Re: LF: Dual-speed WSPR

[Markus Vester <markusvester@aol.com>]

Hi Wolf,

 

thanks for sharing your observations. In theory. as the SNR output is scaled to the same reference bandwidth (2.5 kHz), it should be independent of speed and (on average) identical for both. But as you say, the SNR indications for WSPR-15 consistently showed couple of dB higher than the average of the surrounding WSPR-2 pairs.

 

But if this was due to the noise measurement being contaminated by slightly instable QRM lines, and these were spread across more bins in -15, wouldn't that rather raise the quartile more, causing the opposite effect?

 

My speculation is that the signal measurement may not be not true average power (or total energy), but dominated by the highest power, a bit like a peak-holding S-meter. Then with a fading signal, a higher peak will more likely be caught in a longer observation period.

 

Anyway the real question we want to address is probably a different one: If I have limited power and limited time (say a half-hour opening), and my aim is to be decoded only once by a specific DX station, should I rather send two WSPR-15 sequences, or bet on one of fifteen WSPR-2 transmissions making it through?

 

This will not fully be answered by SNR threshold comparisons, and it depends very much on statistical fading characteristics like power distribution (how likely is which amount of enhancement) and autocorrelation (how long will a peak last). So a better empirical test would probably be a dual-speed beacon which spends equal time for either mode, ie. emitting 7.5 times as many -2 sequences as -15 ones.

 

All the best,
Markus (DF6NM)

 

-----Ursprüngliche Mitteilung----- 
Von: wolf_dl4yhf <dl4yhf@freenet.de>
An: rsgb_lf_group <rsgb_lf_group@blacksheep.org>
Verschickt: So, 31 Mai 2015 9:16 am
Betreff: Re: LF: Dual-speed WSPR

Greetings all,

About the different SNR indications of WSPR-2 and -15 :
The question is how WSPR estimates the "noise" vale .. possibly using the 'lower quartile' method described by G4JNT once upon a time, i.e. sorting the FFT frequency bins from a certain frequency range (which.. ?) by amplitude and taking the <N/4-th> bin, corrected by something, as an estimate for the noise level ?
In that case, because I guess the frequency bins are only 1/8th as wide in WSPR-15 than in WSPR-2, the indicated noise level may depend a lot on the drift rate of all those 'nasty little QRM spikes' (in the spectrum), and this may be different for each and every receiver location.

Here are my decodes for DF6NM and G4JNT (maybe Markus wants to compare)..

WSPR-15:

2245 -2 -1.3 0.475824 0 DF6NM JN59 27

2315 -20 -0.6 0.475810 0 G4JNT IO90 33

2330 -8 -1.3 0.475824 0 DF6NM JN59 27

2345 -17 -0.6 0.475810 0 G4JNT IO90 33

0015 -15 -0.9 0.475810 -1 G4JNT IO90 33

0030 -2 -1.3 0.475824 0 DF6NM JN59 27

0100 -17 -0.6 0.475810 0 G4JNT IO90 33

0130 -11 -0.6 0.475810 0 G4JNT IO90 33

0130 -6 -1.3 0.475824 0 DF6NM JN59 27

0230 -5 -1.3 0.475824 0 DF6NM JN59 27

0245 -20 -0.6 0.475810 0 G4JNT IO90 33

0330 -7 -1.3 0.475824 0 DF6NM JN59 27

0345 -31 -0.9 0.475810 -1 G4JNT IO90 33

(btw interesting to see G4JNT dropping so low at 03:45, guess that one would possibly NOT have been decoded in WSPR-2)

DF6NM in WSPR-2:
2328 -11 -1.5 0.475791 0 DF6NM JN59 27
2346 -11 -1.7 0.475791 0 DF6NM JN59 27
0028 -7 -1.7 0.475791 0 DF6NM JN59 27
0046 0 -1.5 0.475791 0 DF6NM JN59 27

