Nan - voltage

Tesremos wrote: here is your generated error


edit: and now i know how to screenshot!!! winning

It's a learning curve for us all...

OK, that shows the arc-voltage-input is the culprit, so it seems the encoder is sending a non float output.

rodw wrote: So clearly as I suspected, and mentioned before, the calibration values in the GUI are not making it to plasmac.comp.

The calibration values are there, they are 0.000000 which is to be expected on startup as the GUI probably hasn't yet completed loading.

Another one to try...

phillc54 wrote:

Tesremos wrote: here is your generated error


edit: and now i know how to screenshot!!! winning

It's a learning curve for us all...

OK, that shows the arc-voltage-input is the culprit, so it seems the encoder is sending a non float output.

Thats a really big call. What you are saying is that Linuxcnc's code is broken at the core. I don't believe that!

If in the menu, you used show hal configuration to drill down to the Mesa encoder being used for arc voltage will tell you the data type as reported from the Linuxcnc core. If the data type is float, it will give you a float. That float might be infinity but its still a float.

Doing what I proposed would isolate the GUI as the culprit or not. Its a simple change and well worth implementing.

rodw wrote:

phillc54 wrote:

Tesremos wrote: here is your generated error


edit: and now i know how to screenshot!!! winning

It's a learning curve for us all...

OK, that shows the arc-voltage-input is the culprit, so it seems the encoder is sending a non float output.

Thats a really big call. What you are saying is that Linuxcnc's code is broken at the core. I don't believe that!

If in the menu, you used show hal configuration to drill down to the Mesa encoder being used for arc voltage will tell you the data type as reported from the Linuxcnc core. If the data type is float, it will give you a float. That float might be infinity but its still a float.

Doing what I proposed would isolate the GUI as the culprit or not. Its a simple change and well worth implementing.

This is from Tesremos' connections.hal file Note on the screenshot the value shown from arc-voltage-in...

rodw wrote: Doing what I proposed would isolate the GUI as the culprit or not. Its a simple change and well worth implementing.

How can the GUI be the culprit, the only value used from the GUI are arc-voltage-scale and arc-voltage-offset. It clearly shows them both as zero. That would give an arc voltage output of zero.

PHILL!!!

you've done it!

30/30 no nan.


edit: 100/100 no instances of nan

What that one does is look at the arc-voltage-in signal and if it sees an inf instead of a valid float then it starts a one second loop. If it sees a valid float within one second it continues on its merry way otherwise it will give an error message.

Can you get PurduePete to give it a try as well.

It may be worth an ask in the Driver Boards section as to why it could start up with an inf on the encoder output.

So infinity * 0 = 0 so how can it report a nan?
What does halshow report for encoder velocity?

You have not eliminated anything yet.
Don't have a closed mind. There is something going on here that none of us understand.

I still think its more likely that a reported bug in NML which is used by the GUI to talk to the Linuxcnc core is the culprit than the encoder output.

phillc54 wrote: What that one does is look at the arc-voltage-in signal and if it sees an inf instead of a valid float then it starts a one second loop. If it sees a valid float within one second it continues on its merry way otherwise it will give an error message.

Can you get PurduePete to give it a try as well.

It may be worth an ask in the Driver Boards section as to why it could start up with an inf on the encoder output.

When you go back to the First post and look at the screen dumps arc-voltage-in has a valid value but arc-voltage-out has a nan. The original screen dumps still don't correlate with the solution. Maybe the one second delay simply allows the system to sort itself out. eg for the system to load.

I think it should be reported as a bug on the basis that behaviour has changed in master from 2.8. Its a very cludgy work around.