[CW] CORRECTION - Speed vs Bandwidth

[George, W5YR]

Some controversy exists over what is responsible for keying artifacts, but
it seems fair to say that rise and fall times and the actual keyed waveshape
are the principal causes.

In my view, keying speed enters into the picture only if the rise and fall
times are decreased at the higher keying rates in order to maintain
readability and prevent "soft" keying which tends to run the code elements
together. Sending faster, like talking faster on SSB, does not tend to
require a wider bandwidth if the keyed waveform is not changed.

The theory of bandwidth vs rise/fall time is clearly understood and has been
around since the late 19th century, so nothing much new there. Any text on
transform calculus applies.

The remaining element to the picture, I believe, is the actual keyed
waveform or shape. The critical points of the waveform which can account for
bandwidth increases not predicted solely by rise and fall time are the
transitions in the waveform from off to leading edge start; leading edge
finish to on; on to start of trailing edge; and end of trailing edge to off.
If these transitions are abrupt, transform calculus tells us that
high-frequency transients will be produced that can be either prevented or
minimized by making the transitions "gentle" or "rounded" instead of square
and abrupt.

An example is the raised cosine keying waveform which at slow speeds is
virtually without transients but which is difficult to read at higher speeds
or in QRN.

It is encouraging to see that some manufacturers are paying attention to
this aspect of keyed waveform design. In a DSP transmitter, it is almost
entirely a matter of firmware and there is little excuse of having a CW
signal with clicks or any distortion of dot/space timing at any speed.

73, George W5YR
Fairview, TX
w5yr@att.net
http://www.w5yr.com


----- Original Message ----- 
From: "David J. Ring, Jr." <n1ea@arrl.net>
To: "George, W5YR" <w5yr@att.net>; <cw@mailman.qth.net>
Sent: Sunday, March 14, 2004 11:50 PM
Subject: Re: [CW] CORRECTION - Speed vs Bandwidth


> I can't find that.  I know these figures aren't for square waves.
>
> Probably they are adjusted for softness or hardness depending on the
> circuit.
>
> You need a harder waveform for a bad circuit.  The harder waveform makes
> more harmonics and a wider signal.
>
> That would put the waveshape at 5 ms for soft (non-fading circuits) and 3
ms
> for hard (fading) - which I think is about right range, but 5 ms is fine
for
>
> I have seen 20 ms waveshape which of course made the circuit ring like a
> bell - beautiful copy for local work though - fun to copy when signals are
> right and code speed is below about 25 wpm.
>
> On the other hand if the message has to get there, and conditions are
> terrible, I would adjust the waveshape to 3 ms or so - there would be much
> more clicking - but that helps reception.  Something I'd recommend to
> QRPers!
>
> Do you have any sources on this?  I can't find the definative answer!
>
> 73
>
> David Ring
> N1EA
>
>
> ----- Original Message -----
> From: "George, W5YR" <w5yr@att.net>
> To: "David J. Ring, Jr." <n1ea@arrl.net>; <cw@mailman.qth.net>
> Sent: Monday, March 15, 2004 12:16 AM
> Subject: Re: [CW] CORRECTION - Speed vs Bandwidth
>
>
> > David, what is the assumed waveshape?