Re: LM628

Jon Elson wrote:

>

> This sounds very interesting. I had a strong interest in the LM628/629

> For plain linear moves, you just
> give the 'go' command as close to simultaneously as possible, and
> it should do fine. But, on a multi-axis chained move, if one axis
> gets a bit ahead of the other, you want the slow axis to speed up,
> or the fast axis to slow down, just a little. But, you don't want the
> faster axis to go into 'ramp down to stop' mode, or it will really
> muck up the motion. So, I never built a system with the
> LM629 chips, because I couldn't be sure this problem wouldn't
> come to haunt me.

Well, the board I have has 3 628 chips and the buss decoding, timers,
and host interrupt support hardware.

Tech80 seems to know more about the chip than National does! It is true
that you can't change accel on the fly, but you can issue a coast
command then as soon as the busy bit goes low you can change all motion
parameters. They (Tech80) recommend running in velmode and going to
coast at a breakpoint for maximum multi-axis interpolation performance.
They also suggest not waiting for their interrupt parser to figure out
that the busy bit is low (100 microseconds after a write) and just shove
the data with careful timing. Min command update time is about 1000
microseconds using this method.

Position mode will keep axes synchronized within one read cycle but as
you say results in choppy motion. I do observe that it is smoother than
one might think with proper filter tuning however.

Well, the microseconds add up and you can see why I am interested in
the potentially higher performance of EMC.

As for the generation of long straight edges or ways it is just using
the principles of hand scraping but using a hand grinder instead of a
scraper. One can easily take off fractional thousandths with a power
grinder. The process is neither easy nor quick but it is cheap and very
accurate.

Sounds like you are familiar with bearing design but if anyone else
is not I have an autocad drawing of a preloaded duplex angular contact
bearing block for 25 mm ballscrews that costs about $30. These cost
$300+ from ballscrew manufacturers. I decided a big box full of these
would really cut into the beer money so I sentenced myself to stand at
the mill and surface grinder for a few saturdays. I don't want to send
the drawings as attached files but might mail.

Oh, I bought about 11 meters of new THK 45 mm linear rail with 13
Bearings for $1100 from Mann-made Enterprises in Addison ,Il. I'll post
the number when I can find it.

--
Leslie M. Watts
ITW Advanced Technology Group
(847)657-4559
http://www.netcom.com/~leswatts/leswp.html