On 4/29/2013 1:16 PM, Matt Shaver wrote: > On Sun, 28 Apr 2013 12:55:17 -0500 > Jon Elson <el...@pico-systems.com> wrote: > >> Is there a consensus about BeagleBone vs. Raspberry Pi? > > I feel the need to defend the Pi, Cubie, Olimex, et al boards, since it > appears that no one else will :)
Excellent points...I have a few comments of my own in-line: > > Several areas of concern are worthy of our attention, which argue for a > more circumspect outlook regarding the explosion of new ARM based > systems that can potentially host linuxcnc: > > 1. In the case of the Pi, numbers matter I think. I believe you're right about the numbers, and the Pi takes this hands down. That doesn't change the fact that any of these boards could disappear overnight, or be replaced by the "next new thing" at a moments notice. > 2. The PRU is a TI specific thing. Yep...the PRU is a TI specific thing. That said, it's successful enough to be in it's second iteration across a family of several different ARM based SoCs. These parts are also aimed at the industrial market rather than the more fickle consumer market, and I suspect will have a longer life than the SoC at the heart of the Pi, or the Allwinner chip-of-the-day. > 3. Cost: Pi=$25-$35, BBB=$45. I'm not personally concerned drastically about the cost. I liked the 'Bone more than the Pi for CNC before the price got cut in half. > 4. If linuxcnc3 supports distributed processing, and network > interconnection/cooperation, then these cheap little boards could be > spread out in a cluster, which is much less practical with even small > PCs. The small size, low cost, and low power requirements make this > very attractive AND allows each different type of little board to be > applied where it makes the most engineering and financial sense. This > also argues for not narrowing our focus strictly to the BBB. I totally agree here...but the gory details need to be worked out first (see more below for #5). > 5. Most of this debate is moot, because the REAL FUTURE is (IMHO) going > to be in these combo ARM+FPGA chips like the Xilinx Zynq and Altera > Cyclone SoC devices. This type of device will likely render the > existing boards we're looking at now obsolete within a few years. Any > attempt to predict the future past about 5 years is probably futile. I agree with this in spirit...IMHO the ARM + FPGA combination is going to simplify the current dizzying array of dedicated function ARM SoC parts into a more comprehensive ecosystem. But that's the future, not today. Today, what *I* am trying to get from LinuxCNC is a functional replacement for the typical Arduino based 3D printer controller. So what I need is: * LinuxCNC ported to something that runs on a board 'similar' to an Arduino (ie: small, low-power, and fairly inexpensive). In today's world that pretty much means an ARM board. Or maybe some Atom based tablet reference platform, but the ARM boards are easier to come by. * Hardware/software support for 5+ channels of step/dir generation at rates that exceed the current Arduino limits (I'm targeting at least 100 KHz step rate, ideally higher). * ADC support for 3+ thermistors * PWM generation for controlling the heaters (this is low-bandwidth, and can be done by just about anything). So...the Pi falls flat on ADC support and step/dir generation (have you checked the worst-case latency figures for ARM even with a Xenomai enhanced kernel? 50+ uS). The TI AM339x on the 'Bone has enhanced hardware timer support compared to the Pi, but the PRU is really what makes this SoC a great intermediate step between a plain vanilla ARM core and a full-on FPGA solution. I still think you need an FPGA if you're running with encoders and servos (except the 'Bone could _maybe_ drive a 3-axis servo system if you can use the hardware encoders built-in to the SoC, but I'd still prefer an FPGA). Given the Xilinx Zynq is apparently made of solid gold (based on their board and chip pricing), and the Altera SoC parts aren't quite real yet, the 'Bone looks like the best choice for a LinuxCNC ARM platform _TODAY_. But I wouldn't want to be too tightly coupled to *ANY* particular ARM chip/SoC solution...this environment is rapidly changing. As for the FPGA option...in addition to attending the Altera/Arrow SoCKIT seminar next month (where I'll be getting a Cyclone-V SoC evaluation board with ES silicon), I just found out I may be using the Altera FPGA + ARM SoC in an official work (ie: paid job) project. I really do see the FPGA + ARM SoC parts as the future, but there are a few steps to be traveled along the way... :) If you know of another existing board (ARM or otherwise) that can support at least five 100+ KHz step/dir channels and has 3+ ADC inputs (all without having to add external hardware), I'd *LOVE* to hear about it. I'm quite sure I am not aware of every low-cost dev/eval/hobby board available. > Having said all this, I did get a BBB in the mail yesterday, and I will > now play with it... :) *jealous* ...mine are still in transit! :) -- Charles Steinkuehler char...@steinkuehler.net ------------------------------------------------------------------------------ Try New Relic Now & We'll Send You this Cool Shirt New Relic is the only SaaS-based application performance monitoring service that delivers powerful full stack analytics. Optimize and monitor your browser, app, & servers with just a few lines of code. Try New Relic and get this awesome Nerd Life shirt! http://p.sf.net/sfu/newrelic_d2d_apr _______________________________________________ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users
