On Sun, 17 Mar 2013, Tom Easterday wrote: > Date: Sun, 17 Mar 2013 15:41:21 -0400 > From: Tom Easterday <tom-...@bgp.nu> > Reply-To: "Enhanced Machine Controller (EMC)" > <emc-users@lists.sourceforge.net> > To: "Enhanced Machine Controller (EMC)" <emc-users@lists.sourceforge.net> > Subject: Re: [Emc-users] PC for LinuxCNC with Mesa boards? > > On Mar 17, 2013, at 3:30 PM, Peter C. Wallace <p...@mesanet.com> wrote: >> On Sun, 17 Mar 2013, Tom Easterday wrote: >> >>> Date: Sun, 17 Mar 2013 15:15:43 -0400 >>> From: Tom Easterday <tom-...@bgp.nu> >>> Reply-To: "Enhanced Machine Controller (EMC)" >>> <emc-users@lists.sourceforge.net> >>> To: "Enhanced Machine Controller (EMC)" <emc-users@lists.sourceforge.net> >>> Subject: Re: [Emc-users] PC for LinuxCNC with Mesa boards? >>> >>> >>> On Mar 17, 2013, at 1:18 PM, Peter C. Wallace <p...@mesanet.com> wrote: >>>> Other than EPP issues with with the 7I43 and some PCI parallel cards, I >>>> dont think there are any serious compatibility issues with PC -->Mesa >>>> hardware. The real trick is finding current MBs with decent latency. (and >>>> that work with the now pretty old Ubuntu 10.04) >>>> >>>> I'm pretty happy with the ASUS E45M1 PRO (with c6 state and cool&quiet >>>> turned off in the BIOS) >>>> >>>> This gives ~10 usec servo and base thread latencies on a passively cooled >>>> MB >>>> with noticeably better than Atom performance. >>> >>> >>> Peter, >>> Why does motherboard latency matter if using the Mesa card? I thought one >>> didn't need a base thread, just a servo thread. I also have wondered if >>> one even needs the realtime kernel extensions...? >>> -Tom >> >> Because all the servo (or stepgen feedback) calculations are done based on >> the >> nominal servo thread time so if you have a 1.10 ms thread period instead of >> 1.00 the corrections applied by either the PID loop or the stepgen logic will >> be incorrect. > > I see, so the rt can't go away entirely. But a 1ms servo thread time > doesn't require outstanding latency numbers. I thought the critical > latencies were with software stepgen (base thread). I am just trying to > understand if you are saying that low latency (say 10 usec or better) is > required when using Mesa hardware. The Atom boards we are using have fairly > crappy latency for software stepgen (25,000 max jitter), but again, do I > care if I have a 6i25 in there on a 1ms servo thread? > > -Tom >
No you dont need especially good latency numbers with just a servo loop or a hardware stepgen probably 50 to 100 usec would be OK for most requirements. the more latency, the more 'noise' injected into the control loops... Its also should be possible to use the actual thread time instead of nonminal thread time in most feedback calculations to make machines with a hardware 'base thread' less affected by servo thread jitter. That is a a hardware stepgen machine that took 2 ms instead of 1 ms for a servo thread would currently cause a large and completely bogus correction in its current velocity to get applied, when in fact if its was simply slewing even at a high rate, there may be no correction needed. Thats is if correctly done jitter on the servo thread could be made unimportant (except for secondary effects on accelerated motion and loop bandwidth) This has been discussed ad-nauseum in linuxCNC-devel. > > ------------------------------------------------------------------------------ > Everyone hates slow websites. So do we. > Make your web apps faster with AppDynamics > Download AppDynamics Lite for free today: > http://p.sf.net/sfu/appdyn_d2d_mar > _______________________________________________ > Emc-users mailing list > Emc-users@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/emc-users > Peter Wallace Mesa Electronics (\__/) (='.'=) This is Bunny. Copy and paste bunny into your (")_(") signature to help him gain world domination. ------------------------------------------------------------------------------ Everyone hates slow websites. So do we. Make your web apps faster with AppDynamics Download AppDynamics Lite for free today: http://p.sf.net/sfu/appdyn_d2d_mar _______________________________________________ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users
