Problem homing with index

Looking through the forum, I see I'm not the only one with homing problems...

I have a converted benchtop lathe (same as PM1127-VF-LB) running PathPilot, currently v2.5.1, with stepper motors, 5i25, 7i76 and 7i85s.

I have homing working properly, but when I try to use the index pulse from my glass scales, that's when I run into problems.

My part of Z-axis from the .ini:

Here it will home to the switch and then back off it and stop, and all is good. Home is at the tailstock end.

When I try to use the index from the glass scale (index mark is closer to the spindle than the limit switch), there seem to be no rhyme nor reason to where it will eventually stop. Where it will stop can vary a fair bit between each time I restart PP, even without changing anything in the ini-file.

My part from the .ini for index-based homing:

I had it homing to almost where I wanted it with HOME_OFFSET = -0.650, but I wanted it about 1mm closer to the tailstock, so I changed HOME_OFFSET to -0.700. Now it moved about 5-6mm too far and hit my hard stop. Put HOME_OFFSET back to -0.650 and it stopped 10-15mm short of where it should stop. I restart PP between each time I change the .ini-file.

What is the index pulse wired to? Presumably to the encoder index?
Are you sure that there is only one index?

Is joint.N.index-enable connected to the encoder.M.index-enable pin in HAL?

Please attach your HAL file.

andypugh wrote:

What is the index pulse wired to? Presumably to the encoder index?
Are you sure that there is only one index?

Is joint.N.index-enable connected to the encoder.M.index-enable pin in HAL?

Please attach your HAL file.
 

Hi, Andy.

Yes, the index is connected to the encoder index input in HAL The glass scales have index marks every 50 or 100mm, according to the datasheet. On my X-axis scale (5um) it looks like it is every 100mm. On my Z-axis (1um), I'm not sure, it doesn't move very far before it hits the index mark.

My HAL-file is an altered version of the PP file, so it's a bit of a mess, with a lot of commented out parts. Makes it easier to find what I have altered when upgrading to a later version instead of making it a highly optimized file...

Another thing, I have changed num_encoders=2 to num_encoders=5 in the ini-file to include the encoders from the 7i85s, so I have encoder00 to encoder04, but both encoder00 and encoder01 show the same (spindle rotations). Should I change to num_encoder=6? Why was it num_encoder=2 earlier, when 7i76 only have one encoder input?

TiA,
Stig

To test the index pulses:
(assuming that PP isn't _too_ different from LinuxCNC.)
Open a terminal
halcmd loadusr halmeter
Then set the halmeter to look at the zindex-enable signal.

Then
halcmd sets zindex-enable 1

Though change the value in the halmeter

Then jog past an index, you should see the halmeter switch to false.

On a 7I76 with firmware for 7I75/7I85S, multiplexed encoders are used ,
this causes the (non multiplexed) spindle encoder on the 7I76 to have an aliased encoder.
Its not harmful.

Is it possible you have a noise issue? This could cause apparent random index locations
You can test this by setting index enable as Andy suggested and seeing if it ever
gets reset spontaneously (that is, when not actually at the index location)

andypugh wrote:

To test the index pulses:
(assuming that PP isn't _too_ different from LinuxCNC.)
Open a terminal
halcmd loadusr halmeter
Then set the halmeter to look at the zindex-enable signal.

Then
halcmd sets zindex-enable 1

Though change the value in the halmeter

Then jog past an index, you should see the halmeter switch to false.

PathPilot seems to be slightly different from vanilla LinuxCNC, as I couldn't use your halcmd, but when I had started PP, I could do ADMIN HALMETER and ADMIN HALSHOW, where I could run HAL commands. zindex-enable switched to false when passing the index mark. Index marks are every 100mm on Z-axis scale, same as on X-axis scale.

Is it possible you have a noise issue? This could cause apparent random index locations
You can test this by setting index enable as Andy suggested and seeing if it ever
gets reset spontaneously (that is, when not actually at the index location)

Tried jogging back and forth several times between two index marks. zindex-enable stayed true the whole time. The glass scale cable is metal-wrapped and the shield is grounded, but my electrical cabinet is made out of wood instead of steel. Don't know if that could cause any problems.
 

OK backing up a bit, Are you using the step motor drives in a closed loop
with encoder feedback?

Not yet. That is the plan later, but now it's just getting homing working properly. I'm using Chinese inductive switches and since they are not incredibly accurate, I wanted to use the index marks in the scales to get more accurate homing.

LinuxCNC expects the joint position feedback to be zero'ed
at index detection. The standard stepgen does not do this
(it has no connection with index at all) So homing to index on a
step/dir system does not currently work unless you use encoder
feedback. There is new firmware and support in LinuxCNC 2.9
for homing to index on pure step/dir systems if you want to try
it. This does require a direct input for the stepgen index though
it can be shared with an encoder index say on a 7I85/7I85S

So if I want to use encoder feedback, I need to set up PID and tune it before homing to index will work? Trying the new firmware and LinuxCNC 2.9 would probably not play nice with PP, I guess? At least, not without a lot for work? I think PP is based on an early version of 2.8.