[RTTY] K3

[Kok Chen]

On Jan 7, 2009, at 1/7    4:07 AM, J. Edward (Ed) Muns wrote:

> Yes.  In the 15kHz IF.  One frequency that is shifted for MARK and  
> SPACE.

Yes, but it should not be just shifting abruptly between mark and  
space tones.  Sharp transitions between mark and space is what causes  
a lot of rigs to put out a needlessly wide signal when operated in  
FSK.   It does not happen in the region between 0 and -25 dB, but in  
the region between -40 dB and below -- i.e., the rig will only bother  
you if it is coming from a moderately loud station.

Just like the CW boys are concerned about keyclicks on rigs like the  
FT-1000, we have the same problem with FSK sidebands.  During the  
Roundup, one 9-area call sign operating from California had an  
abnormally wide low-level keying sidebands -- I use a waterfall to  
S&P, and I could easily see it on the waterfall, but it was also  
audible as a low level buzzing that accompanies the keying.  It could  
be due to AFSK too; I am not placing the blame on FSK generators,  
since wide keying sidebands is not just restricted to poor FSK  
generators, and I have seen very beautiful spectra coming out of FSK  
rigs.)

A completely abrupt switching (whether FSK or AFSK) between Mark and  
Space will roughly create a keying power spectrum that falls at about  
a rate of 1/(f^2) for an RYRYRY text (i.e., the coefficients of the  
Taylor's series for sine of f, and then squared to get power).  In  
our case, f is a half of the baud rate, or 22.725 Hz.  (The 1.5 stop  
bits make it a little different, and worse, than 1/(f^2)).

What this means is that a "bad" (abruptly switching) FSK signal will  
be down only 20 dB when you are about 230 Hz away from the mark and  
space tones.

There at least two ways to solve this.  One is to use amplitude  
waveshaping on the tone (like what you do with CW).  Alex VE3NEA has  
a good article not too long ago in QEX about using Fourier windows to  
reduce CW keyclicks.  You can also see a detailed plot of another  
Blackman-windowed keying here: http://homepage.mac.com/chen/w7ay/ 
cocoaModem/UsersManual/cwManual/index.html#tx (scroll down to Figure 5).

When it is unwindowed or unwaveshaped, the keying sidebands are a  
function of the baud rate.  When windowed properly, it is a function  
of the window.  It is as if you are applying an envelope in the  
spectrum to depress the far out keying sidebands.

Unfortunately, if the PA has a lot of IMD, part of the good done by  
amplitude waveshaping is undone by the non-linearity of the amplifier  
since the envelope of such an AFSK signal is no longer perfectly  
constant.

That is why Ian and I have alluded to slewing between Mark and Space  
frequencies, instead of switching between them.  Phase continuous  
slewing will create a constant amplitude envelope.  However, the  
manner in which it slews will determine the response far (e.g., 100  
Hz) away from the tones, just as how the different windowing  
functions result in different far-off response (see plots in VE3NEA's  
article in QEX).

73
Chen, W7AY