I just discovered that Maslow exists today, so please forgive if this idea has been covered before.
It seems that both the links on the sled and the ring are designed to allow the sled to rotate, but hopefully stay concentric with the router bit. Great if it works. Does it?
For potentially better accuracy, I thought “why not just stop the sled from rotating in the first place.” If the sled always stays plumb, then you can add additional tools to the sled (like a sharpie) at points other than the concentric center. Those tools just have a static offset from the center.
Take the parallelogram idea from the links, but extend it all the way up to the motors. Get rid of the hard links and replace them with equal lengths of chain. This requires about double the amount of chain and 2 (or 4) additional sprockets, but the motors and software can remain unchanged. It solves the problem of shifting tangent points coming off the motor gears because as it shifts at the motors, it shifts that negative amount at the sled. Note, the dotted lines in the attached picture. Each motor has a double row sprocket, and there is another double row sprocket directly beneath it (plumb, but in plane with the plywood). The original chain path is duplicated by a parallel link a fixed distance beneath it. The sled attachment sprockets are fixed to bolts the same vertical distance apart as the the upper joints. At calibration, those sprockets are free to turn. But once calibration is locked in, they are staked so they can no longer rotate. Hopefully the attached drawing will make things clear. Sorry for my poor handwriting.
Welcome! @slomobile
Perhaps as a new user the upload of a picture failed.
I can’t follow you right now, but will try.
The original sled had fixed brackets and (from the little I know) was far more complicated in the math calculating the cutting bit to stay at position no matter how far the sled tilts.
Forum didn’t like the .tiff file. Changed it to a .jpg.
During calibration, with the sled sprockets free to turn, it should come to rest with the top and bottom chains on the left and right equal effective length provided you count out the same number of links.
For parts, I like https://www.vexrobotics.com/vexpro and http://www.andymark.com because they support the FRC teams I mentor.
This is a very cool idea! I haven’t seen it suggested before but it makes a lot of sense. My instinct is that it would take a good bit of refinement to get a system like this dialed in, but I can’t think of a single reason that it wouldn’t work just like you describe
Welcome, i got a feeling you’re gonna feel right at home here.
That does look like an interesting idea, i can’t really see anything wrong with it.
But the sled rotating isn’t really a problem any more, with the linkages. and the shifting tangents have been calculated out, i believe.
That is an intriguing idea. It would require 4 double sprockets, which are a little more expensive. I would consider separating the chain mount points right and left on the sled, as the centered sprockets would interfere with the dust collection and in some cases the z-axis mechanisms.
Now that FRC competition season is over, Team 1248 is just starting to assemble our kit; perhaps we could modify ours as a proof-of-concept.
this is a suggestion that hasn’t been made before Good job of thinking out
of the box.
the ring/pantograph is to keep the math simple by keeping the chains pointing
directly at the center of the bit. This would change that slightly, but with a
fixed offset (without the second chain, the sled will swing, we’ve had people
try that)
potential issues I see:
it would be interesting to make brackets that hold the second sprocket in just
the right place (and keep the chain driving them tight)
the distance between the sprockets would need to match the distance between the
mounts on the sled
the chains would need to clear the router (the angles get up to only 10 degrees
from vertical). I think this is the biggest problem.
you would have some interesting work routing the slack chain and keeping the
chains from hitting each other.
Teams 5002, 5045, 4764 look forward to your proof of concept!
Separating the sled sprocket mount points should work with these caveats.
Plumb is an easy condition to verify for 2 inline points, more difficult when offset.
The amount of separation practically should be some even multiple of the chain pitch.
Perhaps it could be set up initially with any interference causing parts temporarily removed from the sled,
with the sled rotated so the sprockets line up, perform the usual plumb test,
lock the sled sprockets,
reposition the chains on the sled sprockets by the appropriate number of links.
All 3 brackets that hold the sprockets could be identical(provided they all have a big router hole) and non adjustable(fixed sprocket locations). A great candidate for a mass produced metal or plastic(lexan?) part. Chain slop(backlash) can be minimised with UHMW slider block.
There are several potential solutions for making clearance.
Offset pulley locations,
4 sled pulleys at the sled corners,
3 pulleys where 2 of the pulleys are at locations that would make equilateral triangle(side length a multiple of chain pitch) with the ideal virtual pulley location. adjust number of links in that chain.
