Re: LF: Key Clicks

[Rik Strobbe <rik.strobbe@fys.kuleuven.ac.be>]

Hello Hugh,

Just look at CW signal as an AM signal.
Assume you have very hard edged "on / off" keying  and key
an long series of dots at a speed of 12 WPM : 
12 WPM = 600 dotlengths per minute (according to the "PARIS"
system 1 word = 50 dotlengths), so in fact you modulate an AM TX with a 5
Hz rectangular "audi signal".
The spectrum of this signal will be a carrier with a whole bunch of
sidebands in 5 Hz intervals from the carrier. The first sidebands
(carrier +/- 5Hz) is useful all the other (higher) sidebands are waste
(wasted power + useless bandwidth) and and show up as keyclicks.
As the keying gets softer (longer rise- and falltime of the keying) the
higher sideband will become weaker (less keyclicks). In order to have no
higher sidebands at all you would have to get modulation (= keying)
sinusoidal. In that case the modulation signal has no harmonics and thus
keying will be clickless.
So far the theory, because the above assumes a long (a la limite endless)
series of dots.
Our real-world CW instead is a rather random sequence of dots and dashes.
But regardless how you do the keying the worst condition (largest
bandwidth) is a series of dots. This means that the "useful
bandwidth" of a CW signal is determined by speed and the maximum
"audio component" of the keying can be determined as 
:

Fmax = WPM/2.4    where Fmax = the highest useful  "audio frequency" (in Hz) and WPM = keying speed (in words per minute)

So the ultimate way to get a clean CW signal is to filter they keying
signal through a good LPF at Fmax.
But in most cases a simple RC filter will give the best "effort vs.
result" ratio.

73, Rik  ON7YD

At 10:32 25/08/2004 +0100, you wrote:
> Good morning all,
> I keyed my original LF TX with a relay - crude, but it got me on the
> air.
>
> I know many designs just filter the keying waveform to "round off
> the
> corners" of the envelope.
> But I read that the best envelope shape is a "Gaussian
> waveform". This
> produces the smallest sidebands.
> [
> http://fermi.la.asu.edu/w9cf/articles/click/click.html
> and lots of other
> stuff ]
> RC and LC filter circuits produce exponential waveforms, although the
> waveform diagrams that accompany these circuits in the texts often make then
> look as though they produce a gaussian waveform.
>
> My question is ...
> Is there a circuit that will produce a nice "gaussian wavform" that I could
> feed into an AM modulator to produce really clean CW - or is there no such
> thing ?
>
> 73
> Hugh M0WYE