[James Newton]

I'm wondering if anyone else is interested in developing new ways of encoding position to adapt low cost motors into CNC machines while avoiding the standard method of using encoders on the motors. There are advantages to putting the encoder ON the axis as a linear strip with readers, instead of putting the encoder on the motor. One the motor, any slop, gearing, and non-linearity in the lead screws results in positioning error. With the position encoder rolled out flat on the axis, all of those issues disappear. You can use a cheap non-linear lead screw. You can use a sloppy gear drive. Because the system knows where the axis /actually/ is, it can drive the motor to compensate for any of those errors.

The trick is finding a good way to print and sense an encoder as a linear strip with low cost and fine resolution.

Fine resolution is possible via vernier sensing even with a low resolution strip. And a /type/ of vernier sensing (single track grey code) can provide absolute position sensing as well.

I've collected some data, ideas, and explanations on this web site:
techref.massmind.org/techref/new/letter/news0312.htm
if anyone is interested in reading up on it.

WARNING: Single Track Grey Codes are a horrific time sync for anyone who enjoys mathematical puzzles. You will become sure that the answer is just around the corner... for years of your life. ,o) Ask me how I know. But now, friends of mine and I have finally worked out how to generate the track and sensor positions for very long run tracks.

We have a mast for an 8 bit, 240 position code, and there are much longer ones potentially available. At this point, we could run the 240 code sequence down the length of the axis, and it would give you the rough absolute position in one 8 bit reading, then a low resolution rotary encoder on the lead screw shaft would give you the position within that 1/240th. Non-linearities in that short distance shouldn't be significant. The encoder on the lead screw would still compensate for gear slop between the motor and the lead screw, but not for slop in the nut riding the screw. Still a good thing.

It might be better to run 2 or more of the sequences down an axis since the axis can't jump from one place to the next instantly, it would be a good combination of absolute and incremental encoders. Absolute within each sequence, incremental between sequences. That depends on how closely we can space the sensors.

Anyone willing to give it a try?

We also need help turning the sequence of 1's and 0's into a nice printable graphic in both a straight line and in a wheel. Any SVG wizards out there?

[UA_Iron]

AMS has a magnetic encoder interface that will work for both linear and rotary applications:
https://www.ams.com/eng/Products/Pos...Sensors/AS5311

It has 12-bit resolution per 2mm pole separation on a magnetic tape - though I think it manages 1.5mm tapes just fine with an increase in positioning accuracy.

This read-head attached to a circuit board is around $20 on mouser or digikey. You can get magnetic tape on ebay (bogen electric) for a relatively good price. - ~$100 for a capable linear encoder.

The limiting point of the magnetic position encoders is in the accuracy of the tape itself and not so much the magnetic read-head. The bogen tape is probably ±20um over a given length with a possible drift over the entire length, but would still probably maintain a fairly decent accuracy - definitely would yield excellent repeatability.

I am sitting here with a Balluff magnetic read-head and 1000mm of tape at 5um accuracy. I was quoted at ~$600 an axis in quantities.

[UA_Iron]

Copley controls makes some awesome stepper controllers - they actually turn a stepper into a servo motor (with more poles essentially). They require an encoder on the motor for commutation. The stepnet plus can take advantage of a linear encoder and rotary encoder to close a position loop via the linear encoder and a velocity loop via the rotary encoder. The controller will accept absolute encoders as well.

This is just a brain dump of some technologies I thought were awesome - I realize I might be deviating from your original intention of this thread a bit.

[bhurts]

"Quote"

(There are advantages to putting the encoder ON the axis as a linear strip with readers, instead of putting the encoder on the motor. One the motor, any slop, gearing, and non-linearity in the lead screws results in positioning error. With the position encoder rolled out flat on the axis, all of those issues disappear. You can use a cheap non-linear lead screw. You can use a sloppy gear drive. Because the system knows where the axis /actually/ is, it can drive the motor to compensate for any of those errors.)

I agree that using a linear encoder will allow you to know true position. That does not make the rest of the statements accurate. If you remove the encoder from the motor and there is any slop in the system you will have issues that will be hard to compensate for. When you command a motion from the motor it will move until the slop is taken up but by then it will be wondering why there was a gap in the motion. It will cause tuning nightmares that will change over time as the mechanical backlash changes.
You will have to atleast use a tach so the computer will know the motor is moving and at what velocity while taking up this slack. You will also have major issues while cutting because you used cheap sloppy components. What happens when the tool can pull the table around because of the slack in the motion. You will then have to deal with instant jerks in motion as the tool pulls itself into the direction of cut almost instantly. The control would then have to try to instantly reverse direction to get back in position. Then it will reverse again to move in the correct direction and cause a never ending cycle.

Just my opinions😉.

As for a low cost linear encoder. That's a great idea. If you can get it working I will be at the front of the line to buy plans or complete kits assuming they don't have a low speed limit.

Ben

[James Newton]

"Quote"
If you remove the encoder from the motor and there is any slop in the system you will have issues that will be hard to compensate for.

Agreed. But hard is not impossible. A 32bit ARM0 processor is $5 these days. No reason why the controller can't learn what the slop is and compensate. And we can put a low cost, low resolution encoder on the motor, or a tach as you suggest, and still have a system that costs less than the currently available high precision lead screw and high res encoder / high precision gear train systems.

The nice thing about single track grey code is you can get absolute position with a single 8 bit read. So you can get very, very quick position updates. No decoding of quadratures, no counting, just "where am I? Ok, there I am, I want to be here, subtract, that my error, am I changing direction? yeah? compensate for slop, otherwise, speed = F(error)" and so on. You get the idea.