hostmot2 encoder documentation

Should be fixed now in master and 2.9.

What's the magical incantation to get this version? I installed from 2.9.2 iso. I updated the repo to search 
deb www.linuxcnc.org/ bookworm base 2.9-uspace

deb [arch=amd64,arm64 signed-by=/etc/apt/trusted.gpg.d/linuxcnc.gpg.key.binary.gpg] www.linuxcnc.org/ bookworm 2.9-uspace base
deb-src [arch=amd64,arm64 signed-by=/etc/apt/trusted.gpg.d/linuxcnc.gpg.key.binary.gpg] www.linuxcnc.org/ bookworm 2.9-uspace base

I also tried deb buildbot2.highlab.com/debian/ bookworm 2.9-uspace

This got me updated from 6.1.0-16-rt-amd64 to 6.1.0-17-rt-amd64 and linuxcnc 2.9.2.106.g26e5d but it's still displaying the quad errors. I thought the linuxcnc or buildbot 2.9-uspace would give me all the 2.9 updates including this fix? Not sure what I'm missing?

I don't know how old the buildbot image is but you may have to build
from 2.9 or 2.10 source to get the latest.

I cloned master and built the RIP with your fix which works for the hal_encoder_test.hal script!

When I run the attached config, I still see the problem? (Simplified version of my mill.)  I modified the hostmot version # and quad error string to make sure I was running the new hostmot and I am. So I think it's a step in the correct direction but not a complete fix.

I couldn't figure out how to reopen issue #2680. So I opened a new issue in the tracker. I changed hm2_encoder_read_control_register() to clear encoder errors whenever they are disabled. It still needed the hm2_encoder_force_write(), but with this fix my config works fine. I'm not sure how much overhead the force_write() adds but for my config it doesn't matter since I'm off when this occurs. 

I will push this change, but I wanted to make sure you saw it.

static void hm2_encoder_read_control_register(hostmot2_t *hm2) {
    int i;
    for (i = 0; i < hm2->encoder.num_instances; i ++) {
        hm2_encoder_instance_t *e = &hm2->encoder.instance;
        if (*e->hal.pin.quadrature_error_enable) {
            e->reset_quadrature_error=0;
            if(!e->prev_quadrature_error_enable){ // detect rising edge of pin quadrature_error_enable
                e->reset_quadrature_error = true;
                hm2_encoder_force_write(hm2);
            }
            bool state = (hm2->encoder.read_control_reg & (HM2_ENCODER_CONTROL_MASK | HM2_ENCODER_QUADRATURE_ERROR)) && e->prev_quadrature_error_enable;
            if ((*e->hal.pin.quadrature_error == false) && state) {
                HM2_ERR("Encoder %d: quadrature count error\n", i);
            }
            *e->hal.pin.quadrature_error = (hal_bit_t) state;
        } else{ // quadrature error disabled, set reported error to 0
            *e->hal.pin.quadrature_error = false;
            e->reset_quadrature_error = true;
            hm2_encoder_force_write(hm2);
        }
        e->prev_quadrature_error_enable = *e->hal.pin.quadrature_error_enable;
        *e->hal.pin.input_a = hm2->encoder.read_control_reg & HM2_ENCODER_INPUT_A;
        *e->hal.pin.input_b = hm2->encoder.read_control_reg & HM2_ENCODER_INPUT_B;
        *e->hal.pin.input_idx = hm2->encoder.read_control_reg & HM2_ENCODER_INPUT_INDEX;
    }
}
 

Forcing a control register update every cycle on every encoder 
almost any machine using Mesa encoders is really not something
that should  be pushed.

Let me do some more investigation into why a existing error is not
cleared before the error mask is removed.

 

I agree. It will work as a temp solution for me as I always run with quad errors enabled and machine is off when it's doing the extra updates. I also had a bit & mistake that I cleaned up. I'll change the pull request to a draft. 

Well...I cannot duplicate the error in either current 2.9 or master.
The only way I could generate the "error on error enable"
was to miss-configure the encoder DPLL timer so the read data
gets pipelined by 2 instead of the normal 1 relative to the error
clear.

I tested this cycle  at least 20 times on both versions:

disable check

generate quadrature error

enable check

and was never able to generate an error on enabling the check.
 

I will retest on my lathe config and try a fresh pncconf config on the mill.

Thanks for you patience!

I went back and tested my lathe, mill, and a new config named quad. I build the code with your changes and the extra hm2_encoder_force_write() from the pull request. I test the lathe config on one computer, the mill and quad on another identical computer. These are old (2008ish) computers with dual core 3Ghz. The only changes I made to quad.hal were to comment out the extra x-pos-fb lines and add 

net machine-is-on <= halui.machine.is-on           => hm2_7i92.0.encoder.00.quad-error-enable

All three configurations fail with the master (and work with the extra force write) version of hm2_encoder_read_control_register(). I have a bad scale that's easy to create the quad error with. Every time I create an error while quad-error-enable is false, it will erroneously trigger when machine-is-on. (No reason to try multiple times to catch the problem on my setup.)

I have attached the quad files I am using.