Re: [CAD_CAM_EDM_DRO] Re: Black box ??

mariss92705 wrote:

> Hi,
>
> The "black box" project is what a number of people on this group
> (myself included) have taken on independently to get around the
> bottle-neck of having a PC generate step pulses for CNC motor drives.
>
> Step pulse timing (frequency generation) is a task that places an
> inordinate burden on the PC. The general consensus is this task
> should be off-loaded onto external, dedicated hardware.
>
> This hardware (the black box) would generate the requisite
> frequencies (step rates), freeing the PC to do more intelligent tasks
> rather than being tied down with this donkey work.
>
> Once that had been resolved, the question became how much should the
> black box do? Should it just generate frequencies on command from the
> PC

If this is all you want to do, then I've already done it! I have 4 encoder
counters, 4 digital rate generators specifically designed for step pulse
generation, plug 16 opto-isolated digital inputs and positions for up to
8 solid state relays, all on one board. This is essentially implemented
on one chip, a Xilinx FPGA. (There is one crystal oscillator to stabilize
step rates, one chip to drive some LEDs and properly reset the chip on
power up, and a one-shot to create a watchdog timer.)

This board connects to the PC's parallel port, and uses the IEEE 1284
protocol to keep CPU-controller transfers fast.

It can be used either open loop, by counting its own step pulses, or
closed loop, with encoders on the motors or XY table.

Step frequencies up to 300,000 steps/second can be generated, or
as slow as one every 1.5 seconds. Step timing resolution of 100 nS
is maintained at all speeds. Step pulse width and setup and hold
times between step and direction signals can be programmed by
software. Each timing can be set from 100 nS to 25.5 uS.

> or should it include the motion routines themselves? If so should
> it include a display and a keyboard as well?
>
> If you are of an ironic frame of mind, you would appreciate the full
> circle back to a PC that extension took. That was fine though, it
> explored fully what role the proposed project should fulfill.
>
> There are other parallel efforts that I am aware of, but the
> conclusion I have arrived at is it will require the use of a
> microcomputer no matter what the level of "smart or dumb" the box
> takes.

No, I have PROVED that this is NOT TRUE! There is NO CPU of any
sort in my design. The bus protocol for IEEE 1284 (EPP printer port)
is pretty simple but robust and VERY low overhead. PC motherboards
permit a byte to be transferred either way with only one CPU instruction.
I DEFY anyone to get 100 nS pulse resolution on 4 axes with any CPU
less than a "Cray on a chip".

> Once that is recognized, it becomes a matter of communications
> bandwidth that determines the level of "smarts" the box must contain.
> Concurrent with that is the requirement that it also provide the
> maximum flexibility and value for the price (read cheap).

Using a 333 MHz Pentium II and EMC to control my board, the
control loop reads the current position, updates the net step rates
and writes them out to the rate generators in 50 uS. With this CPU, it
would be practical to run the loop at 10 KHz.

You could run it on a 100 Mhz Pentium classic with a 1 KHz update rate
with no problem.

> It will be very interesting what others will come up with. I belive
> in the concept enough to have budgeted for a project that will take a
> year before it makes a profit.

Two or more of these boards can be connected to the same parallel
port if more axes are needed.

See http://pico-systems.com/univstep.html for more info on it.
I will be demonstrating it at the NAMES show April 27-28 in the
Detroit area.

Jon