Message-ID: <5593AF2E.3030607@freenet.de>
Date: Wed, 01 Jul 2015 11:13:18 +0200
From: wolf_dl4yhf <dl4yhf@freenet.de>
User-Agent: Mozilla/5.0 (Windows NT 5.1; rv:17.0) Gecko/20130620 Thunderbird/17.0.7
MIME-Version: 1.0
To: rsgb_lf_group@blacksheep.org
References: <5591B4B6.6080302@posteo.de> <5591C9E4.4070407@psk31.plus.com> <5591CECF.6000709@posteo.de> <55929BC9.6020804@psk31.plus.com> <5592A3A3.5030805@posteo.de> <55932804.5030800@posteo.de>
In-Reply-To: <55932804.5030800@posteo.de>
Subject: Re: LF: SpecLab's frequence selective limiter: Better performance
 for your WSPR decoder!
Content-Type: multipart/alternative;
 boundary="------------040408080505000604030803"
Sender: owner-rsgb_lf_group@blacksheep.org
Precedence: bulk
Reply-To: rsgb_lf_group@blacksheep.org
Status: O

This is a multi-part message in MIME format.
--------------040408080505000604030803
Content-Type: text/plain; charset=ISO-8859-15; format=flowed
Content-Transfer-Encoding: 8bit

Hi Stefan,

Thanks for the interesting test. I seem to have little luck with the 
dropbox links (they almost always result in some sort of file-not-found 
error) but I can imagine the combination. Maybe there's another link to 
your dropbox storage, or maybe the space characters in the filename make 
it unaccessable ?

I'm curious about what actually causes the spectral spreading of the 
strong carrier (caused by the normal noiseblanker), and why the 
frequency-selective limiter can achieve a steeper roll-off (over 
frequency) than the carefully shaped noiseblanker response. Maybe the 
'window function' of the noiseblanker itself can be improved (by 
something more elaborate than the raised-cosine-shaped edges, and the 
flat 'window ledge' in between) ?

All the best,
  Wolf DL4YHF
   (QRN season and extraordinary temperatures up to 40 °C on their way 
... yucc...)


Am 01.07.2015 01:36, schrieb DK7FC:
> Hi all,
>
> My MF RX-experiments are continuing and i discovered SpecLab's 
> frequency selective limiter:
>
> Today it is a quite noisy night.
>
> As reported previously i'm running two SpecLab instances from the same 
> signal source (VAC1) feeding two WSPR instances, so i can directly see 
> the effect of the NB or other techniques.
> This image shows the experiment: 
> https://dl.dropboxusercontent.com/u/19882028/MF/Frequence%20selective%20limiter.png
> The upper WSPR+SpecLab instance is using the NB, the lower one just 
> makes the necessary frequency conversion and SSB filter setting (474.2 
> kHz "dial").
>
> Using SpecLab's noise blanker (NB) makes an improvement of 3...7 dB 
> tonite, as reported by WSPR-2, so it is worth to run it! Actually 
> there were many decodes by that instance which is using the NB that 
> didn't decode on the other one, for example G3WCB at 22:32 UTC (see 
> text fields). Without the NB there was no trace at all!
>
> However, when running the NB, strong signals will cause the NB the 
> rise the noise in the spectrum. DK2DB is a strong signal here, showing 
> 40 dB above the noise (in 1 Hz), (yes i'm such a strong signal too, i 
> know!!!). You can see the effect on the upper spectrogram at 21:50 or 
> 22:00 UTC. A small signal like G3WCB would be lost in that background 
> noise which is caused by the NB. However without the NB it wouldn't 
> decode as well!
>
> So what is the solution? Maybe something like an auto notch filter but 
> if this is very sharp, it also causes a higher noise arround. Actually 
> it must be something which just lowers the strong signals a bit, maybe 
> 25 dB but leaves the rest as it is. In a phone call with Markus we 
> discussed about this and discovered the the _*frequence selective 
> limiter*_ and it's performance. It can be found in the options 
> register card of the filter control window and must be enabled there. 
> I've the configured a band pass filter including this limiter. _After_ 
> that filter and limiter, the NB is running in the blackbox between L4 
> and L5.
>
> Now, when enabling the filter and defining a limit just about 10 dB 
> above the average noise floor, the signal of DK2DB is reduced to a 
> level which does not cause the NB to rise the noise floor! You can see 
> it at 22:08 UTC. In the list you can see that the displayed SNR 
> dropped from +3 to -15 dB at 22:16 UTC where is reduced the limiter 
> threshold a bit more. *So decodes of the strong local stations do 
> still happen while the NB is not acting on the strong signal, and QRN 
> is significantly blanked, allowing to decode weak signals!* For 
> example, see G3XIZ at 22:44 UTC, showing -17 dB instead of -25 dB, 
> i.e. an improvement of 8 dB! DK2DB has been active to that time.
>
> 73, Stefan
>


