Bill,
Thanks for the detailed explanation.
I was kind of surprised the other day after calling CQ on JT65 and receiving
two call backs on the exact same DF. I guess I know why now. Ed, K0KC
[email protected]http://k0kc.us/
From: Bill Somerville <[email protected]>
To: [email protected]
Sent: Saturday, January 2, 2016 10:34 AM
Subject: Re: [wsjt-devel] WSJT-X 1.7.0 r6333 Decoder Performance
On 02/01/2016 15:03, Ed Wilson wrote:
> I wonder how this performance translates to JT65 and JT9; I can see
> improvement, but nothing as good as -30 dB.
Hi Ed,
the performance of each decoder is discrete. I will try and summarize:
The WSPR-2 decoder in WSJT-X has several performance improvements over
the WSPR application. It was the first to use two pass decoding with
signal subtraction after the first pass, this is the main reason you can
receive multiple spots on the same DF. Steve K9AN authored these changes
with assistance from Joe K1JT for integration into WSJT-X.
JT65 decoding is undergoing a major rewrite and the current development
source is the result of that, it too uses 2 pass with intermediate
signal subtraction decoding. It also incorporates an advanced soft
decision decoding (basically very well informed guessing) for the
signals that cannot be decoded simply from the encoded FEC parity data.
This last part replaces KVASD and is currently able to achieve ~0.5dB
resolution improvements over KVASD. Steve originated that latest soft
decision algorithm and both Joe and Steve have fine tuned it to the
current state.
JT9 had a round of improvements for the v1.5 release but since then not
much has happened. The two signal subtraction mechanism could be applied
there too but the benefit would be smaller since JT9 signals are much
less likely to overlap until the bands reach higher occupancy.
Overall WSPR-2 is the most sensitive of the three on HF, this is due to
the shorter message and slower symbol rate which allows more averaging
to dig deeper into the noise. Occasional -30dB reports occur with JT65
and JT9 but these are rare and beyond the expected range. JT65 and JT4
in WSJT-X will eventually use experience based decoding which is
basically using existing known information like known active callsigns
to do more targeted guessing of decodes, this allows deeper searches for
successful decodes to be completed in a fraction of the time that a
exhaustive search would take. These experienced based methods allow JT65
to gain a few more dB resolution approaching -30dB.
73 & HNY
Bill
G4WJS.
------------------------------------------------------------------------------
_______________________________________________
wsjt-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/wsjt-devel
------------------------------------------------------------------------------
_______________________________________________
wsjt-devel mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/wsjt-devel
