Re: 5-axis machine controller





"D.F.S." wrote:

> Thanks all for the info.
>
> I guess what I'm trying to do would best be served by a 2 step approach.
>
> I have a machine I previously described I intend to use as a Printed Circuit
> Board manufacturing macine for personal projects only.
>
> The controller is toast.
>
> If I can simply get EMC to run it with some very basic controls and a simple
> circuit running off the parallel port of a PC I could then use it to build a
> good control board with all the bells and whistles.
>
> One thing I didn't see previously in the source code was a clear cut function
> that does the actual output of the signals to the steppers.
>
> I could easily adapt the source to drive my simple hardware if it does not
> already, if I could find the actual functions.
>
> I'd hope for something along the lines of --  Move_Axis_Z(Direction, Count)

No, it is a LOT more complicated than that.  There are 12 layers (well, maybe
really only 5) of software between the G-code and the step pulses.  EMC
was originally designed for servos,  so the easiest hack was to simulate
a servo system, and then generate step pulses from the middle of the
simulation.  But, you really don't need to know ANY of that to make
it work.  You don't need to look at the code, it outputs either step and
direction
or stepper phase signals from the parallel port.  There is a table of the bit
assignments somewhere on the Linuxcnc web site.

> I could make that work, but as I said I never found such a function, I'm
> sure they are there.
>
> I'd like to toggle the bits on the port and drive the stepper coils directly
> thru FETs. I figure a 3-to-8 ttl Decoder and 3 latches along with a dozen
> Transistors should do it. I have all that stuff close at hand.

Well, you really don't need that, as EMC will output the stepper phases
directly, as an option.

> What particular distribution and version number of Linux will work
> with what version number of RT Linux and what version of EMC will
> work with that?

I'm using Red Hat 5.2 Linux, with the 0.9J RTLinux patch, and the
20-DEC-1999 version of EMC.  I had problems with later versions
that are related to the combination of that old RT-Linux and the TkEMC
version of the GUI.  Later versions of EMC work with the same RTLinux
when you use the xemc GUI.

> I'm all for tinkering, I do however want there to be a point in time
> where the tinkering is OVER for a particular tool.
> At that point, I simply want it to work so I can USE it and have all the
> time I'm putting into tinkering on a new project go to THAT project not
> the tools themselves.
>

I have been using my system with minimal tinkering for almost 2 years, and
no tinkering at all since exactly one year ago, when the 20-DEC-1999
EMC was released.  (I have tried later versions, but always went back to
that one.)

> By the same token, a failure on the PCB machine would be annoying, a failure
> of a REAL machine tool could kill or maim someone.
>
> Is EMC to that point?

I have never had a "failure" with my system.  I have even had the Linux
OS crash (due to that TkEMC / 0.9J interference) without causing any
disturbance of the real time motion control task.  The screen froze, the
keyboard was dead, but the machine kept cutting to the end of the program,
and then stopped.  But, using the versions mentioned above, I have
a totally reliable system.

I think the 2.2.13 kernel with the appropriate RT patch and the latest version
of EMC is also supposed to be rock solid, but I haven't tried to upgrade
yet.

Jon




Date Index | Thread Index | Back to archive index | Back to Mailing List Page

Problems or questions? Contact