Auto-Calibration Mods?

Hey,

I was thinking about the ‘string level’ concept for calibrating the 12 O’clock sprocket tooth and it occurred to me that an accelerometer might be more accurate than the small 4-6" water bubble levels commonly available.

I know my phone has one thats pretty accurate and cost a $, tops. So I went looking for Arduino boards and it seems that they’re plentiful, @ $10 or so.

I wonder how difficult it would be to create an accurate (> +/- 0.1 degree) level using the Maslow and an Aux pair and something like this:

Leveling the sprockets is so key to calibration, and its difficult to get a good measurement across such small distances on a part that’s at least 6’ off the ground this should be a priority.

An effective means of ‘Auto-Calibration’ will lead to greater initial setup accuracy, portability, and general performance.

thoughts?

4 Likes

How could we arrange the Calibration Process so that the board is mounted on the Sled?

Also, wouldn’t a module like this on the sled give us superior Acceleration Planning / Feedback?

2 Likes

I was thinking about the ‘string level’ concept for calibrating the 12 O’clock
sprocket tooth and it occurred to me that an accelerometer might be more
accurate than the small 4-6" water bubble levels commonly available.

I wonder how difficult it would be to create an accurate (> +/- 0.1 degree)
level using the Maslow and an Aux pair and something like this:

The problem is how accurately can you mount the sensor to the mechanical part
that you are using as the base of the level. The chips are cheap, it’s the
accurate manufacturing/assembly that is hard.

Leveling the sprockets is so key to calibration, and its difficult to get a
good measurement across such small distances on a part that’s at least 6’ off
the ground this should be a priority.

The key is to get a way to magnify the distance so it’s easy to see.

getting something that can have a long, stright edge and hooks over the sprocket
would let us find the level easily.

sin(1 degree) = 0.017
sin(0.1 degree) = 0.0017

If we have something that is a long reach vertically, we are looking for an
error of under 2mm over a meter to be within 0.1 degree

now, note that 1 degree of error in the positioning of the sprocket is 0.17mm of
chain, 0.1 degree of error is 0.017mm of chain. We are aiming for an accuracy of
~0.4mm overall, and the error in chain length can be multiplied by at least two
in the position of the sled.

So I think we are looking to get the sprocket point to within 1 degree of
vertical, not 0.1 degree

1 degree of error is ~1/4" at 12", so if we can accurately hook something to the
sprocket with this sort of span, we should be in pretty good shape.

How could we arrange the Calibration Process so that the board is mounted on the Sled?

which board

Also, wouldn’t a module like this on the sled give us superior Acceleration Planning / Feedback?

acceleration planning is purely a math thing (look at the different line
segments we need to cut and figure out what speed we need to be to make things
work well)

I’m not sure we need to try and sense acceleration, we have good measurements of
our position and so can calculate speed and acceleration from that.

which board

one of these multi axis accelerometer/inclinometer/etc boards.

If we mount something like that on the sled parallel/perpendicular to a known object (like the vertical leg of the vertical pantograph, 90deg ring mount, or along the 45 deg of the wood pantograph) wouldn’t it tell us at when it’s level to the bottom of the frame/floor?

spitballin’ a calibration example:

3D print a sprocket cap on which to mount the sensor, cap slips over the sprocket with a tooth at vertical.

3D print a mount for the sprocket cap that is attached to the sled in the manner described above.

Mount the sensor to the sprocket and auto-level to 12 O’clock sprocket with sensor and software.

repeat for both chains.

Remove sprocket cap and place on sled mount, on sled.

Place chains on 12 o’clock sprocket like the current calibration and feed out both chains equally to ‘center or below’ based on user input.

enough chain is fed to attach the sled, attach chains to the sled.

sled mounted sensors are in a known configuration.

auto level the sled based on sensor with software.

bonus: grok and store frame angle.
bonus: calculate sled pressure on work surface based on sled weight/frame angle, store for later when we determine idealized sled weight/placement.

feed equal amounts of chain to move sled to bottom of workpiece manually so bit touches outside bottom of workpiece.

repeat with top of workpiece.

Assume the frame and workpiece is already leveled during assembly.

isnt that all we need? (in addition to existing ‘vertical height of sprockets above workpiece’ and ‘linkage measurements’)

which board

one of these multi axis accelerometer/inclinometer/etc boards.

If we mount something like that on the sled parallel/perpendicular to a known
object (like the vertical leg of the vertical pantograph, 90deg ring mount, or
along the 45 deg of the wood pantograph) wouldn’t it tell us at when it’s
level to the bottom of the frame/floor?

with the triangulation kits (arms or ring), we don’t care if the sled is level
or not, it doesn’t matter.

spitballin’ a calibration example:

3D print a sprocket cap on which to mount the sensor, cap slips over the sprocket with a tooth at vertical.

3D print a mount for the sprocket cap that is attached to the sled in the manner described above.

Mount the sensor to the sprocket and auto-level to 12 O’clock sprocket with sensor and software.

if you are looking for sub-degree accuracy, you are going to need to calibrate
your 3d printed part and how the sensor attaches to it.

repeat for both chains.

Remove sprocket cap and place on sled mount, on sled.

Place chains on 12 o’clock sprocket like the current calibration and feed out both chains equally to ‘center or below’ based on user input.

enough chain is fed to attach the sled, attach chains to the sled.

sled mounted sensors are in a known configuration.

auto level the sled based on sensor with software.

I don’t see that the angle info of the sled helps you.

bonus: grok and store frame angle.
bonus: calculate sled pressure on work surface based on sled weight/frame angle, store for later when we determine idealized sled weight/placement.

without knowing friction coeficcients and surface area, that doesn’t help.

feed equal amounts of chain to move sled to bottom of workpiece manually so bit touches outside bottom of workpiece.

other than finding the limits (which I have suggested that we do), how does this
help? and how does the sensor help us?

repeat with top of workpiece.

isnt that all we need? (in addition to existing ‘vertical height of sprockets above workpiece’ and ‘linkage measurements’)

no, this has given us no information on the rotation radius or chain sag that
the current calibration figures out.

I have advocated that we should not do the Y offset measurement, but instead
have the user move the sled to the home position (in one link steps so we can do
the ‘marked link’ reset directly to the home position), and then after the
calibration, move the bit ‘around the world’ to define the software limits
(don’t move the sled outside of this area without popping up a warning)

But that didn’t end up being the option selected.