Message-ID: <5093ED35.9020905@princeton.edu>
Date: Fri, 02 Nov 2012 11:56:37 -0400
From: Joe Taylor <joe@Princeton.EDU>
User-Agent: Mozilla/5.0 (Windows NT 6.1; rv:8.0) Gecko/20111105 Thunderbird/8.0
MIME-Version: 1.0
To: rsgb_lf_group@blacksheep.org, rsgb_lf_group@yahoogroups.co.uk
References: <508D92BE.2040500@broadpark.no> <508ED047.1000504@broadpark.no> <CAHAQVWPFQdJQxMtkvzOvGZ-SbzGk3WTptmvMqDzBqpECjzjpgQ@mail.gmail.com> <508EEDA9.7070905@princeton.edu> <50901CB0.2040405@princeton.edu> <50917D09.2030908@princeton.edu> <509299E8.7010608@princeton.edu> <5092D947.1000801@princeton.edu>
In-Reply-To: <5092D947.1000801@princeton.edu>
Subject: LF: Soundcard Levels
Content-Type: text/plain; charset=ISO-8859-1; format=flowed
Content-Transfer-Encoding: 7bit
Sender: owner-rsgb_lf_group@blacksheep.org
Precedence: bulk
Reply-To: rsgb_lf_group@blacksheep.org


Hi all Lowfers,

1. I'm happy to report that AC power came on at my home late yesterday. 
  Things here are slowly returning to normal -- but they still have a 
long way to go.  I know that many in this region are not so fortunate as 
we have been.

2. Unless I hear of some serious problem, another release of WSJT-X is 
unlikely for about 10 days.  My available "radio time" this weekend will 
be devoted to the ARRL International EME contest, and next week I am 
busy with travel and QRL activities.

3. You may be interested in a full explanation of the audio signal level 
issues we have been discussing.  Some have suggested that with "normal" 
audio levels from receiving equipment, the indicated WSJT-X input levels 
are extremely high -- perhaps close to the full-scale "60 dB" mark on 
the WSJT-X green-bar level meter.

The 16-bit digital samples from a soundcard fall in the range -32768 to 
+32767.  The reference level for the WSJT-X level meter is 1 bit, so
"full scale" or 60 dB means an rms level of 10^(0.05*60) = 10^3 = 1000. 
  This may be a "normal" level for the receiver and audio settings you 
have used for other purposes.  That fact might lead you to think that 
the WSJT-X level meter reads "ridiculously high".  In fact, though, 
setting the WSJT-X leval meter to a much lower level of 20-30 dB (with 
the recently added slider set to mid-scale) is still a good idea, and 
here's why.

If the background noise level is set so that the rms output from the A/D 
converter is 1000, you have only about 30 dB of headroom for strong 
signals.  That's not much dynamic range.  Remember, we want to operate 
with AGC off so that WSJT-X can find and work weak signals even when 
your neighbor may be active the band.

A much better setup is to have the rms output from the A/D converter 
more like 10-30 counts, or about 20-30 dB above 1 count.  Yes, this 
means that only a few bits of the A/D converter's range are being used 
much of the time, when you are receiving weak signals plus background 
noise.  But almost nothing has been lost: A/D quantizing noise is still 
negligible with respect to external background noise.  Moreover, all 
those higher-order bits are now available for the headroom needed to 
give you 60-70 dB of dynamic range, rather than just 30 dB.

So: use the digital gain slider, if you must, to make the green bar 
shorter.  But recognize that if you choose to run with your 16-bit A/D 
samples around 1000 or so, rms, you will have sacrificed 30-40 dB of 
available dynamic range.

4. Another matter of interest: several *.wav files sent to me, and 
several screen shots recently posted on the web, indicate that some 
WSJT-X users are running with rather narrow filters (maybe 500 Hz or so) 
in the Rx chain, ahead of the soundcard.  In general this is not a good 
idea.

Unless you need to get rid of some extremely strong QRM -- so strong 
that WSJT-X would not be able to handle it -- wider bandwidths are 
better, up to about 5 kHz.  The normal bandwidth 2.5-3 kHz bandwidth of 
an SSB transceiver is fine, and almost certainly is preferable to a 500 
Hz CW filter.

Why?  Because WSJT-X already does all necessary filtering, down to the 
fraction-of-one-Hz tone spacings used by the various JT9 submodes.  A 
wide bandwidth coming into WSJT-X gives the software a *much* better 
opportunity to do effective noise blanking.  The noise blanker will be 
highly effective, even if strong signals are present.  (Note that I say 
"will be".  I have not yet optimized behavior of the NB in WSJT-X.  This 
is still to be done.)

When running in any of the JT9 submodes, the background noise level 
shown on the WSJT-X waterfall should be nearly constant over the full 
displayed frequency range.  You should *not* see the peaked effect of 
your IF filter(s) on the waterfall!

5. A brief comment for those experiencing sporadic decodes.  With 
steady, non-drifting signals that are time-aligned to within +/-2 s, 
WSJT-X should decode JT9 signals very reliably down to levels within 1 
or 2 dB of the thresholds listed in User's Guide table.  If you are not 
experiencing this level of reliability, there is surely a reason -- and 
we'll help you find what that reason may be.

6. As already stated in the brief User's Guide: "A number of reported 
parameters have not been properly calibrated."  In particular, you 
should take the displayed values of S/N as relative indicators, with 
significant uncertainties -- especially if your passband is not very 
flat, or if there are strong signals present, or ...  I will turn some 
attention to getting these things right, in due course.

7. Finally -- when I can again devote some time to WSJT-X, probably 
starting in the week of November 12, I will be QRV myself.  I'm ready to 
put up a reasonably good Rx antenna for MF and LF, so I should be able 
to listen for transatlantic signals.  I am QRV on 160 m already, and 
have made a number of JT9 QSOs there.  (I note that a number of others 
are playing with JT9 modes on 160 m and HF, even up to 50 MHz.)

That's it for now!

	-- 73, Joe, K1JT