However, the Trak system at one of my clients never requires homing, and both a friend’s AHHA system as well as his modern Akira-Seki VMC don't require homing after e-stop or limit events.
Naturally, Improved confidence in my homing repeatability will help for safety as I'd rather not second guess the ramifications of hitting the e-stop.
I think the key word here is 'require', what's required and what the best operational procedure is are sometimes two different things, and to some extent its also a subjective thing.
And, yes. I do occasionally jog in the wrong direction and hit the limits as my tool change position is at machine 0,0,0
You should try using soft-limits. Mach will decelerate as you approach the soft limit when jogging. Works great.
Also, for servo systems, an uncontrolled e-stop is downright unsafe. I made a custom 1800IPM gantry system for a client this year using a National Instruments motion card, servomotors and Gecko G340 drivers. Controlled motion on e-stop and soft limit events were a requirement. Suddenly stopping the pulse stream resulted in a drive error/shut-down allowing the carriage to freewheel and do (more) damage. (I wrote my own g-code CAM software in LabView for this project - - a far cry from Mach in features but not bad, I think).
You right, safety is the key issue. Many machines also employ axis brakes as a safety feature. What if the control or drive is failing and when you hit e-stop you may not be able to safely decelerate. The old Bridgeport VMC I'm working on now, has axis brakes and the Galil motion control card will allow you to decelerate on E-stop or to go dead stop. I'm considering trying to employ a rapid deceleration with the Galil , flowed by a short time delay before the braking relay (which also removed power to drives) kicks in. If I have to choose, I'll go with the brakes.
Why can’t the index pulse of an encoder be fed in as a Mach home switch?
Right, the roigersmachine error detection board is a great idea for a stepper system with encoder or glass scale feedback.
Why can’t the index pulse of an encoder be fed in as a Mach home switch?
Does anyone know if Mach's encoder DROs are consistent across e-stops and limit events?
To be accurate the the controller needs to see the edge of the index pulse, (edge triggerd). There is no guarantee that all parallel port inputs will latch at the same time so trying to edge trigger on a fast signal, w.r.t another input would be at best, tough. Also the parallel port can start out in an unknown state, if you are making use of the 'charge-pump' you will have the ability to disable the drives when Mach is not actively driving the parallel port. On my little DynaMill I can E-Stop from a stop and restart Mach and never notice the difference. If it was moving I would not take the chance. On the Bridgeport conversion I have the luxury of the Galil with closed loop control. I still would home after an E-stop though.
Does anyone know if Mach's encoder DROs are consistent across e-stops and limit events?
In my experience, I think so. Still, the question is IMHO, should you trust it? Take a look at the G100 and Galil motion control boards if you have needs that go beyond what a few parallel ports will do.