The back side of the UHMW slider blocks could be shaped to guide the upper slack chain outboard from the lower. If the slack ends of the upper and lower chains are joined, the original Maslow solution can be applied.
I originally said “It solves the problem of shifting tangent points coming off the motor gears”. But I was wrong, had it backwards and doubled the error. In order to work as I described, the chain path at the motors and sled must both be over or both be under, not over/under. Only both under seems practical, but changes the route of the slack chain toward the center which alters the slick slider block solution.
I’m trying to understand what is being proposed and I’m struggling with the terminology and the diagram. Maybe I figured it out. “double row sprockets” are basically two sprockets stuck together (i.e., two rows of teeth), correct? In the diagram, is the top sprocket connected via a relatively short loop of chain to the bottom sprocket one “row” so that when the motor turns the one up top, the one below also turns the same amount? Or is there some other chain routing in the diagram I’m not seeing?
You understand it madgrizzle. Everything you said is correct. There are parts of the picture (far left and far right slack chain) where 2 chains happen to follow the same path. I did not draw that detail because I thought it would be distracting from the overall concept.
All 3 brackets that hold the sprockets are identical and non adjustable(fixed
sprocket locations), great candidate for a mass produced metal part. Chain
slop(backlash) can be minimised with UHMW slider block.
3 brackets? are you talking about sprockets on the sled?
There are several potential solutions for making clearance.
Offset pulley locations,
4 sled pulleys at the sled corners,
both of these will drastically complicate the math. With the chains meeting
directly above/below the bit, you have a ‘virtual chain’ that is mid-way between
them that is exactly pointing at the bit at all times, so the triangular
kinematics will work.
If you have the chains attached elsewhere on the sled, you no longer have this,
and the exact position of the bit relative to the chain is much harder to
calculate)
3 pulleys where 2 of the pulleys are at locations that would make equilateral
triangle(side length a multiple of chain pitch) with the ideal virtual pulley
location. adjust number of links in that chain.
explain what you are talking about (diagrams please)
The back side of the UHMW slider blocks could be shaped to guide the upper
slack chain outboard from the lower. If the slack ends of the upper and lower
chains are joined, the original Maslow solution can be applied.
the chain does not move side to side well, so if I am understanding your
proposal, I see problems.
joining the slack of the two chains won’t let you use the original maslow
solution because you have twice the distance of slack (the original and slack
across the top both double the chain on itself to get ~7 ft of chain if 4’ of
space)
All 3 brackets that hold the sprockets are identical and non adjustable(fixed
sprocket locations), great candidate for a mass produced metal part. Chain
slop(backlash) can be minimised with UHMW slider block.
3 brackets? are you talking about sprockets on the sled?
There are 3 places where we need 2 sprockets a fixed distance apart. At left motor, right motor, and sled. It should be possible to use the same base plate part in all 3 locations to fix the locations of the sprockets. 1 part design instead of 2 or 3. Not a requirement to be done that way, just an option.
Offset pulley locations,
4 sled pulleys at the sled corners,
both of these will drastically complicate the math. With the chains meeting
directly above/below the bit, you have a ‘virtual chain’ that is mid-way between
them that is exactly pointing at the bit at all times, so the triangular
kinematics will work.
The runtime math remains exactly the same or adds a constant offset if the user chooses. There is a fixed alteration in chain length that is only calculated once at calibration. The ‘virtual chain’ still passes through the router bit. Any offsets can easily be a constant added to the code when choosing configuration type. The location of the sprockets is not random, but chosen so the sled sprocket is colinear with the ideal chainline. See attached picture. I’ve added 4 sprockets to the original configuration so you can see how everything fits together. A 3 sprocket configuration would use 2 of the new sprockets and 1 of the old. Offset uses any of the cross corner sprockets. Notice how the virtual chainline and the bit position never change. If you prefer the router to be anywhere else on the sled, just add in its static offset. Likewise for adding any additional tools like a marker.
3 pulleys where 2 of the pulleys are at locations that would make equilateral(oops, meant isosceles)
triangle(side length a multiple of chain pitch) with the ideal virtual pulley
location. adjust number of links in that chain.
explain what you are talking about (diagrams please)
The back side of the UHMW slider blocks could be shaped to guide the upper
slack chain outboard from the lower. If the slack ends of the upper and lower
chains are joined, the original Maslow solution can be applied.
the chain does not move side to side well, so if I am understanding your
proposal, I see problems.
joining the slack of the two chains won’t let you use the original maslow
solution because you have twice the distance of slack (the original and slack
across the top both double the chain on itself to get ~7 ft of chain if 4’ of
space)
See picture, I see no problems. Since the slack chain lengths move equally, as long as they don’t interfere at the start, they wont interfere later as long as they have similar paths. Replaced slack sprockets w/plywood in picture, but either should work. Left side is over, right side is under.
why do you want sprockets at the center? wouldn’t you want just a pivot point?
