To: "Discussion of the Lowfer (US, European, & UK) and MedFer bands"
 <lowfer@mailman.qth.net>,
        600 / 630 Meter Group <600mrg@w7ekb.com>,
        "rsgb_lf_group@blacksheep.org" <rsgb_lf_group@blacksheep.org>,
        "rsgb_lf_group@yahoogroups.co.uk" <rsgb_lf_group@yahoogroups.co.uk>
From: N1BUG <paul@n1bug.com>
Message-ID: <22427f06-486d-f7fd-965f-70ed200d2b8d@n1bug.com>
Date: Mon, 26 Feb 2018 07:57:51 -0500
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:52.0) Gecko/20100101
 Thunderbird/52.6.0
MIME-Version: 1.0
Content-Language: en-US
Subject: LF: WSJT-X 1.9 vs 1.8 WSPR decoding test
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Transfer-Encoding: 7bit
Sender: owner-rsgb_lf_group@blacksheep.org
Precedence: bulk
Reply-To: rsgb_lf_group@blacksheep.org

Yesterday's release of WSJT-X 1.9.0rc2 came with this note: 
"Improved decoding performance for WSPR mode, especially effective 
at LF and MF". I wanted to put this to the test so overnight I ran 
four instances of WSJT-X:

1.8.0 on both LF and MF using the call sign N1BUG/1

1.9.0rc2 on both LF and MF using the call sign N1BUG

All spots were uploaded to WSPRnet.

Both decoders on LF were fed the same audio stream. The setup is a 
simple SDR providing I/Q input to a physical sound card, HDSDR 
software output into a virtual audio cable, both versions of WSJT-X 
taking input from the output of that virtual cable.

The MF setup was the same except using a different SDR feeding a 
different physical sound card.

I used identical settings in all four instances of WSJT-X with the 
exception of different input audio source (virtual cable) for LF and MF.

All of this was running in Windows 10.

I watched incoming spots very closely for the first several hours. 
There was not much activity on LF at the time but on MF I saw 1.9 
decode many WSPR transmissions that 1.8 failed to decode. Some of 
these were extreme weak signal down to -32 with barely visible 
traces on the waterfall. Others were not with some clearly visible 
and decoding up to -23 in 1.9, yet no decode in 1.8 despite being 
very clear on the waterfall in that version. I did not see a single 
instance where 1.8 decoded something that 1.9 failed to decode.

This morning I took a quick look at statistics:

MF - During a 12 hour period ending 1145z, 1.9 decoded a total of 
933 WSPR transmissions while 1.8 decoded only 883.

LF - During a 12 hour period ending 1150z, 1.9 decoded 253 WSPR 
transmissions while 1.8 decoded only 183.

In all of this I do not see any obvious spurious decodes from either 
version. No strange call signs or stations displaced on the map from 
where you would expect them to be.

I was not expecting to see such a large difference. I make no claim 
that this result is representative of what others will see. I am 
simply reporting the results of an experiment carried out here.

73,
Paul N1BUG