I put some tape on either motor coupler and ran Y homing from different start positions, very slowly and absolutely confirmed this working in action. I've now squared the CNC, by adjusting one of the sensors until my dial gauge read the same from one end of the linear rail to the other, when sliding it along a square. It's probably not perfect, but it should do for now, at least as proof to myself that I can do it this way.

Notably, I also had to up my idle motor currents a little so it wouldn't go back out of square when they timed out. That's the I value on M906, where you set your motor amps.

@jrow nevermind. Had to flip the signal and ground wires on the other one for some reason.

Actually come to think of it, what would be the cost of manufacturing boards with these additional components instead? Is this something that could be achieved at a low rate of production with a pick and place, or via a PCB service?

I tried a new tact; I increased the resolution of the mesh, so it didn't have to 'turn' as steeply, and now the defect is mostly gone.

@Jrow my proposal would be to include the webcam PCB on the back of the Z axis of a CNC machine, and to mount the laser level permanently to the bed of the CNC.

The warm up routine would include a travel along X and also along Y, measuring the deviation in Z, then using that to build a grid of Z offset for mesh bed levelling. You could also measure Z backlash and compensate for it.

@oliof the discord conversation is in the PrintNC community, which is open to join, but it's pretty well condensed in the youtube link.

Basically; Discord user Byran ON is using a laser level common in the construction industry, and the PCB from a common webcam to detect differences in the level of the top surface of a steel beam. In his case, the steel beam is a mounting point for linear rails for a CNC, and the intention of his device is to manipulate the surface of the top of the steel beam so it's as flat as possible. The webcam PCB includes an imaging sensor with an array of pixels*, which the above linked Github software uses to detect the most illuminated point of the laser via sampling.

Another user has proposed a similar method, using a stretched wire, and an optical device that detects the center of the wire. He previously did this by eye to manually grind a lathe for someone.

• Image of the array of pixels to give an idea of how this produces such an accurate reading:

It seems irrelevant to me whether or not the sensor can be implemented on the Duet, if it can record data on an external device, then transfer that data to the Duet.

BTW, an FYI for the others in this thread. I started a discussion about another sensor option being discussed by the PrintNC community; https://forum.duet3d.com/topic/32134/a-potential-highly-accurate-sensor-for-levelling

@Jrow ok, so this isn't precisely what I'm suggesting. In the case of the CNC for instance, it's not about scanning a bed or work surface, but the variation along linear axis (X & Y primarily) in order to compensate for variation in travel.

The laser level would be mounted and collimated separately, and the machine levelled separately. The imaging sensor would likely be mounted on the Z axis, or the Y, and could be used for a wide variety of tests from measuring Z backlash, to measuring the amount of variation in Z along the X & Y rails.

A mesh could then be calculated from there.

@dc42 should've used the search function!

I just came across a lengthy discussion on the PrintNC discord channel where a user is combining a webcam sensor without a lens and a laser level to accurately measure the variation in the surface of a steel beam that he's using in the construction of his CNC.

You can see a video explaining what he's doing here on his channel; https://www.youtube.com/watch?v=hnHjrz_inQU

And find the discussion of this method here; https://discordapp.com/channels/648972213734604807/1078447519517921392

And the webcam software can be found here; https://github.com/betzuka/laserlevel

This is all very intriguing to me, as I've built a CNC and I'd love to improve it's precision. And this should help, but I immediately wondered if these sensors could be combined with mesh bed levelling, to provide a highly accurate CNC, without the added steps of modifying the CNC gantry. Couldn't a CNC (or 3D printer) simply substitute a sensor like Duet's IR sensor for this webcam / laser level combo, run along the X and Y axis, while measuring the deviation, and then compensate?

Further; and this gets into even more intriguing territory for CNC's; couldn't the controller also measure deflection of the gantry in real time, along with a tool height sensor, to determine (and compensate for) spring forces?

I put some tape on either motor coupler and ran Y homing from different start positions, very slowly and absolutely confirmed this working in action. I've now squared the CNC, by adjusting one of the sensors until my dial gauge read the same from one end of the linear rail to the other, when sliding it along a square. It's probably not perfect, but it should do for now, at least as proof to myself that I can do it this way.

Notably, I also had to up my idle motor currents a little so it wouldn't go back out of square when they timed out. That's the I value on M906, where you set your motor amps.

Yup, I have confirmed homing via this method now. But in trying to install the endstop plugin, I think I messed up my DWC. I had a CNC specific version of it installed, and attempted to update to the current DWC (3.4.1), and I don't have the menus the plugin suggests.

Is it something worth creating a ticket about, regarding M119? Is that a bug; that it doesn't report multiple endstops?

@jay_s_uk I've already homed the Y axis with io5.in only, by switching from io2.in, to ensure that endstop works.

I just am a bit concerned that M119 doesn't report multiple enstops for Y.

@fcwilt I'm a bit nervous here. When I try M119, with M574 Y2 S1 P"io2.in + io5.in" set, I only get an endstop triggered if io2.in is triggered. Due to the nature of the machine I really don't want to try bending it. Is there a way to check that my config is correct beyond what I've already specified?

The documentation for M574 only mentions multiple endstops for Z. I'm worried it's possible that it wasn't implemented for Y.