[hpsdr] Latency

[Glenn Thomas]

Hi Shel.

Dunno. Half a second is a very long time for a 
modern computer. I did an experiment here on my 
SDR1000/XP system. I sent QBF's at 25WPM and 
20WPM from a paddle, using a second receiver as a 
side tone. In the past, I've found that even 20ms 
or so - round trip delay from a mid-altitide (900 
miles) OSCAR - made it impossible for me to send 
CW via a keyer while using the downlink as a 
sidetone. Thus, if there is several hundred ms of 
latency in an SDR system, I should not be able to 
send, using the second receiver as a sidetone.

My result is that I could detect no latency via 
this experiment. I could send QBFs perfectly at 
20WPM and limited only by my ability at 25wpm. 
Thus, the latency on my SDR1000/XP system appears 
to be less than a few tens of milliseconds. In my 
case, the "three or four HUNDRED milliseconds" of 
latency that you cite seems to not be credible. 
Can you point me to a more complete description 
of where the 300ms to 400ms number comes from?

73 de Glenn WB6W

At 12:59 PM 5/23/2008, CT1IZU wrote:
>***** High Performance Software Defined Radio Discussion List *****
>
>Below is an interesting and relevant cut of a 
>recent message on the Star2 reflector. Star2 
>runs a DSP IF at 15Khz on either the old EzLite, H/B AD2181 or Lyle´s dspX.
>
>
>(((((((This note prompted by a review in March 
>Radcom of a couple of SDR boxes -page 51 And 
>some enquiries on general PC-based SDR behaviour 
>on this critical figure of merit.
>
>LATENCY - the time from an event on the input 
>(of an Rx or Tx) until it appears at the output. 
>With these PC/SDR boxes we are talking three or 
>four HUNDRED milliseconds for Rx and same again 
>for Tx. This is comparable with time for sending 
>HI at 25wpm. I had no idea we were talking these big numbers.
>
>You can imagine the impact on trying to work Dx 
>if the soonest your signal could ever get to the 
>Dx after his last key-up was 1 second. And you can
>imagine what a whole second of ambiguity-time 
>would do to doubling in a VOX conversational net.
>
>FYI, STAR latency is approx 12 ms - . But we 
>don't have the burden of an operating system! At least not one that
>Bill G would recognise as such. A major benefit 
>of running software that is only there to 
>support an HF radio. Nothing else.)))))))))))))))
>
>Shel CT1IZU


 1211588894.0