I think I was convoluting it in my head and not doing a very good job of explaining what I was thinking. Yes, even the through-beam raw sensor input will be analog and will require some sort of circuitry to determine if a threshold has been crossed, I definitely just saw two charts from Omron and said “See?” while actually not applying any actual thinking to that part.
Both styles of sensor will have a certain physical range in which they will generate an analog response which the internal circuitry then converts to a digital on/off. For simplicity’s sake, let’s just say that this response changes in a linear fashion from 0-100 as the detection object moves into range.
If the induction sensor has a range of 1mm where it moves from 0% detection to 100% detection, and it can accurately detect a 1% change in the raw signal to then trigger the actual threshold, it will be able to accurately detect down to 0.01mm.
Now let’s say you have a through-beam optical sensor with a beam width of 0.5mm. As an object moves into that beam, the detected light will drop from 100% to 0% (again, simplifying) as the object moves across the 0.5mm of the beam width. If the electronics have the same accuracy as the ones in the induction sensor, this would result in a detection accuracy of 0.005mm.
The caveat being, none of the sensors I looked at gave such a simple number as beam width. The physical lens on the Elite’s homing sensors are 1mm in width, but that does not mean that the beam is a full 1mm wide. Typically the lens of the receiver is a fair bit wider than the beam to account for any inaccuracies in aiming of the emitter, though that is obviously less of a concern in these one piece sensors, so it is possible they are using a full 1mm wide beam.
Following that train of logic, then an induction sensor with a smaller detection zone should also theoretically be more accurate/repeatable and a quick search for a high precision induction sensor seems to bear that out, as looking at the options for this specific model from Omron does show a correlation between sensing distance and repeatability.
Actually getting it through my head that the through-beam is still initially an analog raw signal does however also dispel my initial thought of the sensor being mostly unaffected by dust, as if the trigger threshold is say 50% light received and dust is already blocking 20% of the light, then the flag only needs to move 30% of the way into the beam rather than 50% to trigger it. It would actually be less affected as some of the dust would be in the same area that the flag is also blocking, but just using a very simplified analogy to show it would in fact be affected. The detection area does again come into play in mitigating this too, as a smaller detection area means any percentage of error is a smaller actual position difference.
I guess to sum up these ramblings, I do still feel that the detection mode of a through-beam optical sensor at least has the potential for inherent accuracy advantages over an induction sensor, assuming all else is equal. I suspect that at Masso’s target precision level, the through-beam sensor was the less expensive option.