On Sat, 3 Nov 2012, Jon Elson wrote: > Date: Sat, 03 Nov 2012 17:42:58 -0500 > From: Jon Elson <el...@pico-systems.com> > 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] Revisiting [OT] measuring real-world latency, > was Re: worsk: sim parport component > > (I think) Kent Reed wrote: >>>> >>>> Logic analyzers, oscilloscopes---I know the drill but I no longer have >>>> easy access to such instrumentation. >>>> >>>> Has anyone come up with simple (e.g., inexpensive) ways to measure >>>> latencies in the range we are interested in - say a range of 1K - 100K >>>> pulses per second with pulse widths on the order of microseconds. >>>> >>>> > I think there ought to be timer registers in the CPU that can be read > each dispatch of
> the thread that would give good data. I think this is what the existing latency test uses. It suffers from being much too optimistic as the CPU count register is always available where memory or I/O hardware may be blocked. This is what I noticed on Atom MBs, the actual physical latency at the I/O pin is much worse than the latency test shows (on the Atom it appears to be related to video updates) Only by reading external hardware can the real latency be measured. This hardware should match the final control hardware (at least as far as I/O vs memory, PCI vs PCIE etc) This is why I suggested (at least for people having access to hardware stepgens) setting the stepgen in velocity mode and setting a reasonable step rate. Now if the stepgen feadback position is read, you have a timestamp for jitter measurement. This timestamp has the advantage that it measures the true latency deltas including any delays through (possibly blocked) hardware. > But, some other methods come to mind. > One would be to make a 1 ms delay, maybe with a 74123 one-shot, and a couple > gates to detect under- or over-run of the time between pulses. It could > blink > LEDs. That doesn't tell the magnitude of the variation, though. You can > build > various things with timers that could do this and bin the time between > pulses > to a row of LEDs. > > Jon > > ------------------------------------------------------------------------------ > LogMeIn Central: Instant, anywhere, Remote PC access and management. > Stay in control, update software, and manage PCs from one command center > Diagnose problems and improve visibility into emerging IT issues > Automate, monitor and manage. Do more in less time with Central > http://p.sf.net/sfu/logmein12331_d2d > _______________________________________________ > 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. ------------------------------------------------------------------------------ LogMeIn Central: Instant, anywhere, Remote PC access and management. Stay in control, update software, and manage PCs from one command center Diagnose problems and improve visibility into emerging IT issues Automate, monitor and manage. Do more in less time with Central http://p.sf.net/sfu/logmein12331_d2d _______________________________________________ Emc-users mailing list Emc-users@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/emc-users
