ClearPath HLFB triggering fault on machine enable

LinuxCNC 2.9.

In my gantry table setup, I've got ClearPath SDSK servos connected to a Mesa 7i76e. Tonight I connected the enable and High Level Feedback (HLFB) pins up. The HLFB is configured to assert when the servo is within a specified number of counts of the commanded position; the input deasserts (goes low) when the servo determines that it is out of the specified range.

It behaves correctly, except that the signal is deasserted when the servo is disabled. If I enable the servos, LinuxCNC immediately faults because the HLFB doesn't go high fast enough.

I was hoping to use this HLFB signal as a way to tell LinuxCNC that the servos were out of range.

Any suggestions on how to accomplish this? Maybe there's a way to have LinuxCNC wait a bit after enabling (F2) the machine before monitoring the fault inputs?

Relevant config lines in attached custom.hal: 52, 128, 133.

Thanks!

One possibility is to feed a delayed version of the enable signal (using say "timedelay")
to a "and2" gate that masks the fault for a few ms after the machine is enabled

Note the usually all the enables are driven at once so you may not really need 4 sets of all enable logic.

sometimes it is only a debonce needed as it is only 4-5ms delayd

Hmm. I think I'd rather not delay the HLFB signal during normal operation if I can avoid it, so maybe concocting a delay on machine enable is the better (if much more complicated) approach. I'll have to design it so the conditioned HLFB input is high until some time after the machine enable goes high, and there will just be that small window where the real HLFB signal is ignored.

At 1200 mm/sec, it can move 6 mm in 5 ms. I'm not sure what typical e-stop timing should be, but 6 mm before it even detects the situation seems like a lot.

If you do what I suggested, the fault input will not be delayed
_except_ for the first few ms after the machine is enabled

@Pcw,

If you do what I suggested,
Do you think he is doing you suggested?
Pcw, I think soon or later he will be on horriblevideo's.com

But anyway.
If Tom Kosovo has a job for him, i will follow Tom.

PCW wrote:

If you do what I suggested, the fault input will not be delayed
_except_ for the first few ms after the machine is enabled

Yes, that's what I'm planning to do, thank you for the suggestion. I was responding to the debounce suggestion, which would always delay the fault signal.

Off topic slightly how do you find the clearpath's? Did you replace steppers and if so what power did you go for? Was it like for like torque or could you get away with less?
Also an idea of what you're cutting at feeds and speeds would help me a lot in my choice.
Thanks
Jools

I love them, but I don't know yet if I selected the right ones. I wish there was more high-level communication back to the LinuxCNC controller, but they're no worse than a stepper in that regard (just can't take advantage of their built-in encoders). I never bought Avid's control system or spindle, built all that myself around the ClearPath. When I did a quick price comparison, it ended up costing about the same, although I've added a lot of extra parts that might end up making it cost more (especially the spindle, which was overkill and cost $6k; a Chinese version of it is about $2k).

I purchased four CPM-SDSK-3411P-ELS .

The HLFB signal can be configured to trigger upon up to four conditions being met or exceeded. One is "position good," which is what I'm using, and indicates when the servo's actual position is within some configurable number of counts of the commanded position. Right now I've set that to 100 counts (and my servos are configure to 800 counts/revolution), and it triggers sometimes if I do aggressive stop-start maneuvers. Could be the acceleration I've set is too high, or maybe I need a larger margin, or maybe my servos are a bit too small.

As-is, the servos are FAST. I've only cut one thing so far, and it didn't have any trouble, but I haven't pushed it yet.

The whole system really is too much for the 8020 Avid frame, but that's okay. The process has been fun.