Login  Register

Re: Fwd: New Sherwood report

Posted by Kok Chen on Dec 02, 2010; 7:32pm
URL: http://elecraft.85.s1.nabble.com/Fwd-New-Sherwood-report-tp5793377p5797522.html

On Dec 2, 2010, at 12/2    4:32 AM, Barry N1EU wrote:

> Are you saying that wide sidebands measured with the rig at 4msec  
> rise-time are probably not going to be even wider with the rig at 1  
> msec rise-time?

Not that much, since the really far off keyclicks are mostly from  
higher order discontinuities.

Remember that the RF spectrum for CW is approximated by a rectangular  
pulse that is convolved by the impulse response of the waveshaping  
filter if the filter is linear (i.e., none of the silly diodes and so  
on used to attempt to tame keyclicks).  The short way of saying it is  
that the pulse is filtered in the time domain by the waveshaping filter.

The "convolution theorem" thus states that the Fourier of the  
resultant filtered signal is just the *product* of the Fourier  
transform of the CW pulse train with the Fourier transform of the  
waveshaping filter.

If you assume that the unfiltered pulse train is completely unfiltered  
(i.e., perfectly sharp edges and so on), a pulse train of dits will  
have a Fourier transform that is a sinc(f) (i.e., sine of f divided by  
f) function in the frequency domain.

The spectrum of the waveshaping filter is simply the stuff you can see  
at the Wikipedia page.

Multiply the two to get the RF spectrum of a waveshaped CW, and you  
will find that the sinc(f) is predominated by the waveshaping filter  
for far off frequencies.

Since the power of the unfiltered CW keying signal (i.e.,  
sinc(f)*sinc(f)) dies down very slowly, all the far off spectrum is  
determined by the "tail" of the spectrum of the waveshaping filter.

Without waveshaping, the power of CW keying pulses die down  
asymptotically as 1/( f*f), since the spectral envelope of sinc(f)  
dies down as 1/f .

The same scenario holds true for RTTY.  You can think of an FSK signal  
as two pulse trains, when one is on, the other is off (i.e., really no  
different from a CW signal :-).

With randomly generated FSK, you have the same sin(x)/x spectral  
envelope.

If you now make sure that the mark-space phase transition of FSK is  
smooth -- what many people call "phase continuous FSK"," RTTY improves  
a lot (this is what the K3 and many other rigs and software do).

"Phase continuity" is basically saying there is no first order  
discontinuity of the temporal waveform, but if the Mark and space  
frequencies are different (of course), there will be second order and  
higher order discontinuities unless all mark/space switching are done  
when their carriers are right at zero -- not very practical).

Just like CW, you can also make your RTTY signal friendlier to your  
spectrum neighbors by waveshaping the FSK signal some more.

You can see this process here (those are actual recorded AFSK signals):

http://homepage.mac.com/chen/Technical/FSK/Sidebands/sidebands.html

Instead of two peaks of the FSK signal, you can visualize in your mind  
a single peak for a CW signal that is represented by the one sided  
spectrum, and the spectrum for first oder discontinuitity, second  
order discontinuities, and waveshaped FSK (I had used a simple  
Blackman in the Web page above) pretty much holds for the CW case also.

I don't know of a single FSK rig today that is as friendly to  
neighbors as the last two spectra shown in the web page. The Omni V  
and VI does a little of waveshaping of the keying signal that is  
applied to the varicaps to generate the FSK signal.

I also don't know of any software that applies RTTY waveshaping,  
although it is easy to do for all AFSK software.

You need to be careful, of course. Since any waveshaping will cause a  
slight overlap between the mark and space carriers, too much  
waveshaping and you again run into that pesky transmit IMD problem  
that John (juergen) mentioned.

For that reason (and also not to degrade SNR at the end of the matched  
filter that is discussed in the above web page), cocoaModem doesn't go  
too far off the deep end when wave shaping.  It actually survives  
pretty well through the K3's transmit IMD.  You can see that in the  
last two spectra on this page:

http://homepage.mac.com/chen/Technical/K3/Digital/digital.html

The K3 native (i.e., using a paddle) FSK signal (measure by a separate  
receiver) is in the second last image, and cocoaModem's waveshaped  
RTTY (using K3 DATA-A) is the last image on that page.  You can  
actually see IMD spikes dues to the intermodulation between the now  
overlapping mark and space signal.  But overall, the QRM is still  
below the spectrum from the K3's FSK signal.

But if the K3 transmit IMD can be improved, you can squeeze more RTTY  
stations in during a contest.  The FSK ops will then get the same  
reputation as the FT-1000D CW ops :-) :-).

The RTTY spectra in the second web page has random (well, LTRS Baudot  
characters) bit modulation and I had apply a pretty per-bin filtering  
of the spectra to remove the noise from the receiver and sound card,  
so it appears smoother and less serrated.

But you can see the difference between something that has first order  
discontinuities, something that has no first order discontinuities,  
and something that also try to suppress higher order discontinuities.

Not much happens in the nearby spectrum.  But as you go further away  
from the keyed carrier, the difference between a "good" waveshape and  
a "very good" waveshape is quite large.  And, at some point (as seen  
by the IMD spikes in that last plot), the transmit IMD becomes the  
"weakest link."

73
Chen, W7AY

______________________________________________________________
Elecraft mailing list
Home: http://mailman.qth.net/mailman/listinfo/elecraft
Help: http://mailman.qth.net/mmfaq.htm
Post: mailto:[hidden email]

This list hosted by: http://www.qsl.net
Please help support this email list: http://www.qsl.net/donate.html