At 05:52 PM 9/1/2010, Jim Moritz wrote:
Dear Bill, LF Group,
----- Original Message ----- From: "Bill de Carle" <[email protected]>
I tried turning off the AGC for a while but had to manually cut
the RF gain so much to avoid clipping it didn't seem like a good
idea. With the AGC on at least there is some possibility of
weak-signal energy getting through between crashes if there are
QRN-lulls here and there.
I think you may find that you get better results under these
conditions if you crank up the RF gain until clipping occurs all the
time. This results in some loss of signal level, but a greater
reduction in QRN level on the spectrogram, leading to several dB SNR
improvement overall. I suspect how well this works depends on how
quickly the RX gain stages recover from a short-term overload; if
the amplifiers clip in a well-behaved way, then the QRN transients
during which the signal is lost will have a duration determined by
the impulse response of the IF filter. But the overdriving of the
amplifiers that occurs during the transient might upset the DC
biasing, leading to a long recovery "tail" due to the time constants
of coupling and decoupling Rs, Cs, etc., and greater loss of signal
energy. Operating the receiver with the IF amplifier in this kind
of limiting mode was quite successful with the old RA1792 receiver,
which has gain-block IC IF amplifiers which probably work quite well
as limiting amplifiers.
This kind of clipping does not work well if there are strong
unwanted signals present in the bandwidth being clipped - either the
clipping threshold has to be increased to above the level of the
unwanted signal, making it less effective at reducing the QRN, or
the strong signal "captures" the bandwidth as in an FM receiver
limiting IF, reducing the level of the wanted signal as well as the
QRN. so in Europe the bandwidth around 136kHz has to be restricted
to 1kHz or so to exclude DCF39 et al.
Thanks for your comments Jim. I tried turning off the AGC with RF
gain at max and some extra gain in front of the Rx to make it clip
sooner. Looking at the results of an overnight (Wednesday) session,
I saw no trace of TIL (worse than the performance on the previous
overnight runs) but I did see other weak signals. Of course one
overnight run shouldn't be taken as definitive, especially when
propagation on the path from TIL seems to have been steadily
worsening over the last few nights. Could be it was just a bad night.
One idea I am kicking around is to pass the 50-ohm signal at the
antenna connector through the centre of a toroid with a lot of turns
on the secondary - enough that we see millivolts instead of
microvolts. Then I would do the clipping on the hi-Z side using a
couple of fast op-amps with a variable threshold setting, hoping the
hi-Z clipping gets reflected back thru the toroid to the 50 ohm
side. Not good for HF but at 137 Khz it might work well
enough. Then I could tweak the clipping threshold while watching the
screen to find the best setting. On the other hand, recording first
then playing back a QRSS-120 run repeatedly while changing a software
setting to find the best clipping point would be quite time-consuming.
Since we're trying to adjust display parameters to make it easier for
the human eye/brain to discern a weak signal against a noisy
background, it might be best to maintain linearity as much as
possible then do the clipping after the FFT across all the individual
frequency component amplitudes. I don't know if this would be any
more effective than adjusting the contrast and brightness controls as
we do now. The eye-brain is already pretty good at ignoring all the
background clutter and "seeing" a pattern it is expecting. And the
ear-brain is amazing in its ability to "hear" really weak CW signals
in a band full of louder competing signals. The signals we're
looking for are generally weaker than the noise. I wonder if it
would be helpful to display brightness proportional to the logarithm
of amplitude? I tried that years ago to no great benefit, but the
on-the-fly contrast/brightness sliders we have today to change the
whole screen in real-time might make it worth another try.
73,
Bill VE2IQ
|
