[Lowfer] Sampling Rate and Frequency Calibration

[JD]

JA,

I use a method similar to Jay H's with Argo, and it works well...except I 
don't try to compensate for receiver calibration, only sound card error, 
which is a lot smaller factor at any usual QRSS speed.

Being out in the wilderness where ambient temperature may vary more than 20 
degrees F during a session is a special case.  I'd be forever adjusting 
Argo's offset.  Instead, immediately before and after a run on a given 
frequency, I do a short capture of WWV's 10 MHz carrier in a QRSS3 Argo 
window that I always keep open.  It's easy to establish the receiver error 
in PPM that way, since all of my receiver's internal frequency determination 
is from one oscillator.  At the end of an actual capture, I log the dial 
frequency of the radio and the PPM error, then compute the error in Hz, and 
graphically shift the scale by the appropriate Hz or fraction thereof on my 
"trophy" copy of the capture.  Sounds complicated at first, but it's easy to 
get the hang of, and it gives good results.  (Were I in a temperature 
controlled environment or had a tighter master oscillator in the radio, I 
would simply do what Jay H does.  And I'd play with WOLF more!)

I have noticed that while Argo's frequency requirements for visual modes are 
a lot less demanding than, say, WOLF, it does require at least a little 
knowledge of signal timing to achieve maximum benefit.  The width of the FFT 
bins and the interval over which their energy levels are integrated does 
constitute a form of matched filter.  The response "peaks" in a manner of 
speaking, only when the elements being received are a multiple of the 
specified dot length.  That means both the on _and_ off times.  When a 
station weights the spacing to shorten off-intervals in an effort to save 
transmission time, then that part of the signal exceeds the turnover 
frequency of the matched filter and smearing occurs.  You then either have 
to depend on dogbones to judge the timing of on-intervals, or else drop back 
to a shorter dot length.  Either way works if there's plenty of signal, but 
weak signals will suffer needlessly.  A similar case can be made for 
frequency stability, since the drift spreads energy out over multiple bins. 
A frequency shift during keying may be distinctive when strong enough to be 
seen anyway, and in that sense it is more forgiving; but it is less 
efficient from Argo's viewpoint when the signal is weak.  Just a couple 
thoughts to keep in mind when going for distance with QRSS.

JD

----- W1TAG wrote -----

Jay H,

Upon further thought, you may still get perfection! Matching the
sampling rate is critical for modes where you are trying very hard to
stay in sync with what is being sent. WOLF, with it's ability to build
up copy, is the ultimate example. But an FFT program like Argo isn't
doing that. It is just displaying frequency and amplitude of whatever is
coming in, with no knowledge needed of the timing of any signal.

I'm guessing that your method would work just fine in this case. Anyone
have any thoughts about it?

John, W1TAG