Re: [Emc-developers] Questions about porting EMC2

[EBo <eb...@sa...>]

In the spirit of helping out...

> I see that FreeRTOS has been ported to the x86 platform, and might be a 
> better fit in some ways than linux. Also, if the motion controller were 
> moved to a separate micro controller, there would be no need for a 
> real-time OS (a CPU that only runs one program doesn't need an OS-- 
> and might be better off without one).     

I've programmed FreeRTOS a bit.  One of the things you need to be aware of
is that the standard timing loop is only 1us.  To get reasonable fast motor
speeds you will have to use the fast interrupt driver and roll your own
handler.  On the other hand, that is not to difficult.  I posted some
template code for this to the Make Microcontroler forum.  Other than that,
there are a number of unix specific stuff that will probably require a LOT
work to replace the underlying functionality.  If you go this route I wish
you the best.

> [And yes, I'm aware that the RepRap controller software has been adapted
> to other projects. I plan to play with that for a while before deciding
> whether to dive into EMC2.]

I might be wrong about this, but here are some thoughts about EMC2 and
RepRap.

As far as I know the real-time kinematics in RepRap are fairly simple. 
The implication there is that it will work fine for slow actuators but I am
not sure I would trust it for a stage moving at 1000+ IPM.  The faster
something is moving, and heavier it is, the more damage that can happen
when something goes wrong.  Before using the RepRap code on my Van Norman
retrofit with high speed servos, I would really want to know the limits of
their software.  Maybe it is up to snuff... I just do not know.

In EMC, NIST got the physics right -- so you can trust it to control some
way-out-there stuff.  On the other hand, it is big, bulky by comparison,
and can be a bit of a pain to install/configure.

As a note, I've been thinking about taking a stab at extending RepRap
using the real-time version of the Plan 9 distributed operating/file
system.  A version of the Plan 9 file system was ported to the LEGO
Mindstorm controller (called styx-on-a-brick) in roughly a 12k executable. 
The cool thing about this approach is that each of the axis can be
distributed and controlled via any machine which can communicate via the 9p
files system protocol.  If this interest you give me a poke in 3 or 4
months and I will see if I have time to pick this project up (I'm finishing
my thesis research at the moment, and working on a Plan 9 related project
which was accepted for this years Google Summer of Code, so I am insanely
busy).

  Hope this helps,

  EBo --