Default Probing feedrate?

[paukstelis]

I seem to have a fair bit of trouble with probing feedrates. For this specialized application, My probe routine has its own feedrate (G91 G21 G38.3 Z-100 F150), but occasionally I get much much faster feedrates, despite it clearly sending the F150. I feel like there must be a default feed rate for probing somewhere, but I don't see it in config.h

[terjeio]

I have never seen this - and there is no default feedrate from settings and/or config. The feed rate is taken from the last F parameter encountered.
What is the frequency of this happening and do you have G-code that I can use to replicate it?

[paukstelis]

There is nothing special in the gcode. It does get created dynamically based on user input, but it is just a series of probes:

G91 G21 G38.3 Z-100 F150
G91 G21 G0 Z5 F500
G91 G21 G0 X5 F500

This is just repeated for whatever the length of the probing requested.

I haven't been able to nail down a root cause yet, but one theory is that if the movement before starting a probe was a rotary axis movement at a given feedrate, that is when the problem arises. I've worked around it by just giving a very tiny linear axis movement before each probe at the desired feedrate and I haven't seen the issue (yet).

[terjeio]

Do you use G93 anywhere in the code and what is the frequency of the incorrect probes (one in a hundred, one in ten, ...)?

[paukstelis]

G93 was not used in the probing code directly. However, when I had a series of these failures in a row, it was directly after giving a rotary axis movement that did use G93. Those failures prompted me to ask the question. Prior to that, it was very low frequency.

It does seem that if the feedrate is off on the first probe, it will be off for all subsequent probing calls. I could always immediately tell if there would be a failure on that first probe.

[terjeio]

A G38 probe should fail with an error if inverse time mode is active according to the NIST spec (3.5.9.1):

"It is an error if:
...

• G38.2 is used in inverse time feed rate mode,"

However Grbl (and grblHAL and I assume other forks as well) does not, see this comment.

Perhaps grblHAL should be changed to fail with an error?

If following the NIST spec you should switch to G94 mode before probing, or at least provide a feed rate consistent with the active mode.

[terjeio]

A discussion that could be the reasoning behind for how G38 behaves in Grbl.