0128 1 -1.5 0.475791 0 DF6NM JN59 27

0228 -3 -1.5 0.475791 0 DF6NM JN59 27
0246 -2 -1.7 0.475791 0 DF6NM JN59 27

0328 -5 -1.4 0.475791 0 DF6NM JN59 27

0346 -12 -1.5 0.475792 0 DF6NM JN59 27

From the last transmission (03:30 to 03:45 in WSPR-15): -7 dB
Average between the adjacent WSPR-2 transmissions: -8.5 dB

(-> very similar results as from SM2DJK and DG3LV....  -15 shows 1.5 dB "better")

73,
  Wolf DL4YHF
   (receiver several hundred meters away from houses, but mains-fed -> QRM)

Am 31.05.2015 02:07, schrieb Markus Vester:

Yeah maybe. But at the moment I'd rather keep things as simple as possible, and I anyway wouldn't expect a sharp optimum regarding speed.

 

The idea is to compare the SNR of the slow sequence with the average of the two surrounding fast sequences, received by the same monitor. If the dB results were the same, we could probably exploit the full 9 dB advantage due to the lower threshold (-38 vs -29 dB), otherwise we'd have to subtract the dB difference.

 

Results from first round:

 

 2015-05-30 23:46   DF6NM   0.475791   -11   0   JN59nj   0.5   DL4YHF   JO42fd   359   330 
 2015-05-30 23:30   DF6NM   0.475824   -8   0   JN59nj   0.5   DL4YHF   JO42fd   359   330  
 2015-05-30 23:28   DF6NM   0.475791   -11   0   JN59nj   0.5   DL4YHF   JO42fd   359   330

=> WSPR-15 SNR happened to be 3 dB better!

 

 2015-05-30 23:46   DF6NM   0.475790   -15   0   JN59nj   0.5   SM2DJK   KP03au   1694   15
 2015-05-30 23:30   DF6NM   0.475823   -13   0   JN59nj   0.5   SM2DJK   KP03au   1694   15 

 2015-05-30 23:28   DF6NM   0.475790   -14   0   JN59nj   0.5   SM2DJK   KP03au   1694   15

=> WSPR-15 shows 1.5 dB better.

 

 2015-05-30 23:46   DF6NM   0.475792   +1   0   JN59nj   0.5   DG3LV   JO53gv   502   356 

 2015-05-30 23:30   DF6NM   0.475824   -2   0   JN59nj   0.5   DG3LV   JO53gv   502   356 

 2015-05-30 23:28   DF6NM   0.475792   -8   0   JN59nj   0.5   DG3LV   JO53gv   502   356

=> WSPR-15 shows 1.5 dB better.

 

This is an unexpected result, implying that the advantage of WSPR-15 would be more than 9 dB, instead of less. But on the other hand the dB values may not be telling the whole story, because the threshold for a fading signal might still be higher than -38 dB.

 

All the best,

Markus (DF6NM)

From: DK7FC

Sent: Sunday, May 31, 2015 1:33 AM

To: rsgb_lf_group@blacksheep.org

Subject: Re: LF: Dual-speed WSPR

Markus,

What about the old slow-WSPR by DF6NM? It could be useful to try WSPR-4 for experiments. There is not just 2 and 15. And WSPR-120 on VLF... No problem!

73, Stefan

Am 31.05.2015 01:28, schrieb Markus Vester:

To allow SNR comparisons at different speeds, I will transmit a mixed-mode beacon tonight. It will consist of two WSPR-2 sequences, before and after a WSPR-15 sequence on each odd half hour:

 

hh:28 - hh:30: WSPR-2 475.790 kHz

hh:30 - hh:45: WSPR-15 475.823 kHz

hh:46 - hh:48: WSPR-2 475.790 kHz

 

Transmitter power is currently 25 watts into an antenna with 2% efficiency. Am listening to both modes at all other times.

 

All the best,

Markus (DF6NM)