[RTTY] Crystal filter width preferences for RTTY contesting

[Kok Chen]

On Aug 26, 2009, at 11:08 PM, Jeff Blaine AC0C wrote:

> I think this helps
> to explain why there is quite a bit of satisfaction with the K3  
> using this
> filter on RTTY.

Jeff, bear in mind when you select filters the following two points:

1) optimal receive filtering has nothing to do with FCC 47 Part 2.202  
definition of occupied bandwidth (which is over 400 Hz in the case of  
steam RTTY).
2) keying sidebands from nearby stations can never be removed with a  
receiving filter (comments below).

To see what keying sidebands of an RTTY signals are like, you can take  
a look at (you can see that this topic rears its head every couple of  
years :-):

http://homepage.mac.com/chen/Technical/FSK/sidebands.html

In the above, I had looked at FSK (where you switch between mark and  
phase with no regards of any continuity), phase continuous FSK (where  
the transition between mark and space are continuous, but is not slope  
continuous -- this is why FSK (vs AFSK) sidebands are wider than  
neccessary), an AFSK signal that is well filtered and still provides  
the receiving matched filter with something that is still pretty  
decent, and finally an AFSK signal that is too narrow (this is the  
same case as a receive filter that is too narrow).

Now, when you pass all that through a practical transmitter, the  
behavior is not as obvious, but still visible.  I took some samples  
last night of the FT-1000MP's  transmitter.  The FT-1000MP feeds a  
dummy load through a directional coupler, and I used transformer  
coupling from the directional coupler (to minimize any ground loops)  
into the antenna input of a K3, from which I recorded the following  
two plots.

The first plot is that of the FT-1000MP running in FSK mode.

http://homepage.mac.com/chen/Technical/FSK/foxFSK.png

The data is a bunch of repeated "The quick brown fox jumps over the  
lazy dog" with lots of spectral averaging (each FFT bin is  
individually exponentially filtered).

The second is the FT-1000MP running in AFSK mode (no ALC, naturally)  
with audio from cocoaModem that uses a moderately tight Blackman- 
Harris windowed filter.

http://homepage.mac.com/chen/Technical/FSK/foxAFSK.png

Notice that the transmit phase noise predominates in the spectrum  
beyond 100 Hz away from the FSK carriers.  Even with that, you can  
still see that the AFSK signal is a bit narrower (look at the 600 Hz  
region and the 1600 Hz region and you can tell that the FSK signal is  
wider than the AFSK signal, even though both are using 170 Hz shift.   
The difference would be more obvious with a rig that has better  
transmit phase noise.

In any case, take a look at both FSK or AFSK signals 200 Hz away from  
the FSK carriers.  They are only about 35 dB down from the carriers  
themselves.  I.e., an S9+35 dB signal will clobber an S9 signal even  
if it is 200 Hz away.  No filter is going to fix this problem.  I.e.,  
there is no need to be so anal about filter bandwidths, and in any  
case, let your software demodulator do the heavy lifting in the  
filtering department.  The crystal filter only needs to provide a  
protection against clipping the sound card.

[Note that you can also "just" see the difference of the keying rate  
differences between the two plots.  The FSK plot is keyed at 45.00  
baud (MicroHAM microKeyer II's FSK generator) and the AFSK is keyed at  
45.45 baud.]

 From the last picture in the html page, you can see what a filter  
that is too narrow would do to a matched filter.  In the presence of  
noise, noise would also go down by a little with a reduced bandwidth,  
but not by as much as the peak of the matched filter goes down by  
(this should not be a surprise, all the textbooks on matched filtering  
tells you that :-).  That last picture has a filter cutoff just beyond  
the third harmonic of the keying signal, i.e., 22.725*3 Hz; in other  
words, a perfectly tuned RTTY signal inside a 306 Hz brickwall filter.

Conclusion: if there is a strong RTTY station close by, a 306 Hz  
brickwall filter will not be able to get rid of of QRM if it is more  
than 35 dB louder than the signal you are trying to copy, and in the  
meantime, you are losing a bit in the ability to copy a weak signal  
because you are cutting off too much of the keying sidebands.

Fortunately, even Inrad 250 Hz crystal filters don't really cut away  
much at 300 Hz.

You might want to use the captured spectra (the PNG files) to  
determine your personal rule of thumb of how far you want to be away  
from a nearby loud station during a contest, both for your sanity and  
the sanity of the nearby station (if he is loud at your QTH, you will  
likely be equally loud at his QTH :-).

73
Chen, W7AY