[RTTY] bartg sprint 75 sv1dpi

[Kok Chen]

On Sep 19, 2011, at 3:04 PM, Joe Subich, W4TV wrote:

> The 45.45 baud ("Standard RTTY") profile appears to work fairly well
> except for the Pre-Filter bandwidth.  I have changed the 45.45 baud
> Pre-Filter settings to Taps=192 and FW=100 for a flatter passband and
> sharper skirts.  For 75 baud, I have been experimenting with Taps=256
> and FW=150 to avoid the "too narrow" issues with 75 baud.

An optimal filter for 45.45 baud RTTY consists of two peaks at the Mark and Space tones, each peak is exactly 45.45 Hz wide at -6 dB points.  With 75 baud, the -6 dB points of the two peaks are at 75 Hz.

The narrowest 45.45 baud filter (best QRM rejection) has a Raised Cosine spectrum

http://en.wikipedia.org/wiki/Raised-cosine_filter:

Normally (using beta=1), a composite mark-space filter for 170 Hz shift is 216 Hz wide at -6 dB and drops to infinite attenuation at 261 Hz.

The widest and still optimal 45.45 baud filter (best weak signal reception) is a Matched filter:

http://en.wikipedia.org/wiki/Matched_filter

The Matched filter for 45.45 baud RTTY also has -6 dB points at 216 Hz.  However, the spectral envelope of the filter rolls off very gradually and the filter still has significant response 500 Hz away.  A Matched Filter is therefore viewed as being very wide.

Notice that neither_of_the_above_filters_has_a_flat_passband, but they each have a very mathematically precise passband to satisfy the Nyquist criteria.

You can use other filters with widths in between the two extremes above.  But the shapes have to satisfy the criteria in the 1928 Nyquist paper.   If you use some arbitrary crystal filter (especially ones with poor phase response), you will need to widen the filters to quite a bit more than 216 Hz to avoid the inter-symbol interference (the received Mark bits "bleeds" into the space bits and vice versa). 

For 75 baud, the optimal narrow RTTY filter is 6 dB  down at 245 Hz and have no response outside a 320 Hz width.

Basically, the -6 dB points have to be very precise and the shape of the filter has to also be very precise, or you will need to widen the filter by quite a lot to avoid inter-symbol interference (a signal QRMing itself, so to speak).  

Even though they are wider than the 216 Hz filters I described above, the "250 Hz" crystal filters in transceivers will throw too many errors due to inter-symbol interference -- i.e., they don't print well even when there is no QRM.

A software modem can easily implement a 216 Hz wide FIR filter that is based on the Raised Cosine filter when there is QRM.  They can also easily implement the Matched Filter for DX work (both RITTY and cocoaModem implement Matched Filters).  Similarly, they can easily switch to using a different set of coefficients for 75 baud.

Perhaps someone should persuade MMTTY folks to implement Raised Cosine filters and Matched Filters?  It only involves a few minutes of trivial math and you just feed the resultant coefficients into the FIR filter and everybody benefits (especially people who use low power and small antennas :-).

73
Chen, W7AY