From: Rik Strobbe <Rik.Strobbe@fys.kuleuven.be>
To: "rsgb_lf_group@blacksheep.org" <rsgb_lf_group@blacksheep.org>
Date: Tue, 31 Jan 2012 23:36:46 +0100
Thread-Topic: Re: Opera v qrs evaluation
Thread-Index: AczgR5WSDdQmLlHvSKeTdmVrxokIRAAHOfRO
Message-ID: <BF4A524700075746A6467658DFC7102CB0B68CCD6C@ICTS-S-EXC2-CA.luna.kuleuven.be>
References: <006201cce044$06c16f80$0401a8c0@xphd97xgq27nyf>,<3A9A60CAE4EB4355A5B0A30CDA0F450A@JimPC>
In-Reply-To: <3A9A60CAE4EB4355A5B0A30CDA0F450A@JimPC>
Accept-Language: nl-NL, nl-BE
Content-Language: nl-BE
acceptlanguage: nl-NL, nl-BE
MIME-Version: 1.0
Subject: LF: RE: Re: Opera v qrs evaluation
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: quoted-printable
Sender: owner-rsgb_lf_group@blacksheep.org
Precedence: bulk
Reply-To: rsgb_lf_group@blacksheep.org
X-AOL-SCOLL-URL_COUNT: 0  
x-aol-sid: 3039ac1d40924f286d4c7992
X-AOL-IP: 195.171.43.25
X-AOL-SPF: domain : blacksheep.org SPF : none


Dear all,

I did an atempt to compare Opera with WSPR, QRSS3 and DFCW10.
I did it by generating audiofiles that contain a Opera2, WSPR, QRSS3 and DF=
CW signal of identical amplitude (but slightly different frequency and add =
a known amount of (white) noise to get a known SNR.
I started from -20dB and then down in 1dB steps, till -33dB. For each SNR l=
evel I generated 5 audiofiles.
All was done by a small application I wrote.

Results:

WSPR:=20
- 100% (5 out of 5) decodes as low as -29dB SNR.=20
- 1 out 5 decode at -30dB SNR
- no decodes at lower SNR
- SNR given by WSPR was always within a +/- 1dB range of the calculated SNR

Opera2:
- 100% (5 out of 5) decodes as low as -22dB SNR
- 4 out of 5 decodes at -23dB SNR
- no decodes at lower SNR
- SNR values given by Opera2 differ about 4dB with the calcualted SNR (Oper=
a2 gives -24dB at calculated 20dB ...)

QRSS3 / DFCW10:
Here comes the tricky part. While with the WSPR and Opera software it is si=
mple to see wether there is a decode or not, this is not the case with QRSS=
 and DFCW.=20
For this modes the operator's experience, both for setting the right parame=
ters in the QRSS software as for reading noisy signal, plays a role.
For that reason random characters were generated. The software used was Spe=
cLab with the default settings for QRSS3 resp. QRSS10.
I could easily read QRSS3 as low as -24dB SNR and DFCW10 as low as -27dB SN=
R. But is is likely that more experienced ops do 1 or 2 dB better.
If there is some interest I am willing to put the screen captures on a webs=
ite for those who want give a try in decoding the QRSS / DFCW signals.

Conclusion: I was a bit surprised by the +/- 6dB difference between WSPR an=
d Opera2.
Both have about the same transission duration. But WSPR transmits more info=
rmation (call + 4 digit loactor + power) compared to Opera (only call). So =
that should be in the favour of Opera (more redundancy in the FEC).=20
But Opera seems to use some kind of Manchester coding, so 50% of the bits a=
re sacrified to synchronisation.
And finaly is WSPR a 100% duty cycle mode versus 50% for Opera, so the aver=
age power in 3dB lower for Opera. So that might be an explanation for the 6=
dB difference.=20

Based on the above Opera2 and QRSS3 seem to be competitive. DFCW10 outperfo=
rms them and WSPR is even a bit better than DFCW10.

But: these result are based on a relative low numer of measurement (5 at ea=
ch SNR level) and with artificial white noise. So it certainly is not writt=
en in stone.

73, Rik  ON7YD - OR7T
________________________________________
Van: owner-rsgb_lf_group@blacksheep.org [owner-rsgb_lf_group@blacksheep.org=
] namens James Moritz [james.moritz@btopenworld.com]
Verzonden: dinsdag 31 januari 2012 19:35
Aan: rsgb_lf_group@blacksheep.org
Onderwerp: LF: Re: Opera v qrs evaluation

Dear Eddie, LF Group,

In an attempt to actually make a comparison, I have set my RX to the 500kHz
WSPR frequency, with Spec Lab going also. Since it has been running, G3ZJO'=
s
QRSS3 has been about 10 - 15dB SNR in 0.3Hz FFT resolution, so say about 6d=
B
above the level needed for "good copy". Opera has so far reported SNR
between -23 and -27dB - I don't know how this relates to the detection
threshold for the Op4 mode. I'll leave it going for a while...

Cheers, Jim Moritz
73 de M0BMU=