Velocity problems

eslavko wrote:

as you already calculated I need to decrease rotary limit under 30 deg/s. I can accept this for jog. But you (Andy) are wrong in term that this disturbance is only in jog mode. It is in MDI too. And can be problematic for G00 (and fast G01).

I _think_ that things are different there in the JA3 branch.
I have been playing about with a JA3 config to try to prove this to myself, but I think I might have been editing the wrong files, and / or working in an old JA3 or something.

eslavko wrote:

I search a lot but JA3 branch seems to be mistery. I find near nothing, and afraid if I install it I cant configure it as ini is different but no documentation found

Yes, I am finding the same problem, though the sample configs in that branch do have INI files in the new format. It may be that the documents exxist but that we haven't found them.

And stability of JA3 is probably in question for "unatendet" run of machine.

Stability of JA3 is reported to be fine and at least two LinuxCNC integrators have sold JA3 machines to customers.

So, if you can figure out how to configure it, it will be as stable as the normal release (as it gets all the same bug-fixes) but the way it handles non-trivial kins is better.

You can experiment with JA3 without breaking your existing configuration by compiling it as a Run-In-Place system.

The basic steps are:
Install Git to download and manage the LinuxCNC source code
Install a number of other packages to be able to compile LinuxCNC
Configure the Git branch to joints_axes3 and compile.
In a terminal, typeto switch to the run-in-place configuration instead of your normal installed configuration. (and it really does have to be dot space dot slash at the beginning of the command )
In the same terminal, type linuxCNC to start the special version (the menu will still launch the standard installed version)

If you think that you want to try this (and for your machine I think you should) then start at section 2.1.2 here: wiki.linuxcnc.org/cgi-bin/wiki.pl?Instal..._or_8_04_from_source
Then follow sections 2.2, 2.3 and 2.4.
Then, before following the commands in section 2.5, switch to the joints-axes3 branch:[code] git branch --track joints-axes3 origin/joints-axes3
git checkout joints-axes3[/quote]
Then follow the section 2.5 instructions to build the version.

Git is a very clever (and complicated) piece of software that keeps track of all the different files in all the different versions of a software package and switches between them, So you can step backwards through the development phases, or look at different branches of the "tree". It is a tree because different people work on different things at the same time from the same base-version and their versions diverge until they are all combined together in a release version.

This is likely to take some time. Probably a whole evening to get a working JA3 running a demo config. However at that point you will not only be able to run JA3, but any other branch, and you will be able to make any changes you want to LinuxCNC and run those too.

andypugh wrote:

I _think_ that things are different there in the JA3 branch.

Having experimented, they aren't

It looks like even in the JA3 version you would need to set the C and B axis speeds low enough for XY and Z to be able to keep up.

One advantage in JA3 (which might make this acceptable for you) is that the AXIS and JOINT velocity limits are different things.

This means you can have a C axis limit of 30deg/sec but a JOINT-5 limit of 1000 degrees per second. So, you can switch to joint mode to rapidly position a rotary, then switch back to world mode for a coordinated move. (This does work a _lot_ more smoothly in JA3)

How do you jog?

eslavko wrote:

andypugh wrote:

How do you jog?

For now with keyboard.

If you were analog jogging then it should be fairly easy to pass the jog speed command through a limit component related to the cosine of the B-axis value (I have realised that both Z and XY velocity limts are dependent only on B angle)
The same functions connected to feed-override ought to be able to limit coordinated movements.

This is not a general solution, but might work in your case.

I still think that the B and C axis limits being discussed will be far less of a problem in practice than you think.