What if it was mounted at the top center, same height as the motors? Cuts the length down a lot. Reduces “sag” somewhat due to shorter length of cable, but is pretty vertical when sled is near middle of sheet. Also electronics closer, out of way of sawdust, helps pull sled up, not down or to a side.
With this we can still find any x, y location with shorter cables, more vertical. Still needs a linear sensor with angular measurement.
A quick calc suggests that this would be better, but would still only resolve 0.72mm when the bit was in a lower corner (longest line).
Does anyone have experience with Kevlar thread? Choosing a strong (enough) light-weight line with a small minimum bend radius, low stretch and good chafe resistance is one of the tasks. Do we know what the Goliath is using? They seem to be fair leading their line out a radius-relieved hole, a very nice way to solve the end-point-location question, but bend and chafe (line and orifice) are considerations.
Take a look at the SN65HVD75, I believe that’s what is in the encoder. No thru-hole version, but the 8-SOIC isn’t hard to hand solder.
I was looking at this.
It says .04% accuracy. At 80" thats 0.032", 0.81mm. repeatability is half of that.
mount that on a rotating shaft with a good angular sensor and we got it.
I must be doing something wrong - looking up that part at Mouser, they want $830 for the mechanism, without an encoder. While I agree that ‘worry about the cost later’ is a good approach, this is outside my reach.
Sorry, I did not look at pricing. My bad.
Even in the top center, you have a very small angle to detect.
David Lang
for everyone else watching, I picked up the max485 chip.
David Lang
Great brain-storming! Here are some thoughts based on your ideas and some research of my own:
Sensor Options
Mounting Options for the sensor(s)
Note: Chains in red. Wire/cable in blue (from sensor to top-center of router).
Next time I will post details on sensor specifications/costs.
An an option for the "DIY’’ version is the AMT23 series absolute capacitive encoders. They are roughly $55 each, and count 16384 shaft turns in addition to 16384 absolute positions within a turn. The interface is RS485, which should be stable in our noisy environment. A Synchronous Serial Interface version is available as well, which would use fewer gpio pins. Since they are capacitive rather than optical they should be immune to the our dusty environment.
One thing I’ve noticed about the draw-wire rigs I’ve seen for sale is that they seem to expect the line to pull straight away from the unit in a single direction. We need somewhere near a 70 degree arc of motion if the units are placed at the ends of the horizontals.
I was thinking that the “distance” sensor needed to be mounted in an assembly that can swivel to keep the line pointed at the bit. I think of a downrigger (fishing) type of action. It contains the “reel”, has a long post with an eyelet on the end to keep it oriented toward the load (fish), and swivels accordingly.
It is just not “electronic” like we need. But because that type of mechanism could measure distance, and with an angular sensor, could as one assembly triangulate the position of the bit from any location it is mounted.
The main problem is the angular sensor has to be just as accurate, if not better, than the distance measurement.
Keep thinking.
I called the original company and they quoted me $700 for a the MK88 model after asking many times. First questions I asked is what is the ballpark price and he couldn’t tell me. I dislike it when companies do not give a straight forward price, makes me feel like I’m going to the doctor. anyways told him I needed something 1/10th that price. he said he would look into it and get back to me. I would not be surprised if they never called back.
Has anyone thought of using a camera to detect position? Or would it not be accurate enough? I.e. similar to https://www.jjrobots.com/air-hockey-robot-a-3d-printer-hack/
Very interesting. Maslow is very slow compared to an air hockey puck. I wonder where the camera would need to be mounted. Out in front to see the entire area.
Reads like this developer has done a good job with his segmentation of code modules.
If this could give the accurate location of the bit, it might be what @dlang was talking about to move the closed loop servo point out to the sled.
Nice find. Now we’ll need to see what the Maslow experts think about this possibility.
Has anyone thought of using a camera to detect position?
yes, frequently. the current discussion is happeing in the optical calibration
thread
Or would it not be accurate enough? I.e. similar to Air Hockey Robot (a 3D printer hack) – jjrobots
well, we are needing 6000+ positions horizontally, with a normal 4x3 camera,
that is about a 20 megapixel camera in an ideal world.
however, you really need 2x-4x this resolution to control the system, so that
would be a 80-320MP camera
and then there is the problem of getting the camera mounted exactly perfectly to
see the entire work area and nothing else (and not be angled at all), which
would increase the resolution requirement.
in any case, it would be far more expensive than the current mechanism.
David Lang
This looks good! About $50 each:
CUI AMT23 Modular Absolute Encoder
https://www.mouser.com/cui-amt23-encoders/
Does anyone have CAD plans to make a retractable reel to attach to the encoder? (Or know where to buy one?)
Check out this incredibly detailed analysis of draw-wire (and other) measurement systems!
For a cabled system, would something like this create a triangulation kit without moving parts? If you fixed one end of the cable to the top of the sled, feed it to a free turning pulley located where the current motors are, and return it to a spool/encoder/motor placed directly across from the fixed point on the bottom of the sled, such that a line drawn between the two passes through the center of the router bit, will it maintain a triangle where the bisection point between the fixed end and the spool end is maintained at the center of the router bit. You’d need some way (e.g., eyelet) to fix the point the cable feeds out of the spool, but there’s no “moving” parts like what exists with the various triangulation kits… drawing below:
And, if you add a pulley to the bottom of the frame and fed the cable around that before returning to the sled, would that not still maintain a triangular relationship between the fixed end of the line, top free spinning pulley, and spool/encoder/motor… drawing below
That’s a very interesting idea! That solves the problem of the changing angle very nicely.
I think the latter, assuming this works, would improve performance in the bottom corners.
