[RTTY] RTTY Bandwidth. Was: 18 Mhz Band

[Joe Subich, W4TV]

Chen,

Thanks for the empirical confirmation.  When I posted my original reply
the 340 Hz value was from memory rather than bothering to consult the
ITU (also NTIA "Appendix J") documents.  For the benefit of others, the
ITU/NTIA formula simplifies to <baud rate> + 1.2 * <shift> where baud
rate is in bits per second and shift is in Hertz.  Thus for RTTY with
170 Hz shift the "necessary bandwidth" in Hz is 204 + baud rate.

For practical purposes, US amateurs can operate within 100 Hz of the
band edge if their transceiver displays the FSK mark frequency (e.g.,
Elecraft K3, Yaesu FT-1000/2000/5000/9000 series, etc.).  When using
USB AFSK, it is necessary to subtract the *higher* of the two tones
*plus* 100 Hz (or the "center frequency" plus 200 Hz) from the band
edge to determine the highest acceptable ("safe") USB "dial" frequency.
*note:* - 100 Hz for FSK and "highest tone plus 100 Hz" or "center
frequency plus 200 Hz" are approximations meant to be easy to apply
in the heat of battle ... and intentionally err slightly (50 - 75 Hz)
on the side of safety.

73,

    ... Joe, W4TV


On 4/9/2012 2:55 PM, Kok Chen wrote:
>
> On Apr 8, 2012, at 4:25 PM, Joe Subich, W4TV wrote:
>
>>> Not quite sure how close to 110 one can get, but probably less than
>>> 109.5 is too close.
>>
>> With a theoretical bandwidth of around 340 Hz for 170 Hz shift 45.45
>> baud FSK, one could get as close as 18,109.9 (MARK).  For USB dial
>> frequency (AFSK), one would want to stay below 18,110 - MARK - 85 Hz
>> which, for 2125/2295 tones (18,110 - 2.295 -.085) is about 18,107.6 KHz.
>
> I did a bit of literature search and there are some conflicting numbers out there regarding "occupied bandwidth."
>
> The ARRL Web site's DominoEX page (http://www.arrl.org/domino) cites a bandwidth of 453 Hz for 45 baud RTTY.
>
> This information is repeated in the Appendix of the Fourth Edition of Steve Ford's "ARRL's HF Digital Handbook," (ISBN 0-87259-103-4) and also mentions that the ITU Emission Designator for RTTY is 450HJ2B (450.0 Hz).
>
> However, if you refer to the ITU web page on Emission Designators (http://life.itu.ch/radioclub/rr/ap01.htm), the "necessary bandwidth" formula for 45.45 baud, 170 Hz shift RTTY yields a value of 249.45 Hz.
>
> Moreover, there is an overriding rule for US hams in Part 2 of Title 47 of the United States Code of Federal Regulations (amateur specific rules are in Part 97), which you can find here: http://cfr.vlex.com/vid/1049-measurements-occupied-bandwidth-19845042 , specifically Part 2.1049 which specified precisely how to compute "occupied bandwidth."
>
> Part 2.1049 states
>
> "The occupied bandwidth, that is the frequency bandwidth such that, below its lower and above its upper frequency limits, the mean powers radiated are each equal to 0.5 percent of the total mean power radiated by a given emission shall be measured..."
>
> together with some specifics, for example CW transmitters are measure with a keyed rate of 16 dots per second, the -35 dB point for harmonic distortion, etc.
>
> 2.1049 defines the occupied bandwidth as the bandwidth where 99% of the average power lies, and more specifically, 0.5% lies below the bandwidth and 0.5% lies above the bandwidth (since the spectrum may not be, an in general, is not, symmetrical).
>
> So, I set off last night to create a program that generates continuous phase RTTY (which presumably all modern RTTY FSK transmitters comply with, and certainly, all AFSK software modems should be using). Then, grabbing frames from it to perform an FFT.  Each FFT has a resolution of 0.5 Hz, and approximately 1000 seconds worth of spectrum is averaged.
>
> Here are the results (for 45.45 baud, 170 Hz shift RTTY):
>
> For an RYRYRY sequence (arguably the largest bandwidth hog text sequence, which is why we use it for testing purposes), the "99% bandwidth" rule gave 314.4 Hz for 1 stop bit, 300.3 Hz for 1.5 stop bits and 295.4 Hz for 2 stop bits.
>
> When I used the "Quick brown fox" text sequence, I got an FCC occupied bandwidth of 269.0, 256.0 and 249.5 Hz (for 1, 1.5 and 2 stop bits, respectively).
>
> Indeed, by using 2 stop bits, the simulation for the FCC "occupied bandwidth" with "average text" (not worst case text like RYRY) yields a result that is practically spot on the formula that is given by the ITU site.
>
> (I got curious and tried "<CR>CQ CONTEST W7AY W7AY CQ" and got 262.2, 256.3 and 250.0 Hz, HI HI)
>
> Anyhow, this confirms that the "occupied bandwidth" of steam RTTY is closer to 250 Hz (you can use 270 Hz if you use 1 stop bit) than it is to 400 Hz or 453 Hz.
>
> It is possible that the wider 400 Hz numbers that I cited above are the result from non-phase continuous RTTY generation (i.e., computed from the convolution of a sin(x)/x function with the Mark and the Space carriers), but in practice, all RTTY signals today are of the phase continuous variety.
>
> Even the "Keyed AFSK" generators in the FT-990/1000D are phase continuous by nature of the pseudo D/A convertor that is built from a ring counter.
>
> Remember that this has nothing to do with the RTTY receive filters.  The numbers above are the pro-forma bandwidths of a transmitted RTTY signal to comply with FCC rules.  When unfiltered, the keying sidebands of a transmitted RTTY signal is much wider.  However, most of the keying sideband fall within the "0.5% rule."
>
> The narrowest RTTY receive filter that incurs no inter-symbol interference is a pair of Raised Cosine filters that puts the total bandwidth at about 215 Hz.  As Victor Poor K3NIO mentioned in an 1964 RTTY Bulletin article, you can get close to the performance of the Raised Cosine by using an analog 3rd order Butterworth filter whose signal bandwidth is 225 Hz.
>
> Both of these are narrower than the "occupied" transmit bandwidth and you do lose a little of SNR because of it.  The "optimal" RTTY filter is a Matched Filter, which lets through all keying sidbands (K3NIO also alludes to this fact in his 1964 article).  If you don't use either the Raised Cosine or the K3NIO filters, the receive filter would have to be much wider to avoid inter-symbol interference.  I have earlier found a whole class of receive filters that has no inter-symbol interference, and approaches the performance of the Matched Filter, which I described here:
>
> http://w7ay.net/site/Technical/Extended%20Nyquist%20Filters/index.html .
>
> 73
> Chen, W7AY
>
>
>