"Mike Coslo" wrote

The question kind of states it. I suppose that the BW might be wider as
the speed increases.

There are three different bandwidths that come into play. They are
"necessary bandwidth", "effective (or actual) bandwidth", and "apparent
bandwidth"

Necessary bandwidth in hertz for copying a morse signal is defined as Bn=BK
where B is modulation rate measured in Baud, and K is an overall numerical
factor which depends on the allowable signal distortion. The commonly used
values of K are 3 for non-fading paths, 5 for fading paths, and 8 for
fading/multipath smearing. From the formula you can see that higher speeds
(Baud) require more bandwidth, just as you supposed. The nominal "necessary
bandwidth" presumed for CW is 100Hz which is based on 25WPM (20 Baud) over a
fading path (B=20, K=5). Quite honestly, "necessary bandwidth" is primarily
an academic exercise and planners tool, as it ignores some practical 'real
world' issues and doesn't answer the question raised in your subject line.

Effective bandwidth is an actual on-the-air measurement of the width of the
signal at some designated level, most commonly -60dB referenced to the peak.
To understand what is being measured, you need to recognize that Morse is
sent as an amplitude modulated carrier (AM) and that it contains sidebands.
Like any AM signal, those sidebands extend nominally plus/minus the carrier
at the frequency of the modulation, or BW=2M. Modulation of this signal
contains two components.

The first component is the baud rate of the actual on/off keying (see
"necessary bandwidth" above). Were it only for this component, measured CW
signals would be very narrow, 100Hz, and dependent totally on keying speed.

The second modulation component is related to the rise time of the radiated
signal. Fast rise times (where the RF envelope resembles a square wave)
generate signals rich in harmonics and as these harmonics mix with the
primary signal and each other in the transmitter stages, they produce sum
and difference signals which become part of the sidebands of the radiated
signal. The sharper the rise time and the more non-linear the transmitter
stages, the more energy there is in the harmonics, and thus the bandwidth is
wider (as measured at -60dB skirt points). Controlling this component of
bandwidth can take the form of regulating the rise time (shaping in the
keying circuit) and ovoiding overdriving of transmitter circuits.

The third kind of bandwidth is "apparent bandwidth". This bandwidth is
determined by the effective bandwidth (see above) AND the performance of the
receiver environment. If a receiver were "perfect", then effective and
apparent bandwidth would be equal (the receiver would perfectly reproduce
the desired signal in the form it arrived at its antenna and would reject
the effects of all non-target signals present.)

But receivers aren't perfect (well, maybe my Sherwood equipped R4C is
close). Extremely loud signals (your neighbor 3 doors away) will sound
("apparent") several hundred kHz wide, because your receivers AGC will pump,
RF and IF stages will be overloaded, and the faster he sends the worse it
will be. I'm giving the obvious extreme example, but just to make the
point. Many times just some reasonable adjustments of your receiver such as
turning off noise blankers, reducing the preamp level, or turning your
antenna will reduce the apparent bandwidth down in line with the actual
bandwidth of the transmitted signal.

73, de Hans, K0HB