Re: LF: Simultaneous decoding of WSPR on LF

[Andy Talbot <andy.g4jnt@gmail.com>]

WSPR, as in all the WSJT modes captures the signal synchronously;  an entire block at soundcard sample rate, so isn't affected by IRQs and the like.  In effect, all decoding is done off line, in teh few seconds after teh transmission ends..

Decode failures are most serious when due to sampling rate errors. which is nowadays not much of a problem with WSJT-X adopting the standard 48kHz Fs.  But was a problem with the old WSJT trying to get to 11025 with modern sound cards.

The block decoding uses the sampling rate to define symbol timing, so at the end of a 48 second block, even 0.5% error in sampling rate leads to symbol misalignment at the end.

The other error mechanism, is just timing itself.  The block capture should start on the minute interval, as should teh transmission.  The decoder can handle slippages here of a few seconds either way, but not more than about 3 or 4s IIRC.

Andy 'jnt

On 22 October 2016 at 16:37, Graham <g8fzk@g8fzk.fsnet.co.uk> wrote:

Hugh ,

By 'time'  I was more  concerned  with the  synchronisation , along the time  line ,  ' clock jitter '  in  computer speak .. if sections  of the time  line are displaced due to  IRQ  action or similar , you  could reduce the  system's ability to reach  maximum  - s/n

Opera requires  ~ min of  40%  capture of the  Tx signal ,  along the  time line ,  either  randomly , blocks or 40/60 split , using manchester coding , each data 'bit'  is  self-sync , so it impossible to  'loose' lock.. where as a conventional system would need to  establish  sync at the start ,...... as long as 40% of the opera recovered [or sent !] plays ball ,  it  will decode ,  [ or detect for  dynamic, has the  same 40% requirement]

or in wspr  case , sync is  provided  externally If the  time line is  then jumped , the  system will  no  be  able to recover the  data  , and  as in the   2 pass  wspr , unless  the  two  signals  are  exactly  the  same  in time  and  coherent phase , the  expected  +3 dB gain cannot be  realised .

73-G,

--------------------------------------------------
From: "LineOne" <dhchurch@lineone.net>
Sent: Saturday, October 22, 2016 11:20 AM
To: <rsgb_lf_group@blacksheep.org>
Subject: Re: LF: Simultaneous decoding of WSPR on LF

I checked the PC time and it seems that the Realtek software stops "net time" from loading on occasions so, for another reason as well, I eventually found some older Realtek software that doesn't do this. The software supplied with the motherboard will not share the "line out" with any software if it's used for, as I do, a small speaker amplifier. It has to have the Realtek drivers loaded otherwise no audio at all.

73,  Hugh

On 21/10/2016 23:27, Graham wrote:

May be  down to  time  Hugh ,

Opera  is  free running, so absolute time  reference has no effect on the system  performance ..

Stephan   should have the  solution, his night  garden uses similar things .

The other is  cpu loading ?

73-G,

--------------------------------------------------
From: "LineOne" <dhchurch@lineone.net>
Sent: Friday, October 21, 2016 11:10 PM
To: <rsgb_lf_group@blacksheep.org>
Subject: LF: Simultaneous decoding of WSPR on LF

So far I've been decoding Opera32, WSPR2 and WSPR15 simultaneously on an old IBM "Thinkpad" in the shack. This weekend I've managed a wireless connection to a desktop PC in the house so only a receiver is running in the shack leaving nice, clean signals.

Decoding Opera 32 is going well (3000km best so far this evening) but the PC, with the same operating system and radio software, will not decode any WSPR at the same time. I haven't yet tried running just WSPR-X (for WSPR2) on it's own but it does show a good signal on the waterfall if I arrange a 1.5kHz heterodyne in the shack. Why would a supposedly superior computer refuse to cooperate, I suspect the Realtek on-board sound card and software is  not actually up to the job?

73, Hugh, M0DSZ