Re: EMC compatible computers

[jmkasunich-at-ra.rockwell.com]

John Sheahan wrote:

>> Do you mean a dedicated box near the machine, connected
>> to a PC that is running EMC?  Or a dedicated box near the
>> machine that _IS_ the PC running EMC?

> I reserve the right to decide case by case..

That's one of the great things about an open solution like EMC.

> CPU's are cheap. Monitors and chairs and desks cost.

KVM switches are a _good_ thing ;)

> But I would consider non-real-time on the unix box
> used for surfing/mail/work while dedicated hardware
> (rather than rtlinux) does the rt stuff.  Either is
> a valid solution - but I think I'd expect a much
> longer useable lifetime from the diskless pure HW
> solution.

True.  There is a difficult balancing act here.
On one side, you could put almost all functionality
in dedicated hardward optimized for long life.  That
route has high hardware development time and cost,
and high hardware prices, but can offer better
reliability.  The other side is to use hardware only
for the things that cheap PCs simply can't do.  Much
lower hardware cost, both development and purchase,
but more likely to be broken by the continued
"progress" of the PC.

>> 1)  Allen Bradley 6180 industrial PC.  Picture at:
>>     http://www.ab.com/opinter/eoi/icb/rac6180.html
>>     200 MHz Pentium, 128MB, in industrial enclosure.
>>     Has 14" flat screen, full alphanumeric membrane keypad,
>>     all sealed, you can splash it with coolant and metal
>
> Australia generally has a less endowed surplus market
> unfortunately..

I've been lucky.  My employer is like many large corporations,
it's cheaper for them to trash perfectly good equipment than
to find a use for it.  The amount of waste is just disgusting.

>> Used 200MHz motherboards are free, but if you want to
>> microstep, you might want something faster.  I like
>> the idea of spending $200 for Jon's board, and being
>> able to use a $0 motherboard.  Others might prefer to
>> put the $200 into the computer, so they have enough
>> speed to generate step pulses in software with no
>> external "helper" hardware.

> Sure - i'd usually choose the HW version too.
> I've been doing other things for a while, not familiar
> with Jon's design, and he has every right to own his
> implementation details.

> But I would suggest that the hardware description language
> code (RTL) behind the implementation is the thing you want
> to maintain.  If (when)  the FPGA goes away - just resynthesize
> for the new one.
> Its the hardware version of  source code - just recompile
> for the new platform.    Thats what gives hardware long-term
> maintainable for me.

As long as you have the compiler for the new FPGA.  Your
approach is certainly good for any but the smallest garage
shop.

>> One perspective - there are two kinds of maintainable
>> hardware in the long term.
>>
>> A)  Simple boards built with commodity parts, with
>>     complete documentation.

> Whats a commodity part now?  transistors, inductors,
> resistors are fine.   IC's come and go.

7400 series logic has some staying power.  The families
may change (TTL, LS, HC, AC, etc), but if you stick with
the mainstream parts, the design can be pretty stable.

> FPGA's come and go like any IC. The RTL has worked
> for 15 years and loosk to have that long left.

Yes.  For any design that requires LSI, RTL is probably
the only long term solution.  That still means that you'll
have to re-do the schematic and board layout, and build
a new board when the new FPGA comes out.

Mariss at Gecko is bucking the trend toward LSI.  He uses
only commodity parts, and avoids programmed parts of any
kind.  His designs have higher parts count than modern LSI
based designs, but the parts are cheap and small.  He also
uses the "simple and dumb" approach, doing in hardware only
what must be done in hardware, and letting the PC do the
complex stuff.  It will be interesting to see where he is
in ten years compared to other manufacturers...

> (he says with a 120 year old screw cutting lathe in the
> next room - which unfortunately reached EOL a while back....)

I have a 28 year old lathe (South Bend) and a 62 year old
mill (Van Norman).

>> C)  Mass market parts that adhere to a standard interface
>>     and can be replaced with work-alike units later,
>>     IDE disk drives, for example.  This is chancy though,

> Hmmm. These are decade-long solutions only. Thats not long
> compared to my lathe. Not that I have a much batter solution.

It is a very difficult problem.  Perhaps we simply have to
live with the fact that machine controls will need to be
completely rebuilt or replaced several times during the life
of the base machine.  I don't like it, but it may be the truth.
Electronics is just so ephemeral compared to cast iron.

John Kasunich