Hi Mike,
Thanks for the comments.
> On Jun 28, 2015, at 10:10 AM, Michael Black <mdblac...@gmail.com> wrote:
>
> I'm curious...can you test what happens if you make it a one-sided test? That
> would eliminate local minimums.
> if((smspec[j]>smspec[j-1]) || (smspec[j]>smspec[j+1]) &&
> (npk<200)) {
My current thinking is that spectral shape (peaks and mins) should be ignored
altogether, because almost anything can happen when two or more signals with
carrier frequencies within a few Hz of one another are superposed.
>
> Also...what's the distribution of the snr values for the successful vs
> unsuccessful decodes? Maybe there's a better value to be had there.
The snr_min that I’m using corresponds to -33 dB in 2.5 kHz bandwidth, or -7 dB
in the bandwidth occupied by the WSPR signal. We know from much experience
(both with real signals and with simulations) that we are not going to decode
anything weaker than this.
>
> 73
> Mike W9MDB
Steve k9an
>
> -----Original Message-----
> From: Steven Franke [mailto:s.j.fra...@icloud.com]
> Sent: Sunday, June 28, 2015 9:47 AM
> To: WSJT software development
> Subject: Re: [wsjt-devel] wspr two-pass decoder status and .c2 file problem
>
> Joe,
> Frequency-sorting has now been committed.
>
> It dawned on me that it is no longer really appropriate to insist that a
> decoding candidate correspond to a local maximum in the average spectrum. I
> tried this simple change:
>
> $ diff wsprd_exp.c ~/Builds/wsjtx/lib/wsprd/wsprd_exp.c
> 848,850c848,849
> < for(j=1; j<410; j=j+2) {
> < // if((smspec[j]>smspec[j-1]) && (smspec[j]>smspec[j+1]) &&
> (npk<200)) {
> < if((smspec[j]>min_snr) && (npk<200)) {
> ---
>> for(j=1; j<410; j++) {
>> if((smspec[j]>smspec[j-1]) && (smspec[j]>smspec[j+1]) &&
>> (npk<200)) {
>
> On my 20m wav files it gives us another 10% over the current wsprd_exp with
> subtraction and two-passes, or a whopping 47% improvement over the current
> default wsprd:
>
> Default wsprd: 1315 in 163s
> wsprd_exp: 1758 in 270s
> no peak requirement: 1934 in 1487s
>
> With this change, a case that yields 16 decodes from the current wsprd_exp
> now yields 19 decodes from one .wav file!
>
> It is dog-slow as it stands. But I’m sure that there are things that we can
> do to speed it up significantly. This might be a good application for
> openmpi, no?
>
> Steve k9an
> ------------------------------------------------------------------------------
> Monitor 25 network devices or servers for free with OpManager!
> OpManager is web-based network management software that monitors network
> devices and physical & virtual servers, alerts via email & sms for fault.
> Monitor 25 devices for free with no restriction. Download now
> http://ad.doubleclick.net/ddm/clk/292181274;119417398;o
> _______________________________________________
> wsjt-devel mailing list
> wsjt-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/wsjt-devel
>
>
> ------------------------------------------------------------------------------
> Monitor 25 network devices or servers for free with OpManager!
> OpManager is web-based network management software that monitors
> network devices and physical & virtual servers, alerts via email & sms
> for fault. Monitor 25 devices for free with no restriction. Download now
> http://ad.doubleclick.net/ddm/clk/292181274;119417398;o
> _______________________________________________
> wsjt-devel mailing list
> wsjt-devel@lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/wsjt-devel
------------------------------------------------------------------------------
Monitor 25 network devices or servers for free with OpManager!
OpManager is web-based network management software that monitors
network devices and physical & virtual servers, alerts via email & sms
for fault. Monitor 25 devices for free with no restriction. Download now
http://ad.doubleclick.net/ddm/clk/292181274;119417398;o
_______________________________________________
wsjt-devel mailing list
wsjt-devel@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/wsjt-devel
