Re: [CAD_CAM_EDM_DRO] Re: Black box ??

----- Original Message -----
From: "Jon Elson" <elson@...>

> I seriously doubt it. It will happen, but it will be much longer than you
think
> before they disappear form ALL computers. And, the PCI cards may not be
> that hard to use, especially in my particular setup.

Ok, allow me to rephrase this. Parallel and serial ports will dissapear very
soon on the majority of computers as purchased today. That is the Wintel
platform aka the "PC" from the top 10 PC manufacturers.

> USB may not work at all for the servo-like interface that I have been
> using in all of my designs. There is a limit to the handshaking rate of
> 1 KHz on USB, and that is by the bus's timing, not anyone else's.
> It might work fine if you throw out all the overhead and just use the
> bare hardware. USB requires a CPU on the outboard device, mine
> doesn't. Hot swap causes a temporary lockup of the bus, so you can't
> allow any hot swapping while the machine is running.

Hot plugging or unplugging of anything is not something you should do when
a machine is running.

One way to do this with USB for steppers is let the CNC software do the
toolpath calculations, pump the resulting step / direction sequences to a
bank of FIFO ram that the USB micro will then sequence out to the motor
interfaces. The micro can also service GPIO for limit switches and control
functions. As long as your service rate to the USB device is high enough
that the FIFO doesn't run dry, you have no pauses in the machine motion.
Tool changes and other such interruptions of course would be a pause time
where you would not fill the FIFO until after the tool change was serviced.
1Khz is plenty fast for that cycle type.

For servos, hang a DSP chip off the USB micro and let it do the PID
algorithms and interpolation calculations. Just send it a list of 'goto
locations'. Plenty of appnotes on
TI's DSP site for examples. The micro in the USB interface just acts as a
link
between the PC host and the DSP engine.

> Ethernet is not time deterministic, but it should have enough bandwidth
> for most home users. But, it still can't come anywhere near one machine
> instruction for each byte transferred, in small handshake transfers of a
> dozen bytes or so.

If it's connecting to a 'black box', it only has to service the box fast
enough that the 'box' doesn't run out of work to do and stop the motors
waiting for data.

Like any other problem, put 3 engineers in a room, ask them to solve a
problem and you'll likely get 3 different solutions.

later,
Larry E.