On Mon, 9 Oct 2017 12:36:20 -0400 Gene Heskett <ghesk...@shentel.net> wrote:
> On Monday 09 October 2017 11:45:58 Nicklas Karlsson wrote: > > > > > There ARM M is tiny and dirt cheap and is very good for real-time > > > > work. I have a robot with four motors and encoders and get 44,000 > > > > interrupts per second > > > > and much better real time latencies then from any Linux based > > > > solution on any platform. > > > > > > > > My solution for robot control is a hierarchy with ARM M for the > > > > lowest level, then > > > > Raspberry Pi 3 connected by high speed serial. > > > > Right now I am at this point with high speed serial communication to > > ARM for the lowest level. > > > > > I did succeed in building a fully rt-preempt 4.9 kernel on it today, > > > took about 4 hours, but I've not yet attempted to boot it as I saw > > > some stuff go by during the build (when I wasn't checking my eyelids > > > for leaks :) that I yet need to turn off. I think, other than fine > > > tuning the kernel build, that convincing the hm2_rpspi driver that > > > it should run on the rock's rk3328, arm64 quad core SoC, running at > > > up to 1.5GHz should be the last major hurdle to making linuxcnc run > > > on it, ... > > > > Happen to know how many SPI ports there may be? > > > This driver that Bertho Stultans wrote, can do 5. Over 2 pin groups IIRC. > However since I was the lab rat, only the SPI1 set has been extensively > tested. There is hardware support for 5 SPI ports, or? > the "rates" are the write, and read, clock rates, corresponding IIRC to a > 41 megabaud write to the 7i90 rate, and a slower 25 megabaud rate for > reading back the data from the 7i90. I figured around 100kbit/s full duplex is what is needed per axis, it should be possible to squeeze into a 1Mbit/s CAN network but it's at the limit. UART would work but setting baud rate for speeds above 1Mbit/s might have accuracy problem. 5-10Mbit/s SPI give me about 5-10 times more bandwidth than needed. > Both of those rates could probably be improved with stronger, as in > faster rise and fall times in the pi's pin drivers, they are a tad puny. Delay is a problem for the signal coming back. > The interconnect cable between the pi and the 7i90 is only about an inch > long ... No need for driver here, in these cases SPI make sense, fast and cheap. Regards Nicklas Karlsson ------------------------------------------------------------------------------ Check out the vibrant tech community on one of the world's most engaging tech sites, Slashdot.org! http://sdm.link/slashdot _______________________________________________ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users
