EMC2 running on Raspberry Pi?

I would like to see a LinuxCNC distro for the stock Raspberry Pi.
I do not get why this hasn't happened yet. There's a real-time kernel patch, right?
www.frank-durr.de/?p=203

I don't why it seems like everyone is recommending doing unfocused processes and for support for a specific pieces of hardware instead of getting everyone together to support a standard Pi, with the aim that it be as flexible as LinuxCNC itself.

To see this on both the LinuxCNC forum and the Rpi forum, I feel that I've just walked in on my parents have a moment of failed intimacy. There's something going on here, but I don't know what : )

I had Maschinekit running on RPi2B (latest model). Latency is really bad, ~ 50000ns. And you need a level converter between GPIO and almost anything you try to communicate with (If you don't the GPIOs die sooner or later). You'll need a display and keyboard, too, so you end up somewhere in the range of 120.- just for a crawling slow barebone machine with the equivalent of a parallel port and bad latency. So while it's nice to play with, it's a pain to actually work with.

Better get a used T60 with dockingstation (total ~ 30.- if it's in really nice shape), install Libreboot and have a latency of ~ 2000ns. Or get a T42/T43 and have a latency of ~ 8000 ns out of the box.

Actually, 50,000 nS latency isn't terrible. My first LinuxCNC machine had that latency and was OK even with software stepping. I think that expectations of acceptable latency have moved too far.

linuxcnc.org/docs/2.7/html/install/latency-test.html

Says:
"So, what do the results mean? If your Max Jitter number is less than about 15-20 microseconds (15000-20000 nanoseconds), the computer should give very nice results with software stepping. If the max latency is more like 30-50 microseconds, you can still get good results, but your maximum step rate might be a little disappointing, especially if you use microstepping or have very fine pitch leadscrews. If the numbers are 100 us or more (100,000 nanoseconds), then the PC is not a good candidate for software stepping. Numbers over 1 millisecond (1,000,000 nanoseconds) mean the PC is not a good candidate for LinuxCNC, regardless of whether you use software stepping or not."

I designed and made a batch of level-convertors for Pi (and then went on to never use them).
photos.app.goo.gl/BRkFbfVLtNOBYBW22

I used the GTL2000DL chip (22 bit bidirectional logic level shifter)
datasheet.octopart.com/GTL2000DL%2C518-N...datasheet-134572.pdf
Though it looks like that chip might no longer be current.

But to continue: I am not disagreeing that unless you are really short of space, there is no reason to use a Pi when a small x86 machine will work so much more easily and better.

andypugh wrote: But to continue: I am not disagreeing that unless you are really short of space, there is no reason to use a Pi when a small x86 machine will work so much more easily and better.

Agree totally with Andy, but a USFF PC like a Gigabyte Brix which I use is not much bigger than the Pi (5" square) and comes with a VESA mount so you can mount it in the back of your monitor. Sure you need to buy an ethernet card as well but its a nice neat solution.

I am still running my 3D printer using uspace LinuxCNC on a Pi3, ST Microelectronics L6470 stepper drives directly on the Pi GPIO and an USB-connected STM32F103 microcontroller that controls the heaters and fans (the Pi lacks analog inputs for the thermistors anyway and the 'blue pill' boards are only $1.50-$2 on ebay). GUI is a custom one which is tailored to 3D printing, fits the Pi foundation 7" 800x480 touchscreen and which lacks the Gremlin preview (even a regular PC would suffer from previewing 50MB of short segment G-code)

This setup has been chewing G-code for hundreds of hours without any issues.

Still, I do not think the Pi is ready yet to serve as a reasonable generic CNC controller. It is simply too slow to run a full GUI comfortably, especially when there is a lot of (OpenGL) previewing going on, reliability of the SDcard as rootfs storage is questionable (my 3D printer runs an almost readonly rootfs with a tmpfs overlay), and fairly highres LCD touchpanels that plug into the MIPI port either do not exist or when they do they are terribly expensive.

As soon as you move up from a 3D printer to a lowend router or so a regular x86 provides more oomph for little more money. If it is the small form factor and tinker-ability you are interested in, maybe boards like these might work.

twoflowers wrote: I had Maschinekit running on RPi2B (latest model). Latency is really bad, ~ 50000ns. And you need a level converter between GPIO and almost anything you try to communicate with (If you don't the GPIOs die sooner or later). You'll need a display and keyboard, too, so you end up somewhere in the range of 120.- just for a crawling slow barebone machine with the equivalent of a parallel port and bad latency. So while it's nice to play with, it's a pain to actually work with.

Better get a used T60 with dockingstation (total ~ 30.- if it's in really nice shape), install Libreboot and have a latency of ~ 2000ns. Or get a T42/T43 and have a latency of ~ 8000 ns out of the box.

I agree with all of the above, except on RPI3 the latency is 150000 to 300000. I have the servo thread at 500000 and i never get latency errors.
Usually i use Dell optiplex series desktops (from 40 to 80 euro a piece ) or mini ATX Asus boards with core2duo processors, Latency from 4000 to 15000.
An older Fujitsu Siemens workstation with 8 cores does 1500 to 1800, Celsius is the model but i do not know the number.

How I wrote before, there is one plus for RasPi and that is SPI. Over it is much more easier to communicate with board with external step generator, dac or encoder counters than over ethernet or pci. At least for hobbyists or enthusiasts.

Ethernet is not an option for the RPi anyways as ethernet is via usb.

ozzyrob wrote: Ethernet is not an option for the RPi anyways as ethernet is via usb.

I still wonder who made that decision !!!