I think Ekki's calculations are correct. The minimum bandwidth around each
signal (mark and space) is at least 1.6 times the baudrate to keep
distortion of the square pulse at an acceptable level, resulting in about
73Hz for a 45 baud signal. Total transmitted bandwidth comes to about
316Hz. I think we may be using the same source. My wisdom comes from a
German book, "Amateur-Funkfernschreibtechnik RTTY" by DJ6HP (published
1979).
"Brickwall" filters need to be about 320 Hz wide to pass the RTTY signal
without distorting it too much. Since some of the spectrum between mark and
space is not used, the ideal filter would not just be 316Hz wide but rather
consists of two individual filters that are about 75 to 80Hz wide, which is
rarely implemented at the IF level in an amateur rig.
When I first started with RTTY over 20 years ago, we used active filters
composed of cascaded op-amp circuits (741's worked well). Mark and space
had separate filters, each 80 Hz wide (nominal). I think most sound-card
based RTTY decoders emulate this type of filter, we just don't see this
detail usually. The radio's filters are merely another IF in the processing
chain, with more or less limited selectivity.
Having said all that, in my Icom 745 I used a 270 Hz filter on the 9MHz IF
and a 250Hz filter on the 455kHz IF. Even though they are both narrower
than the optimum 320Hz bandwidth, they worked ok. What probably happened is
that I lost some sensitivity as a tradeoff for the selectivity, but I never
really thought about it much.
In the IC-756 I used a 350Hz filter on 9MHz and a 250Hz filter on 455kHz.
Same thing - it worked just fine. The slopes of the filters are not steep
enough, so plenty of signal gets through. I could shut off the 250Hz filter
if I wanted to get some more bandwidth. However, since I use FSK
exclusively, I need to tune precisely anyway and the added bandwidth was
rarely needed. I moved the 350Hz filter to the mobile rig for CW and never
missed it for RTTY in the 756.
The IC-756PRO is a totally different matter, though. Here, if a filter says
it's 250Hz, that's all you get. No slopes (to speak of). The PRO's RTTY
filter is 350Hz wide, matching the theoretical 316Hz quite well. Icom
actually took it a step further and implemented a twin-peak filter. This is
as close to perfect as I've seen. The PRO's RTTY filter carves out mark and
space with their narrow bandwidth and suppresses the unused spectrum in
between. Note that the twin-peak filter of the PRO is only available in FSK
mode. If you use AFSK, you're limited to the standard filters, which should
be set to about 320 to 350 Hz wide. If you go any narrower than 320Hz, you
start limiting the received signal which will cause increased distortion
when the decoder attempts to recreate the original square wave form.
Luckily, I much prefer FSK (for a variety of reasons I don't want to
elaborate on right now), so I get the benefit of the near-ideal IF-level
twin peak filters. I think they represent the optimum selectivity based on
the composition of the RTTY signal.
The bottom line is that you are receiving a signal that's about 320 Hz
wide, so the filters should match this bandwidth unless they can select
mark and space separately.
73,
Michael, K1JE
|