Re: LF: WSPR Reports

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

Some tests that were done a while back on this issue by several WSPR users, suggested, that the WSPR noise was measured in a bandwidth a little over the 200 Hz needed for the signalling- something like 250 - 300Hz.   

 

300Hz wide CW filters made a fraction of a dB difference, compared with an SSB filter used as a reference;

a 200Hz wide filter gave 0.5 - 1dB as far as I remember,  and a 100Hz or narrower filter gave ridiculous errors.   So WSPR reallt ought to be decoded with an IF filter in excess of 300Hz - ie a medium - wide CW filter.

 

The typing of frequencies into windows is not important to the S/W; only affecting the reported and subsequent tansmission frequency.  All the WSPR decoding software does is look at audio tones in the range 1400 - 1600Hz.

 

I really would encourage people to NOT use narow filter when using the mode.    Strong signals within an SSB bandwidth do not influence teh decoding process PROVIDED they a) do not shift AGC (which should really be off anyway) and do not cause overload / clipping of the A/D.    A bog-standard 16 bit soundcard can manage 80dB dynamic range, so things would have to be pretty serious to exceed that in 3kHz.

 

Andy

www.g4jnt.com

 

On 4 January 2011 21:37, Piotr Mlynarski <sq7mpj@toya.net.pl> wrote:

Andy Talbot pisze:

Looking at the WSPR database, F6CNI is decoding me consistently in the 0dB S/N region.   At 391km this can't be realistic, and most likely comes from using a narrow filter - sub 200Hz - before feeding audio to the decoding software.    The WSPR software does need a flat audio response in the 1400 - 1600Hz range, and preferably over the complete SSB bandwidth.
 Andy
www.g4jnt.com <http://www.g4jnt.com/>
 

Andy, LF group

i do not have at hand the wspr source code at this moment but as i remember it is a  'simple' addition/substraction of 1500
in the decoding/encoding scheme as the 'real'  band in the program is between  (-100 Hz; +100Hz)
As i do not  TX (yet!  :)  )  on  137 khz  and use WSPR  only for  RX'ing it is , say,  'convenient' ,  to put your TRX  into CW mode
and use narrow filters to get rid of qrm otherwise being hard to kill  when using wide (ssb) passband

the issue of feedeing the wspr database with correct freq. is solved by typing in the rx window the difference between your bfo offset
and the rf carrier frequency
suppose, you are using ssb mode  and the rf carrier is on 137.500 kHz so in order to rx this,   you type in the rx window 136.000 khz
tx window 137.500 ; your trx is in usb mode and in the range 1400-1600 you look for the wspr signal

now , you set your trx in CW mode and your BFO offset is 600 Hz , so you type  136.900 in rx window  instead of nominal 136.000
and the report yields correct freq. spot   /136900 Hz + 600Hz = 137500Hz/
(probably, in some rx hardware the left/right injection of bfo may be an 'algebraic'  problem... )

Finally,
all the wspr S/N reports are presumably reffered to a 2500Hz wide passband.  of course, the  s/n report when using 500/250 cw filter is useless...
but.... all the modern, black,  trx boxes have variable SSB filters so it is hard to believe that everybody is using a unified 2.4 or 2.7 khz filter
also, those modern boxes have another feature named PBT ( pass band tuning) Using it you may  cut-off a good part of your 2500 Hz filter either from below or above
the desired freqency  so again, the reported s/n in such cases is questionable...
in my view , the wspr is a weak signal mode - the real challenge is , therefore, to get positive decode of a transmitted signal using all accessible measures
(filters,amplifiers, etc...)   thus making the reported s/n  rather of secondary importance ..
the above statement was reffered to 137/500 khz bands.
on classical HF bands things look probably better , being more unified ..
                                                                                                                              73, Piotr,sq7mpj
qth: Lodz  /jo91rs/