No, if you lower the sled, the chains will wrap around the 4 sprockets on the corner, and the ‘virtual chain’ mid way between them will no longer point at the bit.
I see what you are talking about now, but I think it’s drastically over complicating things, you can do far better just nesting the chains will probably work better.
why do you want sprockets at the center? wouldn’t you want just a pivot point?
The sprockets are there for 2 reasons.
To cancel out the shifting point at which the chain departs the motor pulley. As the sled lowers, engaging less of the motor sprocket, it engages more of the sled sprocket, and vice versa, thus eliminating that source of error without any math.
To enable sled X position adjustment, quantify Y position error, and allow rotational alignment despite misplaced center of gravity. The sprockets have either a shaft clamp or set screw and are on a fixed axle(bolt).
During setup, the set screws are loose, sled sprockets can move.
Motors can pay out the amount of chain that should theoretically result in sled being centered on the workpiece.
Now you can manually move the sled to the actual X center of the workpiece and tighten the top setscrew so it can no longer turn. You have now aligned the software’s logical X center with the workpiece physical X center eliminating that initial error.
Now, note if the sled hangs plumb by dangling a plumb bob from the top shaft and see if it aligns with the bottom shaft. If not, add weight to the high side till plumb. Then tighten the lower set screw to lock in that alignment. If for some reason, you are not able or dont want to balance the cg, you can just manually set it plumb and lock it in.
Make sure X is still on the mark, if not, readjust.
Now you can physically measure the error in the Y direction and feed that info to the controller firmware.
This procedure probably takes longer to explain than to do. It may be unnecessary as software can guess at location as has been described in current documentation. But this provides an empirical means to eliminate or quantify static initial error. Something I have yet to see described elsewhere. But I’m new and might not have found that section yet.
One source of systematic error that remains is chain catenary. Chain sag. I believe there is a solution to that as well. Current documentation on calibration calls for cutting a sheet of plywood. While actual cutting is necessary to confirm that everything is staying aligned in the presence of load and vibration, it shouldn’t be a first step. We want a good indication that Maslow can hit any target on the workpiece before we make sawdust.
Since the sled no longer rotates and we can add multiple tools, I propose adding this line following sensor to the sled for systematic workpiece calibration. Carefully stretch black electrical tape in straight lines around the perimeter of a workpiece or backer board, and now Maslow has some ground truth features it can follow. A calibration routine can have the sled follow the line of black tape around the workpiece, noting intersections with white tape at measured intervals, and comparing the actual position with the calculated position based on motor encoder counts. This information can be used to build up a table of correction factors across the workpiece. Interpolation can fill in the blanks between plotted points. This eliminates the effects of catenary and various other possible error sources. It does not eliminate dynamic error such as chain whip, router bounce/tug, frame creep. However, as a nondestructive automated loggable test, we can compare old test results to new to identify specific areas of change which we might be able to trace back to a shift in a specific frame joint or other cause.
explain what you are talking about (diagrams please)
No, if you lower the sled, the chains will wrap around the 4 sprockets on the corner, and the ‘virtual chain’ mid way between them will no longer point at the bit.
I don’t understand what you are saying here. As far as I can tell, regardless of sprocket number used, the same motor movement would result in equivalent sled movement. Please point out any situation where that is not so.
Edit: I see it now. It works in the picture with that fixed angle of chain, but while operating at different chain angles, the 4 colinear sprockets are no longer colinear. Good catch.
Another thing I was slow to see, this does look like a Maslow built entirely of chain with no frame pictured. I lost access to NX and have been too set in my ways to learn another CAD package, so you are stuck with incomplete bad drawings for now.
I see what you are talking about now, but I think it’s drastically over complicating things, you can do far better just nesting the chains will probably work better.
[/quote]
Maybe I drew it poorly, it seems simple to me. I was not suggesting both the over and under configuration for a single machine. Just showing 2 different possible configurations on the same sheet of paper. What do you mean by nesting the chains?
Good job of thinking out
