Re: Black Box Goals

(38538) Luc,

Not only is it possible to do with an 8-bit processor, it has
actually been done before.

I designed a product that did exactly that with a 4MHz Z80 that sold
many hundreds of units. The differences were the update rate was 256
per second and had only 2 axis. Otherwise it has the same specs and
capabilities as in the "Black Box Goals". The unit is still being
sold today; it is the Centent CN0170.

The difference is it has a proprietary command language and is
overpriced in my opinion. In fact I am borrowing from my own work of
a decade ago for the all-important motion algorithms to speed up the
design cycle.

The difference today is the processor I'm going to use (Z-80 based;
what a surprise) is 6 times faster and has 16-bit harware multiply.
I'm trading some of that speed in for a 4 times higher update rate
(1,024 vs. 256 / second). The Z80 required writing muliplication
routines. The best I could do was about 900 clock cycles for a 16X16
multiply; it now takes 12 clock cycles and saves an enourmous amount
of time. This savings I'm trading in for an additional 2 to 4 axies.

The other difference today is surface mount technology. We have an
SMT production line which permits much lower prices due to automation
and minaturization. Small is cheap, big is expensive.

By using a standard language (G-Codes) and modularization (have only
the parts you need in it), makes it a viable proposition.

The "Black Box Goals" does not cut new ground, it improves on what
has already been demonstrated to work. This is not a "pie in the sky"
thing.

(38484) Bill,

That is the way I'm thinking of going with this. Open "architecture"
on both the hardware and software. Every additional axis places a
burden on what is essentialy an interrupt driven routine. It all has
all of 500 to 750 uS to execute all the axies, yet leave enough time
to have the background program to do its thing to keep the interrupt
routine "fed".

(38485) Ron,

You bet, I'm going for the first 4 G-Codes only. The currency of
exchange has to be steps; it can't know what your leadscrew pitch is,
whether the motor is a stepper or servo, what the steps per inch is.
The serial protocol is undefined yet; I only know the bandwidth will
be low.

(38495) Ray,

The only way it will get skinny is if it ia a very tiny piggy bank.
This will be cheap.

(38502) Chris,

The advantage with a connector per axis card is it is easily
identifiable as belonging to a particular axis, it is easy to switch
a motor/drive from one axis to another, unplug an axis while the
others stay connected, etc.

(38505) Ron,

Feature creep or flexibilty? I want to keep options available in the
hardware design stage. Hardware design changes are very expensive
later on when some feature that was not initially included becomes a
necessity. Once a design is comitted to, it is very hard to "open it
up" to incorporate an oversight. On the otherhand, some feature that
turns out to be uneeded simply winds up as an unpopulated area of a
printed circuit board. Better the latter than the former.

(38530) Mr. Sausage,

The package will be very small, maybe about the size of a Mitchner
novel in paperback. Black anodized aluminum enclosure of course.

(38537) Mike,

If the thing is alive, it has to service an interrupt 1,024 times a
second. This results in an I/O R/W every interrupt. This can trip a
re-triggerable one-shot that lights an LED.

Mariss