|
The SNR threshold 50% successful WSPR decodes is
between -29 and -30 dB (2.5 kHz), wheras the QRSS messages required a couple of
dB more (-27 to -28 dB). In Rik's challenge, completely random messages
were used, and only completely correct readings were counted as success. Due to
the structure of Morse code (dashes are better visible than dots), partial
"decodes" are often possible at lower SNR, which often allow conducting a QSO
using some a-priori information and guesswork.
Rik also looked at Opera versus WSPR, and found a 6
dB deficit for Opera at same peak power. That was probably still in an early
stage of Opera development, and the decoding abilities have been improved since
then. My own tests with Opera v1.5.6
http://df6nm.bplaced.net/opera/Success_rate.png
got
50% successful Op-32 decodes at -40 dB average SNR. This scales to -28 dB (av)
or -25 dB (PEP) at Op-2 speed, i.e. a 1.5 dB improvement since Rik's blue curve.
However, at same average power, Opera-2 is still 1.5 dB weaker than
WSPR-2 (or 4.5 dB weaker at same peak power).
Including the volume of conveyed information, WSPR
wins another 2.52 dB (50 bits versus 28 bits), and it is also slightly shorter
than Op-2 (110.6 vs 122.4 s, another 0.44 dB). Thus alltogether the difference
is 4.5 dB at same average power (i.e. Op needs 2.8 times the energy per bit), or
7.5 dB at same PEP (with a given TX, Opera needs 5.6 times as long to send the
same amount of information). Minimum Eb/N0 values are about +7.9 dB for WSPR and
+12.4 dB for Opera.
Regarding correlation detections, my measurements
using coherent signals showed that opds can go about 8 dB lower than the Opera
decoder. For comparison, Opera's Dynamic Deep-Search believably claims to go 5
dB below the decoder.
Sorry for reiterating this topic
again...
Best 73,
Markus (DF6NM)
Sent: Sunday, November 15, 2015 10:37 AM
Subject: LF: WSPR DECODES
MF
I have been observing WSPR signals this past couple of days on 474.2 Khz
and although most are Decoding there are a number of weak signals visible on
the waterfall that do not decode. I am in a quiet location so noise is not
a problem. My clock and input are set up as specified.
Had these station been using QRSS the copy would be perfect.
also
the same applies to Opera signals visible on the waterfall but do not
decode, usually weak.
QRSS has the advantage that the raw signal observed is immediately readable
on the screen even the barely visible.
G3KEV
|
