On Oct 20, 2005, at 12:41 PM, Mike wrote:
> Good point about the "keyclicks" from the FSK set-ups. Any ideas
> on what causes the 'clicks? There were a number of stations that
> had them in the recent JARTS Contest.
Ian, G3SEK and I had traded some private email in the past over this.
You can view FSK as synchronous OOK (on-off keying - a.k.a. "CW") by
two carriers that are spaced by the RTTY shift.
When you transmit something like an RYRYRY sequence, you are putting
out almost exactly 50% of one tone and 50% of the other tone (except
that the stop bit skews the probabilities a little since it is 1.5
bits wide and you will therefore see one of the cross ellipses being
slightly more favored in that direction).
A perfectly alternating 45.45 baud ones and zeros will produce a
waveform that is a square wave at 22.7 Hz. This square wave has odd
harmonic components at 22.7 Hz, 3*22.7 Hz, 5*22.7 Hz, with the
amplitudes falling by 1, 1/3, 1/5, ...
I.e., if you are just looking at this square wave, if we reference
the first harmonic (fundamental) at 0 dB, the 3rd harmonic will be
about 9.5 dB down, the 5th harmonic will be 14 dB down, etc. We can
make up a table for "45 baud square wave" thus:
22.7 Hz 0 dB
68.1 Hz -9.5 dB
113.5 Hz -14 dB
158.9 Hz -16.9 dB
204.3 Hz -19.1 dB
249.7 Hz -20.8 dB
...
Because of the 1.5 stop bits, the RTTY signal has even harmonics too,
even if you are transmitting RYRYRY and definitely has even harmonics
when you are transmitting random text or diddles. But using square
waves should give an illustration of what is happening.
Keying a CW carrier with a 22.7 Hz square wave will produce sidebands
of the carrier that falls away from the carrier as the above table.
FSK consists of two carriers that are keyed synchronously (when one
is off, the other is on). Between the mark and the space, the two
sidebands add and the shape of the spectrum is more complex
(especially since 170 Hz is not a multiple of 22.7).
However, in the passband outside of the mark and space, the above
table is a reasonable approximation, since the mark and space are 170
Hz apart, and the influence of the sidebands from the tone that is
further away is pretty small.
From the table, you can see that when you are alternately switching
between mark and space, even when you are 250 Hz away from the mark
and space carriers, the FSK signal is only about 20 dB down. Thus
the -20 dB bandwidth of an unshaped FSK signal will be 250+170+250 Hz
wide, or 670 Hz!
These are analogous to keyclicks in CW. There is a need to limit
these harmonics.
It is easier to eradicate keyclicks in CW, since you only have a
single tone, and simply changing the attack and decay time of a CW
keying signal will do that nicely.
Waveshaping is trickier with FSK since this process might allow both
tones to be simultaneously transmitted (albeit not at full power).
What is the problem of two simultaneous carriers, you ask -- well
IMD, of course. Any nonlinearity will cause IMD byproducts that can
extend well beyond even the square wave drop off!
So, unless there is some attempt to minimize the overlap of the two
tones, you are faced with the same problem AFSK is faced with, the
IMD from the transmitter.
The Ten-Tec Omni V and VI actually does FSK the old fashioned way, by
modulating varicap into some one of the mixers' oscillator. You
probably can improve the keyclicks since the rig was probably
designed so that you can still use the faster baud rates (75 baud).
With DDS rigs like the FT-1000MP, you are at the mercy of the
manufacturer of the DDS chip, and can just hope that they'd slewed
between the mark and space tones in some nice fashion. Since the
rigs also don't have a baud rate menu, I suspect they are slewing at
faster than what is optimal for 45 baud RTTY (since they also have to
work with 75 baud RTTY). Just a guess on my part.
In my experiments, cutting off the fifth harmonic and above causes
practically no damage to print when demoduating an RTTY signal. That
means that you can strongly filter the transmitted signal so that it
only falls inside a (100+170+100) Hz passband or 350 Hz or so.
Those who receive with 300 and 350 Hz IF filters can attest to the
fact that 350 Hz passbands are fine for 45 baud RTTY.
It turns out that limiting the transmitted passband is easy to do if
you are using AFSK (you have to use phase linear filters so that
there is no group delays between the demodulated mark and space
signals at the demodulator -- but this is a piece of cake in
software, merely by using symmetric FIR filters).
I have no idea if all the modem programs actually has an output
filter for the transmitted AFSK that is adapted to the baud rate in
use; I am only familiar with the code of one program and that program
has an output bandpass filter.
Over the past two years, I have come to be convinced that if you can
put out a much cleaner signal in AFSK (blasphemy!) than you can in
FSK, and I have not used FSK since. You still have to make sure you
don't drive the rig into crazy IMD levels and when you do, an AFSK
signal can be much uglier than a bad FSK signal.
73
Chen, W7AY
_______________________________________________
RTTY mailing list
RTTY@contesting.com
http://lists.contesting.com/mailman/listinfo/rtty
|