--------------040408080505000604030803
Content-Type: text/html; charset=ISO-8859-15
Content-Transfer-Encoding: 8bit

<html>
  <head>
    <meta content="text/html; charset=ISO-8859-15"
      http-equiv="Content-Type">
  </head>
  <body text="#000000" bgcolor="#FFFFFF">
    <div class="moz-cite-prefix">Hi Stefan,<br>
      <br>
      Thanks for the interesting test. I seem to have little luck with
      the dropbox links (they almost always result in some sort of
      file-not-found error) but I can imagine the combination. Maybe
      there's another link to your dropbox storage, or maybe the space
      characters in the filename make it unaccessable ?<br>
      <br>
      I'm curious about what actually causes the spectral spreading of
      the strong carrier (caused by the normal noiseblanker), and why
      the frequency-selective limiter can achieve a steeper roll-off
      (over frequency) than the carefully shaped noiseblanker response.
      Maybe the 'window function' of the noiseblanker itself can be
      improved (by something more elaborate than the
      raised-cosine-shaped edges, and the flat 'window ledge' in
      between) ? <br>
      <br>
      All the best,<br>
       Wolf DL4YHF<br>
        (QRN season and extraordinary temperatures up to 40 °C on their
      way ... yucc...)<br>
      <br>
      <br>
      Am 01.07.2015 01:36, schrieb DK7FC:<br>
    </div>
    <blockquote cite="mid:55932804.5030800@posteo.de" type="cite">
      <meta content="text/html; charset=ISO-8859-15"
        http-equiv="Content-Type">
      Hi all, <br>
      <br>
      My MF RX-experiments are continuing and i discovered SpecLab's
      frequency selective limiter: <br>
      <br>
      Today it is a quite noisy night.<br>
      <br>
      As reported previously i'm running two SpecLab instances from the
      same
      signal source (VAC1) feeding two WSPR instances, so i can directly
      see
      the effect of the NB or other techniques.<br>
      This image shows the experiment:
      <a moz-do-not-send="true" class="moz-txt-link-freetext"
href="https://dl.dropboxusercontent.com/u/19882028/MF/Frequence%20selective%20limiter.png">https://dl.dropboxusercontent.com/u/19882028/MF/Frequence%20selective%20limiter.png</a><br>
      The upper WSPR+SpecLab instance is using the NB, the lower one
      just
      makes the necessary frequency conversion and SSB filter setting
      (474.2
      kHz "dial").<br>
      <br>
      Using SpecLab's noise blanker (NB) makes an improvement of 3...7
      dB
      tonite, as reported by WSPR-2, so it is worth to run it! Actually
      there
      were many decodes by that instance which is using the NB that
      didn't
      decode on the other one, for example G3WCB at 22:32 UTC (see text
      fields). Without the NB there was no trace at all!<br>
      <br>
      However, when running the NB, strong signals will cause the NB the
      rise
      the noise in the spectrum. DK2DB is a strong signal here, showing
      40 dB
      above the noise (in 1 Hz), (yes i'm such a strong signal too, i
      know!!!). You can see the effect on the upper spectrogram at 21:50
      or
      22:00 UTC. A small signal like G3WCB would be lost in that
      background
      noise which is caused by the NB. However without the NB it
      wouldn't
      decode as well! <br>
      <br>
      So what is the solution? Maybe something like an auto notch filter
      but
      if this is very sharp, it also causes a higher noise arround.
      Actually
      it must be something which just lowers the strong signals a bit,
      maybe
      25 dB but leaves the rest as it is. In a phone call with Markus we
      discussed about this and discovered the the <u><b>frequence
          selective
          limiter</b></u> and it's performance. It can be found in the
      options
      register card of the filter control window and must be enabled
      there.
      I've the configured a band pass filter including this limiter. <u>After</u>
      that filter and limiter, the NB is running in the blackbox between
      L4
      and L5.<br>
      <br>
      Now, when enabling the filter and defining a limit just about 10
      dB
      above the average noise floor, the signal of DK2DB is reduced to a
      level which does not cause the NB to rise the noise floor! You can
      see
      it at 22:08 UTC. In the list you can see that the displayed SNR
      dropped
      from +3 to -15 dB at 22:16 UTC where is reduced the limiter
      threshold a
      bit more. <b>So decodes of the strong local stations do still
        happen
        while the NB is not acting on the strong signal, and QRN is
        significantly blanked, allowing to decode weak signals!</b> For
      example, see G3XIZ at 22:44 UTC, showing -17 dB instead of -25 dB,
      i.e.
      an improvement of 8 dB! DK2DB has been active to that time.<br>
      <br>
      73, Stefan<br>
      <br>
    </blockquote>
    <br>
  </body>
</html>

--------------040408080505000604030803--