On 7/24/2012 9:17 AM, Ralph Stirling wrote: > I don't think it's that simple, Kent. I do not recall > having any realtime delays or errors when I was > using a USB microscope camera on a D510MO with > LinuxCNC. My experience may not be of much > value, though, as I was using a 7I43 with hardware > step generation, so I only had a servo thread. > > For people needing to use USB devices with LinuxCNC, > an FPGA device like 7I43 or 5I25 would be a good $80 > investment to eliminate the base thread. > > -- Ralph > ________________________________________ > From: Kent A. Reed [kentallanr...@gmail.com] > Sent: Tuesday, July 24, 2012 4:08 AM > To: Enhanced Machine Controller (EMC) > Subject: [Emc-users] USB badness, was Re: BASE_PERIOD and Intel D525MW > boards > > On 7/24/2012 5:55 AM, andy pugh wrote: >> On 24 July 2012 03:09, John Stewart <alex.stew...@crc.ca> wrote: >>> I've got a D525mw, and every once in a while it'll throw up an RTAI error. >> I found that my D510 was very reliable for latency unless I had the >> webcam plugged in. >> > There doesn't seem to be any pattern emerging yet as to which USB > devices cause problems and which do not. For want of a better > explanation, I've assumed some USB devices upset realtime performance > because their drivers enable DMA activity. > > If this hypothesis were true, then devices requiring fast service of > large data blocks likely would be bad choices. > > Without more explicit guidelines, my approach has been to avoid USB > connections like the plague. > > Any thoughts? > > Regards, > Kent >
Ralph: I never said it was simple :-) Since the device itself must request DMA service, two USB cameras or what-evers may behave entirely differently with respect to upsetting the real-time applecart if one uses DMA to transfer its data and the other not. I said certain devices *likely* would be bad choices, but so far it doesn't appear to me that one can tell for sure except by trying a desired device in service. I'd be happy to have my conjecture proven wrong. I certainly don't think this is an issue just for Intel Atom-equipped motherboards, whether Intel or some other brand (fer instance, instead of an Intel D510 I have an ASUS AT5NM10-I that I'm pleased with; it's the one in the latency-test results table on the wiki). I suppose it might be more of an issue with certain BIOSes. Again, I'd be happy to be proven wrong. So far, all I see are individual "it (works | doesn't work) for my lashup" messages. Too little information to make an engineering judgment. Regards, Kent ------------------------------------------------------------------------------ Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ _______________________________________________ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users
