multi spindle discussion

[c-morley]

Current behavior AFAIKT with a bit of testing.

M3/M4 will start all spindles as long as their respective S codes are already set, adding a $ with a number will start a specific spindle.
M5 stops all spindles - adding a $ with a number specifies which spindle to stop
After an s code has used a $ to specify a spindle, plain s codes after that will only control that specified spindle.
You cannot put a $ with a number with out another spindle control code -ie you can't preset a specific spindle to be modal in this way.

My thoughts - Mostly copied from dev email list...

Setting the s code to a particular spindle (with %) will from then on always controls that spindle with a plain s code. Not sure if that is on purpose but i wouldn't think so.

If you don't set the s code with $ then only spindle 0 will start and stop with m3 m4 m5.
meaning a program running on a single spindle machine runs the same on a dual spindle machine.
But that relying on not setting the $ code seems a bit risky considering linuxcnc can be remapped and use subroutines (that could set it) or maybe left over from a manual setting (I didn't try that)

I did find that using the python module to control the spindle (like all the python screens do) the stop command seems to assume spindle 0 so pressing stop button only stops spindle 0. Which may or may not be whats expected but there is no python command to stop all spindles, unless you added one I didn't check.

I still feel that with mcode and python module, specifying all spindles with say -1 is a better practice.
We should fix the python module to be able to specify all spindles for stopping.

I feel that with s code, it's probably a bug that it remembers the last specified spindle and continues to adjust it with a plain s code.
It would be good practice and consistent if specifying $-1 would set the spindle speed for all spindles.

[andypugh]

Making S modal was deliberate, but on balance I think it was a mistake.

It is probably also true that if someone really needed to set up a group of spindles and start / stop them together then they could be expected to read the manual to find out how.

statetags and various Python interfaces do not, at the moment, report on more than the default spindle. The overlap between those using these interfaces and those using multiple spindles is probably small.

[andypugh]

M19 stops all spindles regardless of $ number. Should it?

[c-morley]

m19 - I wouldn't think so.
Yes I thought maybe it was modal so tried $1 on a line by it's self and that was not accepted. I agree I don't think that is the way to go.

Qtvcp now has support for multiple spindles if some one wants to add them.

I just think it's important to iron out the details before releasing so we don;t have to break behavior on the next release.

[andypugh]

See 3139328

[c-morley]

oh great I'll test that out.

[c-morley]

That is a great improvement.
s code will not accept $-1 - I think it should.
I'll see if I can figure out how to get python to do the right thing with -1 command.

One other question.
I was entering commands in MDI and if I entered a S or M3/M5 wrong it stopped all spindles.
Do you think it should?
I think it should do nothing other then complain.

[andypugh]

s code will not accept $-1 - I think it should.

I suppose that wanting all spindles at the same speed is quite likely on multi-spindle (copy) routers and dual-spindle lathes.

I was entering commands in MDI and if I entered a S or M3/M5 wrong it stopped all spindles.
Do you think it should?

Probably not, but that is built-in error behaviour.
(And if you ignore the error and type in the feed move anyway, you deserve all you get)

[c-morley]

Well It means if you set all your spindles speeds up and then go to adjust one but get it wrong it turns them all off and you need to set them again. you don;t deserve that :)

[c-morley]

So it looks like there is a NML message send for each spindle to control ie there is no message to control all of the spindles at once.
So to fix up python I need it to send a message per spindle if the user specifies -1

Does that sound right to you?

[andypugh]

But, short of not doing the check, I don't see a way to change that.

[andypugh]

Yes, it's a message-per-spindle.

[c-morley]

ok thanks for confirming. I'll work on the python code

[c-morley]

Hit a snag - python doesn't know how many spindles are available. If I send a command for the max amount of spindles, motion gives error on the ones not available.

I guess we could silently ignore the unavailable spindle commands?
Other suggestions?

[c-morley]

Or maybe I should forget it and document that in python there is no 'all spindles' command

[andypugh]

Does Python have a way of finding it out? It seems like it might be a useful thing to know.
It's a bit of a mess anyway, as (like in a couple of other places) the motmod modparam num_spindles and INI NUM_SPINDLES have to match, but the two halves of LinuxCNC can't see each other to check this.

[c-morley]

status reports the number of spindles and the python module can reads that.
but it doesn't automatically. and i think it would be painful to make it do so.
it would seem that maybe the motion component should know how many spindle there are and then you could just tell it to stop all the spindle etc. - maybe i'll look into that possibility.
it seems that motion and the interpreter don't have anyway to know what each other knows about spindle numbers.

