On Tuesday 18 June 2019 09:50:37 am Greg Bernard wrote:

> I see the 7C81 board is now listed for sale on Mesa's website.  So far
> the only example I've seen of anyone actually successfully running
> LinuxCNC on the Raspberry Pi is Gene Heskett and his account of
> accomplishing this seems too daunting for me to attempt. It seems to
> me that the combination of the 7c81 and the Pi would have wide appeal
> if only there was a ready-to-run image. I'm wondering if the
> developers have any plans to make this available?
>
A ready to run image? Not that I've ever seen. Linuxcnc is so versatile 
that a ready made image for each machine/architecture variation would 
occupy a terabyte or more per release, and then sure as tootin, your 
choice would be left out. If I go put the jessie boot card back in the 
socket, I am running master at version 2.9.0-pre0 on what was originally 
a jessie-lite with a sortof realtime kernel. Both the .ini file and 
the .hal files and the pyvcp stuff were/are from my fingers although I 
have pestered these kind folks to distraction over details at times.

To me thats one of the advantages of linuxcnc, you actually learn 
something, both from writing your own config stuff, and from writing 
your own code to get your own idea done.  This of course is made 
possible by my retired status, no $dayjob since 2002, other than caring 
for my patient but dying of copd lady of 30 years if she makes it to 
December 2nd.  And this weather is hurting her, bad.

Unforch the odriod that was building linuxcnc for me apparently died, so 
now I have built my own right on the pi, using an ssd as workspace so 
I'm not wearing out the u-sd card near as fast.  There are instructions 
on how to do it at

<http://linuxcnc.org/docs/devel/html/code/building-linuxcnc.html>

And they work well on the jessie install, but I'm hitting a lockup point 
during the build when booted to stretch. May be related to building 
uspace on a non-realtime kernel, but thats a wag, not a swag just yet. I 
haven't collected enough evidence as the logs are clean.

So I'm changing direction in favor of seeing if I can build an 
installable but a much later version of a realtime kernel right on the 
pi. I need the practice unless the kernel folks can build a better one 
for amd64 stretch, it likes to ignore the keyboard until rebooted about 
every 2 to 4 days. Thats why I run the bleeding edge, I bleed and report 
the problems, hopefully before they get installed on your cash cows.

The reason for the push to stretch is the much speedier video, definitely 
a lot snappier than the jessie framebuffer which only makes 6 to 7 
frames/second.

> On Sat, Jun 15, 2019 at 10:16 AM Andrew <[email protected]> wrote:
> > сб, 15 черв. 2019 о 18:01 Peter C. Wallace <[email protected]> пише:
> > > I've had LinuxCNC running 24/7 on the 7C80+RPI for about 11 months
> > > now with no
> > > hardware issues. This is with a 40 pin cable about 2" long between
> > > the
> >
> > RPI
> >
> > > and
> > > 7C80. If you have interface problems. its very likely a SI (
> > > Signal Integrity )
> > > issue with the cabling.
> >
> > Thanks for your reply!
> > I actually had a decent link when 7i90HD was not connected to any
> > external hardware. But when I tried to control some steppers as I
> > did with Beaglebone, it was a disaster.
> >
> > > The (7C80,7C81,7I90) SPI interface uses the FPGAs GCLK. This means
> > > it can run up
> > > to 50 MHz or so with a fast host, but it does require good signal
> > > integrity.
> >
> > This is great. So when are 7Cs expected in the store? And what is
> > the price range?
> >
> > For random/longer cable connections using a async SPI interface in
> > the FPGA
> >
> > > that
> > > simulates the synchronous interface would allow better tolerance
> > > of SI issues (
> > > by oversampling / filtering the clock and data in signals ) at the
> > > cost
> >
> > of
> >
> > > lowering the maximum clock speed to say 8-10 MHz. I have not
> > > bothered
> >
> > with
> >
> > > this
> > > since it works reliably with short cables (Actually it works with
> > > longer cables
> > > if the clock source impedance is correct for proper series
> > > termination, I can
> > > drive a 7I90 in SPI mode at 50 MHz send/recv with another FPGA
> > > card and a 2 foot
> > > cable)

An interesting observation, but how is it determined that the clock src 
impedance is the problem? I know its fussy, and a few picofarads to 
ground seems to help. I left a 10x scope probe hanging on it for a 
couple months while Bertho Stultans and I played driver development, 
with Bertho coding and me testing.  Yes, it is that sensitive, and my 
best scope is a tad slow at 1 GHz, a dual trace sampler so I likely 
wasn't seeing a truly accurate waveform. I got the impression the gpio 
drivers in the pi were a wee mite puny, with rise/fall times that didn't 
seem to push the scope much. Ribbon cable is closer to 130 ohms than 120 
but I didn't have any 130's to try.

Cheers, Gene Heskett
-- 
"There are four boxes to be used in defense of liberty:
 soap, ballot, jury, and ammo. Please use in that order."
-Ed Howdershelt (Author)
If we desire respect for the law, we must first make the law respectable.
 - Louis D. Brandeis
Genes Web page <http://geneslinuxbox.net:6309/gene>


_______________________________________________
Emc-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/emc-users

Reply via email to