Re: Survey Please read

> From: hansw <hansw@...>
> This is no criticism, but you nailed it. I don't want the hassle of all
> of this,
> I know Windows well enough the do everything I need, form writing the
> device drivers
> through to high level programs. For my CNC part, I like the FlashCut
> idea, but think they made it
> out of reach for the hobbyist, so I'll just do my own version. Pity the
> people doing EMC did not think of that instead of
> fixing themselves to a OS because of the need to "apparent" more
> accurate timing.

Actually, both Flashcut and Ahha (and maybe others as well) use external
hardware to help them generate their step pulses. As far as the EMC goes, the
main consideration was low cost. That meant using off the shelf components,
preferably mass produced PC compatible items like parallel port boards. I
guess we didn't appreciate the magnitude of the problem until we started to
test the code. The biggest problem is maintaining high resolution control of
the step frequency as the step rate increases. As an example, suppose you are
generating your step output by looping through a section of code decrementing
a delay counter on each pass, and toggling the step bit each time the counter
underflows. Let's say you are using a delay of 100 and you need to speed up
slightly, so now you switch to using 99. This is a 1% change, no problem. As
you accelerate and go faster and faster the delay values get smaller and the
difference between a delay of 5 and a delay of 4 is 20%. This is a big
problem due to the inverse relationship between speed and torque in stepper
motors. You reach a point at which the motor doesn't have enough torque to
make the jump to the next step rate and it stalls. The only way to increase
your resolution is to get through your code loop faster, and there you are
limited by the speed of your computer. Offloading all of this CPU intensive
stuff to external hardware is a great idea! Something like an 8254 CTC chip
on a plug in board, or even one of the new DDS chips like Analog Devices
makes could be used to generate step pulse trains at higher resolution, and
greater rates than can be done in software alone. If you go that route
(external hardware) there is the additional problem of accumulating the count
of step pulses for each axis so that the control software can keep track of
position. Also the comm link between the external hardware and the PC has to
have sufficient bandwidth to allow reading all the count registers and
writing step frequency updates at least every few milliseconds.

> It would be neat to see some actual figure on this constant claim for
> accurate timing, I remain unconvinced.
> This is not a flame for any EMC nut to jump on, it's a statement of
> doubt.

Well, here is a pretty good study (with actual figures) of the real time
performance of the standard version of Windows NT:

http://www.hyperkernel.com/paper1.html

From the above link:

"Windows NT, throughout our tests, is shown to be fairly deterministic when
the real-time application is the only application running on the operating
system. Once other NT applications are introduced or a mouse movement or any
other user interaction occurs, the Windows NT determinism begins to degrade."


Test data on the performance of RT-Linux can be found in these two articles:

http://rtlinux.cs.nmt.edu/~rtlinux/papers/usenix.pdf

http://rtlinux.cs.nmt.edu/~rtlinux/papers/lj.pdf

> I don't believe something just because someone told me they think it's
> needed. And if it's government funded I believe it even less, remember
> all that scare about electromagnetic radiation from power lines.... Well
> a couple of months ago the guy that has been doing the "research" was
> debunked, from the new cip I heard ( that;'s all one gets here i the
> USA) he was making up the "data"..... That was government money also...
>
> That does not mean to say the same thing is happening here, there are
> needs for stadnards, but I'm not so sure it needs to be funded by tax
> payer money. They tried stuff like that in other countries and called it
> something else...

A short time ago, in another forum, I made remarks addressing a similar issue
which I believe I can reuse here:

*****
Well, from:

http://www.nist.gov/public_affairs/guide/glpage.htm

' The National Institute of Standards and Technology was established by
Congress "to assist industry in the development of technology ... needed to
improve product quality, to modernize manufacturing processes, to ensure
product reliability ... and to facilitate rapid commercialization ... of
products based on new scientific discoveries." '

<snip>

I guess I should state for the record that I'm a registered Libertarian. If
the Libertarian party ever becomes the majority party on the US, we will seek
to drastically reduce the size of the Federal government. As far as the
Commerce Department and NIST are concerned the only relevant reference is the
Constitution itself:

Section 8. The Congress shall have Power:

[3] To regulate Commerce with foreign Nations, and among the
several States, and with the Indian Tribes;

[5] To coin Money, regulate the Value thereof, and of foreign Coin,
and fix the Standard of Weights and Measures;

Section 9, paragraphs 5 and 6 having to do with interstate commerce may also
apply here (they might as well, they've been stretched to imply that the
government has authority over everything that might conceivably have been
transported across state lines!).

Until such time as the Libertarians achieve political dominance, and the
future of NIST becomes a matter of debate (or an asteroid strikes the earth
and kills us all, an event of equal likelihood), the logical thing to do is
help the folks at NIST to "ensure relevance and focus in [their] programs"
(from http://www.nist.gov/public_affairs/nist_mission.htm ).
*****
I'm not sure this is a sufficient reply to what you have said, however it's
3:30am and I'm not in good enough form to do any better right now.

> After all of this I perhap will try loading that RH5.2 that I got with a
> book for $5 and a second hand book store...
> But it will be on a straight single CPU system. I have not heard of
> Linux doing mutli CPU's.
> BUT THEN IS HAVE NOT BOTHERED TO RESEARCH IT.

For a nice overall comparison see:

http://www.sunworld.com/sunworldonline/swol-08-1998/swol-08-linuxvnt.html

Matt