Quoting Robert Dennison <[EMAIL PROTECTED]>, on Thu 24 Apr 2008 06:55:34 PM PDT:
> Hi Jim, > > Good to hear from you! > > Okay let's test the breakin at 40 words per minute at 5 characters/word > and 3 symbols/character that's about 100 msec of total processing time > per symbol. That seems pretty reasonable. At 40 WPM a total time > budget of: > > 10 msec T to R This is probably a bit pessimistic. I'd think a decent fast relay can do it in a couple milliseconds. > 20 msec buffer processing I don't think you need to allow for this. > 50 msec DSP chain > 5 msec audio buffer fill & xfer > 5 msec d/a conversion (very generous I'd think) The audio buffering (on both ends) & d/a conversion should all be simultaneous with the rest of the dsp, so I think total latency from wiggle of voltage on antenna port to wiggle of voltage on speaker wire is going to be in the sub 50 millisecond range. Consider 512 byte buffers at 48 ksps.. it takes about 10 milliseconds. If you did NO processing and just passed the buffers through.. that would give you 10 millisecond latency. So, to get to 50 milliseconds, you'd allow 5 buffer copies (presumably with some operations going on during the copy). I don't think dttsp actually does quite that much. > 10 msec air to ear > > seems reasonable. > > So let's see 3x10^9 clock cycles/second X 50 msecs = 1.5x10^8 clocks per > symbol for processing... hmmm... I'll bet our flex programmers do it. I doubt you can actually get one instruction per internal clock. If nothing else, the memory bus will limit you. Though.. dttsp isn't all that huge.. it might fit in cache. > > With at least a 3.0 Ghz machine, I'd say careful testing will show the 5K > can do inter symbol breakin at 40WPM. Certainly with a 2.5 Ghz Core > Duo... Considering I've written Windows programs that stream 10-20 Megasamples/second (16 bit samples) to and from disk with much slower processors on a 66 MHz bus, I'd agree. (and no, not much else was going on while the transfer is taking place) _______________________________________________ FlexRadio Systems Mailing List FlexRadio@flex-radio.biz http://mail.flex-radio.biz/mailman/listinfo/flexradio_flex-radio.biz Archives: http://www.mail-archive.com/flexradio%40flex-radio.biz/ Knowledge Base: http://kb.flex-radio.com/ Homepage: http://www.flex-radio.com/
