on Amazon )
This does assume that your frame is level to the earth, but I think that is a best practice to begin with.
1 Like
One thing to note that I saw was if you don’t have the chain split in half the extra weight of one side verses the other will skew how the center level will sit. The one side will be heavier and not balance itself out. I put mine dead center of the chain then adjusted it. I would think a few links off wouldn’t make that much difference. Two feet on one side and the rest on the other will have a large unwanted affect.
I’ve made some progress on automating the computation of the chain tolerance correction values. Basically all it does right now is take a manual measurement of the distance between the motors and a measurement with the left and right chains.
From those it should be easy to compute the values. It’s still a work in progress, but I’ve got a pull request open to add what I have so far:
I’m getting really consistent values for my left and right chains which might explain why my accuracy test was better than what we’ve seen from other tests:
I would love to see what values other folks get if anyone has the time. It may be that as the chains wear they get more similar which could be why mine are so the same.
I’m also unclear on why I’m getting 3012 now when I got 3009 before. I changed a lot of things including how hard the chain pulls when measuring which could be a factor.
Lots to explore, any and all thoughts welcome!
3 Likes
We should experiment with this, going from a light pull to having both motors at
100%
Pulling the chains tight like this from the motors just makes me nervous. It’s like smashing the break then hitting the gas in a car. It will do it but definitely wasn’t designed for that. Have other ways of measuring the chain been explored recently?
it’s measuing the distance between the motors as much as measuring the chain.
but as for your concern, we already run the motors at full power during normal
operation (in fact, we have seen problems with some cuts because the motors
could not keep up), so this isn’t much more, if any more, stress on the machine
than normal operation would be.
It’s not something to worry about.
With v1.11 and v1.12, when I ease the chain close to tight, and then got to “pull chain tight”, the machine is pretty smooth, and does not seem to be straining too hard. Although there were exceptions, most of the time repeated measurements with the same settings yielded results that were +/- .2mm.
1 Like
I have a thought about this. It looks like the average error for the long measurements is almost the same as that of the short measurements. This means there is no source of variation (error) that gets larger with increasing distance.
I briefly divided the sources of error into two categories, ones that grow proportional to the measured distance, and those that are independent of the measured distance.
- Proportional to distance
a) incorrect distance between motors
b) incorrect chain pitch
c) incorrect radius of rotation - Constant for all distance
a) measurement inaccuracy
b) router bit wobble
c) frame flex
Calibration primarily addresses errors that vary proportional to measurement distance, aka the error increases with increasing distance between cuts. @bar might be reaching the limit of what can be achieved through calibration enhancements. Next steps might be things that address other sources of variation.
Another observation. There are two groups of long measurements, 900 mm and 1905 mm. All the 900 mm measurements measured long. Why is this the case? There might be some information in this. It could also just be statistical in nature.
4 Likes
chain sag and chain tolerance are two factors that have more affect at longer
distances, but not linearly proportional to distance in the case of chain sag)
incorrect rotation radious results is a constant error in the effective length
of the chain, it’s not proporational to distance
frame flex depends on the force involved, so it will be highest at the top
center, and lowest in the bottom corners
David Lang
Calibration primarily addresses errors that vary proportional to measurement
distance, aka the error increases with increasing distance between cuts.
@bar might be reaching the limit of what can be achieved through calibration
enhancements. Next steps might be things that address other sources of
variation.Another observation. There are two groups of long measurements, 900 mm and
1905 mm. All the 900 mm measurements measured long. Why is this the case?
There might be some information in this. It could also just be statistical in
nature.
if the motors are at a different distance than the machine thinks (or the
rotation radius is off), then the triangles that are formed aren’t what the
system thinks, and that translates to vertical errors.
you should play around with the simulator in ground control, it lets you
introduce error to each of the measurements and see what the resulting cuts
would look like.
I’m wondering if we should do:
tape measure with no tension
chain measurement full power
tape measure while still under tension
chain measurement with the other chain at full power
tape measure while still under tension
the first one to detect flex in the machine compared to under tension, and the
last one just to give two measurements of such a critical value (average them if
they are close, ask if the user is really sure if they are not close)
1 Like
That sounds like a good idea, at least until we have gathered enough measurement to determine if all five measurements are needed.
How can we add precision to the mounting of the router to the sled, so that the spindle is right on the point of the triangle formed by fhe chains? With the ring it is pretty easy to discover an error… with the sled stopped, turn on the router, manually drop the bit to -.10”, and rotate the sled on the ring… if the bit does not move and the cut stays at 1/4”, the mounting is accurate. If the hole gets bigger, or forms some kind of a “C” shape (like mine) then it is off. Fixing the error is not easy, with the motor mounts in wood and small corrections needed. Believe me I have tried.
I’ve been thinking about this… the only thing I’ve come up with is…
- Mount the router base so that the bit is as close to the center as possible
- Let the router drill a 1/4" hole (this hole and base mount become the absolute zero/center for the sled)
- Remove router base and router from the sled
- Put a long 1/4" bolt through the hole in the sled with a washer on either side of the sled and secure with a nut. It would be absolutely critical that this bolt is square (90 degrees) to the sled.
- Mount the ring as centered as possible using the bolt as a guide.
Now for the fun part! to make this dead on!
- Make a router circle cutting jig out of a piece of scrap (I’m thinking 1/4" plywood). Put a 1/4" hole in one end and slide that over the bolt in the center. Support it with another nut on the bolt so the jig is at the same height as the ring
- Mount the router on the other end of the jig so that the bit just barely touches the inside of ring at the rings farthest point from the center bolt.
- Use the router to skim the inside of the ring; making it perfectly concentric with the bolt. (Your 1/4" up spiral carbide bit will work fine for this; just go slow and make sure you are not taking off to much … I’m thinking no more that 1/64th of an inch).
- Remount the router in the center and every thing should be square and concentric.
Only problem is you sacrifice your ring in this process, but if you only take off a 1/64th or 32nd of inch … you could probably repeat this a couple of times before you would need to buy a new ring.
This is just my theory at the moment and I haven’t tried it, but I may when I get around to mounting my ring kit.
EDIT: In further thinking, I don’t think this will work. I think the mounting tabs on the ring or the brackets will get in the way of the router making the inside cut.