On Jan 19, 2013, at 2:48 PM, Carter wrote:
> a) FSK that is generated internal to the rig where the manufacturer 'keys' an
> audio tone -- an 'internal' AFSK if you will, that the rig manufacturer calls
> 'FSK'. This can be OK as hopefully the rig maker is controlling many of the
> variables such as level, mark/space switching, etc versus the operator
> jamming his sound card into the mic jack.
We used to call it "Keyed AFSK" where the FSK keying line actually keys an
internal AFSK generator in the rig.
The FT-1000D, and my old FT-990 are of this variety.
> b) What I call 'true' FSK as implemented in my old Drake T-4XB and TenTec
> Omni 6. This true FSK switches a varicap diode off and on, changing the
> frequency of the VFO by adding capacitance. IMHO, this is the best/cleanest
> way of getting on RTTY.
Notice that even the Omni VI (and Omni V) only keys a low (I forget now,
something around 4 MHz?) frequency oscillator by switching a trimmer
capacitor, that oscillator is mixed up to the proper RF frequency.
The capacitor switching method has its own pitfalls, I still remember when
VY1JA transmitted with something like 100 Hz shift after his Omni V made a
journey by truck from Vancouver to Whitehorse -- must be bouncy roads over
there :-).
There are quite a few more "FSK" methods.
(c) use the FSK function in direct digital synthesizer chips. This is used in
my FT-1000MP and probably many other rigs of the 1990-2010 era.
(d) generate FSK by using DSP. This is popular for the current generation of
rigs. The Elecraft K3 is one of these. The FSK is generated numerically at
around 15 kHz I.F. converted to an analog signal, and then get mixed up to the
right RF frequency.
Of these, (a) can obviously be modified to include an AF bandpass filter after
the internal AFSK generator. Method (b) can also be waveshaped in the same
manner we used to waveshape CW keying pulses to reduce keyclicks. Usually with
some RC time constants.
There is unfortunately nothing we can do about the method in (c) (FSK in the
DDS VFO).
The best chance we have is the DSP FSK. Not only is it easy for the
manufacturer to implement a filter (or a waveshaping window) it does not cost
any more hardware if there is enough memory already. The manufacturer can
simply send out a firmware update and bingo, you get click-free FSK.
Notice that Elecraft is already using DSP to waveshape their CW. So they know
about getting rid of keyclicks this way. They just need to also do it for FSK.
> 2) AFSK:
> The second -- and more subtle -- issue is that "zero crossing" switching
> between mark and space should be implemented as Irv Hoff W6FFC (SK) used when
> he created his AK-1 (?) AFSK unit that went along with his ST-6 demodulator.
Switching at zero crossings ensures that the energy of the keying sidebands is
small (since the first order discontinuities occur when the power is near
zero). But there are higher order continuities too.
With DSP techniques, we have an easier way to reduce sidebands, and that is to
use phase continuous switching. Like the zero crossing method, there is no
voltage jumps when you switch between Mark and Space. However, again just like
switching at zero crossings, there still are higher order discontinuities (the
slope of the signal is different from before you switch to after you switch;
but the second derivative is different, the third derivative is different, etc).
With phase continuous switching, you need not worry about introducing jitter to
the data bits that the zero crossing method creates. With a 2 kHz AFSK tone,
the jitter can be up to 0.5 ms.
To see the difference between the completely free-for-all switching and the
phase continuous switching (which produces the same spectrum as the Hoff
method), compare the first two spectra in
http://www.w7ay.net/site/Technical/RTTY%20Sidebands/sidebands.html
Fortunately for us, all the FSK methods also switch phase continuously, so they
produce the spectrum in the second figure (as also shown in Andy K0SM's article
to represent FSK), not the really, really ugly one in the first figure.
However, there is still significant keying sidebands (which is wasted power,
just as CW keyclicks is wasted power). Except for Matched Filters, one uses
these keying sidebands. Again, like the CW keyclicks, the FSK keyclicks are
just QRM. The third image in the web article shows what filtered or waveshaped
AFSK looks like.
Of all the software modems, MMTTY, 2Tone and cocoaModem all have some form of
AFSK filtering. (cocoaModem sets the -6 dB point of the transmit filter at 351
Hz). There are probably others that I don't know of. As of this writing, I
believe fldigi is the one software modem which is known not to apply filters to
its AFSK output.
Even with DSP methods, there are two ways to produce narrower AFSK signals.
One is to use waveshaping, like the way the K3 reduces CW keyclicks. This is
used by 2Tone. The other is to use bandpass filters. This is used by MMTTY
and cocoaModem.
73
Chen, W7AY
_______________________________________________
RTTY mailing list
RTTY@contesting.com
http://lists.contesting.com/mailman/listinfo/rtty
|