[andypugh]

Just document that Python can't do it, and move on then.

[c-morley]

https://github.com/LinuxCNC/linuxcnc/commits/motion-all-spindles-command
This teaches motion that -1 means all spindles, so now python can send -1 to command all spindles.
please take a look when you have time.

[andypugh]

Is this compatible with the existing stuff? (including da3c1c1 )

I am a bit nervous of the big block of duplicated code. How about:
int s0 = spindle;
int s1 = spindle;
if (spindle = -1) { 
    s0 = 0;
    s1 = settings->num spindles - 1;
}
for (i = s0; i <= s1; i++){
.... one common block of code to maintain....
}

edit: I kept considering doing this with the other changes for -1, but thought that in that case clarity of code just won out.
edit2: it needs to be <= in the for loop.
edit3: but then we need to set s1 one less than spindles.

[c-morley]

yes I didn't like it either, but it was clear. I can add a comment to describe the somewhat strange code.

[c-morley]

meaning I'll use your changes with a comment

[andypugh]

I am not clear if this replaces the changes I made? I think it does?

[c-morley]

It doesn't replace your code. But probably you could change the code to send one NML message using -1 as the spindle number rather then several.
Also still need to modify s code to use %-1

[andypugh]

I did S several minutes ago: da3c1c1

[c-morley]

ok I redid it as you asked and tested with my system a bit. please take a look.

[c-morley]

ha hard to keep up :) i'll pull it in and test.

[c-morley]

all seems to work fine !

[andypugh]

I tried to pass a -1 for all spindles deeper in to the depths:
4b7b487
But it doesn't work at all. (whereas prior to that commit is was all nice)

emccanon uses the spindle number as an index into the spindles array:
https://github.com/LinuxCNC/linuxcnc/blob/master/src/emc/task/emccanon.cc#L1888
So I would need a loop in there for spindle == -1, and I would rather have the changes in interp not canon.

[c-morley]

fair enough - so far having multiple motion messages is not anything anyone is complaining about.
I'm not sure if single stepping, steps through each spindle message or ignores them completely.
that's the only possible difference I can think of at the moment.

Are you happy with the motion code, so I can commit ?

[andypugh]

Already committed.

Next question: CSS with more than one spindle? I reckon you could.

[c-morley]

Ya I guess that could be possible.
i could see them slaved to different axes too. - just gets deeper doesn't it
But maybe we should cross that bridge later.

[andypugh]

CSS is explicitly slaved to X.
Which seems like missing a trick, it could work on cutter diameter for mills....

[c-morley]

yes on a mill with a sub spindle on the bed, it might be desirable to slave the sub spindle speed.to Z position

[c-morley]

I modified my second spindle (which is table mounted spindle) from velocity to positional mode.
So now it's an indexer. Running by using the S code as the positional request in degrees.
This was easy to set up and works well. I do notice it you ask for S0 it cancels spindle enable.
Does this really need to happen? I can't see with a regular spindle expecting S0 would turn off the spindle, I would expect it to set the requested RPM to 0. There is not much difference in a velocity spindle but a big difference in a positional spindle.Thoughts?

[andypugh]

I can't help feeling that using S to control position is a bit eccentric when LinuxCNC supports 3 rotary axes.

S = 0 has always turned off the spindle in LinuxCNC. It seems a bit of a big step to change that.

[c-morley]

I'm using it as an indexer not a rotary axis - motion does not manage the indexer.
It's much easier to add to the config and it seems to work well - the S0 is an annoyance in this sistuation.

I know S0 has always done this, but I'm suggesting maybe people relying on this behavior is much rarer (I can't think why you would) and this is a major release so changing is an option.

[c-morley]

To put this another way;
If one expects that S0 automatically adds a M5 (which it basically is doing), then why does issuing S1000 not automatically issue M3 or S-1000 not automatically issue M4. (and I am not wanting this - it's just for discussion)

I've never seen a gcode program like:
M3 S100

S0
M2

It's another case for being explicit - S code is for adjustment, M is for state
Thanks for considering.

[andypugh]

I am inclined to leave this as it is now. Maybe think about changing for 2.9, but it might make sense to leave the current slightly odd behaviour consistent.

[c-morley]

yes I agree - I have a patch for master that needs a little more testing but so far works fine.