ATC senor options

There are some with quite long forks:
uk.rs-online.com/web/p/slotted-optical-switches/1730303/

But you could machine rings pointing in the axial direction, if you really wanted to.

I think that a 4th "strobe" signal to indicate that the disc was in a valid position would have been a good idea. With the current arrangement it is hard to tell the difference between 001 and (for example) 111 that is just coming in to position. (Also, in this application there isn't really much point using Gray-code, but neither does it do any harm).

Without a strobe I would say that you do need to synthesise a home signal, and the all-on position does seem like a good way to do that.

For the LinuxCNC application I would not bother with any external circuitry, do it in HAL or inside the component that controls the changer.

I don't think that the LinuxCNC "carousel" component can work with this encoder style, as it looks like there are times as the holes move in to place where (for example) the bit2 hole is in front of the bit0 sensor?

If this isn't the case then it probably can be made to work. Are the forks deep enough that the holes can be drilled on three different radii so that each sensor only ever sees its own hole? if that can be done then the strobe signal can be synthesised in HAL from "NOT all off".

Thinking about this, it would probably be good to add a new parameter to the carousel called something like "ignore-zero" for systems like yours, where if no sensors are active the position feedback is ignore (as when strobe is not set) . I don't think it is too late to add this for the 2.8 release.

Yes but a config that only needs to configure carousel would allow the changer to be handled by a standalone HAL file that won't confligt with existing loadrt and addf statements

Ross77 wrote:

The index sensor was going to be a single slot for accurate homing, are you talking about a 4th bit of grey logic to confirm position?

Yes. If you had that then there would be no need to home, the system would know where the disc was any time that it was in a valid position. That would require a slot at each position, though.

Thinking about this, it would probably be good to add a new parameter to the carousel called something like "ignore-zero" for systems like yours, where if no sensors are active the position feedback is ignore (as when strobe is not set) . I don't think it is too late to add this for the 2.8 release.

So would that work with our current config?[/quote]

Not if you can have the disc part-way aligned with a sensor active. You either need no sensors active if out of position, or a strobe to say when the disc is in-position.

Much smaller slot detectors exist: uk.rs-online.com/web/p/slotted-optical-switches/6666511/ is almost certainly _too_ small but would potentially work nicely with a 3D-printed target disc: a360.co/2ZpE9sM
(That design is actually for a slightly bigger version with a 3mm slot width, and is binary coded.)

Ross77 wrote:

That's a pretty neat solution, I suppose it really needs the sensors on a PCB for alignments, compactness. and to add the transistors to boost to 24v output.

Yes, but then a batch of 10 PCBs costs $5 from China:
www.seeedstudio.com/fusion_pcb.html

If I can get out current 3 sensor arrangement to work on 3 separate radi then could we get that to work with Carousel and then rework the sensors later?

Yes, I think so.

I also see that carousel is for dc motors with Geneva gears, is it easy to modify the code for stepper motors?

Take a look at the enhanced 2.8 version of "carousel", it has a lot more pins and can be connected directly to a velocity-mode step-generator:
linuxcnc.org/docs/2.8/html/man/man9/carousel.9.html