Bill W7TI wrote:
> The real villain here is excessive receiver bandwidth.
If there is a lack of very large dynamic range of signals
within a passband, the filtering in software is actually
superior to that of the filters you can buy for today's
rigs.
But!
The problem is when the dynamic range exceeds that of the
sound card's capability, DSP filtering becomes mush, and
that is when you need to narrow down the bandwidth of
the RF/IF stages.
One of the reasons we can do so well with something like
PSK31 is that there are tight filters in the software. In
the PSK31 implementation which I wrote for the Mac for my
own use, I had used very long FIR filters, with floating
point math.
That, together with PSK being superior to FSK to start with.
Remember that with digital modes, the BER, (bit error rate
or, how clean the "print" is) is not a linear function of
S/N ratio. There is a very steep threshold where you go
from good BER to unusable BER. Right at the threshold,
bumping your power by 3 dB can make the difference between
good copy and guessing every other character.
There is a huge advantage of PSK over FSK as to where this
threshold is.
Traditional FSK goes from good copy (1 error in hundreds of
bits) at a S/N ratio of 10 dB or so, to crummy copy (1 error
in 10 bits, or just about every character has an error) at
S/N ratio of 5 dB. Whether you can still guess that the DX
has answered your call is question at this S/N ratio :-).
Binary PSK goes from the same good copy at a S/N ratio of
6 dB to crummy copy at a S/N ratio of 0 dB.
Thus, for practical purposes, you can get good print on
PSK when incoherent FSK is completely garbled, even if the
bandwidths are the same.
However, the bandwidths of PSK31 and RTTY are not the same.
The difference in performance is further amplified by the
fact that a PSK31 signal is much narrower than an RTTY FSK
signal (i.e., there is probably a difference of some 8 dB
in S/N ratio between optimally filtered PSK31 and steam
RTTY).
From the above, there is probably a 13 dB or so difference
in signal power needed between an optimal RTTY setup and an
optimal PSK31 setup. A QSO at 10 watts using PSK31 is going
to take 200 watts using RTTY. (Or, equivalently, typical
barefoot rig vs a 2 kW RTTY monster station).
[Someone check this, I am doing it from the "back of an
envelope." :-) It has been a long time since I last looked
at the spectrum, and thus bandwidth, of a PSK31 signal.]
BTW, unless you are ZD8Z, you can't easily get an extra
13 dB of directivity out of an antenna system, either.
One can see why PSK31 is a darn good solution for a modest
ham, like moi.
But all that DSP filtering and demodulation isn't going
to be very effective if the signal given to the "sound
card" has a larger dynamic range than the sound card
can handle.
Hence the chorus here of "narrow the receiver bandwidth!"
73
Chen, W7AY
|