Hello Rick, >When I use PSK31, isn't the bandwidth pretty much set by the baud rate The baud rate and the windowing (square, cosine...) chosen...
>down to say 100 Hz, is this changing anything in terms of its practical >ability to work deeper into the noise? There will be no change if you have only gaussian noise, but if you have QRM it is another story. >Or do they use a 3000 Hz BW for testing purposes and compare modes that way? Yes to compare. For example, I want to compare modes at S/N=-10 dB: I send a a signal of 1 mWatt and 10 mW of noise in 3KHz (so 3.33 mW per KHz). Now among this noise you can send your 1 mWatt signal in the way you want (RTTY, PSK...), the bandwidth you want (within 3 KHz) and also the coding you want. The judge will be the error rate: 2% is good, almost 100 % is bad. >Doesn't this tend to favor the wider modes when it comes to claims of SNR? No it is indifferent. 73 Patrick ----- Original Message ----- From: Rick To: digitalradio@yahoogroups.com Sent: Sunday, October 28, 2007 1:20 AM Subject: Re: [digitalradio] Re: QEX Article on HF Digital Propagation The part that I don't fully understand is the bandwidth calculation. When I use PSK31, isn't the bandwidth pretty much set by the baud rate and width of the signal? Often it is expressed as around double the baud rate or ~ 60 Hz. Now if I have my rig wide open with the 3.6 kHz bandwidth and tighten it down to say 100 Hz, is this changing anything in terms of its practical ability to work deeper into the noise? Or do they use a 3000 Hz BW for testing purposes and compare modes that way? Doesn't this tend to favor the wider modes when it comes to claims of SNR? 73, Rick, KV9U Rud Merriam wrote: > Jose, > > Just as you were posting this message I was stumbling on a web site that > agreed with your comment. > > With further searching I think I have the relationship. The QEX article has > the statement that to go from the 3kHz bandwidth used you "subtract 34 dB > and add 10 log of the desired bandwidth in Hz". But I think he has it wrong. > > > My search found that you adjust by taking 10log(BWoriginal/BWdesired) and > adding it to the given figure. I think the author neglected to consider that > the power of the signal is unchanged during the calculation. The result is > you need to add 19.82 dB to the reported values to obtain the SNR for a > 31.25 Hz signal. > > As proof (I hope <g>): > > Signal: 3000 Noise (3kHz): 3000 SNR(dB): 0 > Signal: 3000 Noise (31.25Hz): 31.25 SNR(dB): 19.82 > > Where the noise is 1 Watt-s per Hz. > > The article reports that PSK-31 work down to -12 dB in AWGN this actually > means it work to 7.82 dB. The channel capacity for that SNR per > Shannon-Hartley is 88 bps. PSK-31 attains less that half the channel > capacity. > >
