Message-Id: <3.0.6.32.20041127113429.00c39b50@magma.ca>
Date: Sat, 27 Nov 2004 11:34:29 -0500
To: rsgb_lf_group@blacksheep.org
From: Bill de Carle <ve2iq@magma.ca>
In-Reply-To: <41A89AAD.2080804@usa.net>
References: <000b01c4d3fc$f0716880$6507a8c0@Main>
 <000b01c4d3fc$f0716880$6507a8c0@Main>
Mime-Version: 1.0
Subject: Re: LF: Re: MSK etc and stability
Content-Type: text/plain; charset=us-ascii; format=flowed
Sender: owner-rsgb_lf_group@blacksheep.org
Precedence: bulk
Reply-To: rsgb_lf_group@blacksheep.org
Content-transfer-encoding: 8bit

At 04:18 PM 11/27/2004 +0100, Alberto wrote:
>[..]  I was thinking of a scheme more or less like the 
>following, for the
>137 kHz scenario, for demodulating a BPSK transmission.
[..]
>A PC program can now mix down, with a tunable NCO, that range to zero IF 
>and apply a complex FIR lowpass with a BW of just a few Hz. What do you
>have now ? A rotating vector (ideally it should not rotate, if the zero
>IF is actually zero), whose phase is instantaneously defined by
>atan(Q/I). If you do a long term averaging of the speed of variation of
>this phase, you can then more or less easily measure if the instantaneous
>phase is leading or lagging this averaged value.
>This is just the recovered 0 or 1 bit of  the phase modulation applied 
>to your carrier.
[..]
>As said, just a wild idea. If there are valid reasons to trash it, 
>please speak up, so I won't
>waste any more time thinking at it.   Thanks.

This is basically how the "PSKL" mode in COHERENT/AFRICA  was done.
Everything works nicely when the SNR is good, but one problem is that
it is very difficult to accurately measure phase when the signal is
extremely weak.  I found that in place of long term averaging it is better
to fit a linear regression line (assuming the speed of the variation is
constant over the averaging period) - the slope is what we want and it
does a better job of taking out the effects of jitter - but still when
the signal is way down in noise, even the best guess can be off by
a mile.  You can't afford to assume the parameter stays constant for a
long time because the filter wouldn't track.  And for a sufficiently
short measurement time to be useful in BPSK decoding you generally can't
measure accurately enough to be more useful than just going to differential
encoding then comparing two consecutive measured phases to see if they
differ by more than +/- 90 degrees.

Bill
VE2IQ