Jim,
Yes good points , note the narrow front end ..looks good on a plot
..no use for s/n LC in front of the mixer = more noise , same argument
over tuned/un-tuned rx antennas ... look at these min-whips on mf
''with A/D conversion immediately after a "roofing" filter for data
signals'',
I would think, these days , the cost of a good roofing filter compared
to a A/D convertor , pushes this argument one step on to eliminating
the filter as well , taking a few pointers from the 'world of radar' the
sooner you can digitise the better ,
AGC , off and on are not quite the full story, attack and hang time
serve to produce IMD by altering the transmitted pulse shape ,
switching from short to long on the ra1778 introduces app 10 dB more
IMD on a psk31 signal by tracking the envelope, the some what
'mechanical' sample and hold AGC in the ra6790 makes little difference
For my $ the sound card is an area most overlooked , SURE 444 V XTAL
mic syndrome and speech clipping , you 'have' to have good audio
before you (rf) clip .. but how many assume because its clipped than
the start quality is not important ?
G ..
hobby .?. more like being at work !
--------------------------------------------------
From: "James Moritz" <james.moritz@btopenworld.com>
Sent: Sunday, November 07, 2010 11:41 AM
To: <rsgb_lf_group@blacksheep.org>
Subject: Re: LF: WSPR T/A hole discussion
Dear John, Roelof, LF Group,
I think this problem and the discussion just reflects the fact that
receivers are not really designed for use with digital modes, particularly
ones that are very narrow-band compared to the traditional phone/CW modes.
There does not seem to be a fundamental reason why WSPR signals of greatly
differing strength should not be successfully simultaneously decoded,
provided the transmitted signals are "clean", and the receive stages up to
and including the DSP processing have good linearity and low phase noise,
etc.
There has been an obsession in the amateur world for receivers with
enormous dynamic range, but this is only normally applied to the RX front
end, up to the point where the final IF bandwidth is achieved - the
designers assume that the IF bandwidth is similar to the wanted signal
bandwidth, and filtering represents all that can be done to remove
unwanted signals. It is assumed that all unwanted signals that it is
possible to remove have been eliminated after this stage. The post-filter
stages only need to be linear enough not to significantly impair the
wanted signal, and there has to be a trade-off with other inherently
non-linear functions too, such as AGC.
But for many modern "digital" modes, this is no longer the case -
normally, several signals will be present in the RX IF and audio channel
simultaneously. In the case where spread-spectrum or CDMA techniques are
used, multiple signals inherently use the same bandwidth. Linearity must
be maintained throughout the receiver. So RX design lags behind the
requirements for modern transmission modes. I think in the long term this
means seperating the audible reception channel from the data reception
channel, with A/D conversion immediately after a "roofing" filter for data
signals, and AGC, etc. confined to the "analogue" channel. For current
receivers this can't easily be done, and the best practical thing to do in
my experience seems to be to ensure the AGC has been disabled, and that
the RF/IF gain is operated at a much lower level than would normally be
used for audible reception. This applies to QRSS, etc as well as WSPR. I
have found that with careful gain adjustment, a lot of the problems due to
strong local signals can be eliminated.
In the case of WSPR, the other practical thing that can be done is to
reduce the TX duty cycle - so if a 20% duty cycle is used, 80% of time
slots will be unnaffected by a local ground-wave-blasting station; if
there are two such stations, 64% of slots remain unaffected, and even for
three, 51% are OK.
Cheers, Jim Moritz
73 de M0BMU
|
