msushi wrote in another thread:
"Noise is measured in volts per root Hz. To get the noise over a
bandwidth, you multiply by the square root of the bandwidth. So using
a 10% narrower filter reduces the noise by 4.8%. So narrowing the
bandwidth helps, but it is not a cure".

Indeed, it is not a cure. Narrowing the bandwidth of a filter also
narrows the audio response, so the narrower the filter is the poorer
will be the audio response and the more difficult it becomes to render
a rare faint DX signal readable. It is vital to my MW DXing of AM
stations that I get as fuller sound as possible. These signals are
very faint and often severely garbled by atmospherics. At best it is a
great challenge to read and identify these signals with enough
information to request and get a QSL. The fullest possible sound is
necessary to clearly hear all the speech harmonics and sibilants, so
as a result I usually DX in ECSS USB mode, but use a comparatively
wide 3.0 Khz filter setting on my Icom IC-756PROIII.

Then there are filters and there are filters, from the cheap ones
fitted to most portables right up to the Collins mechanical filters
that can be fitted to an AOR7030. Filter characteristics are very
important to the end result of rendering audible those faint DX
signals. The filters in most portables have very wide skirts that let
in all sorts of interference, from plain noise to monkey chatter from
adjacent stations. Even more expensive filters on some of the lower
end tabletops have very shallow skirts in addition to poor skirt
characteristics.

I once owned a new Kenwood R5000 to which I added the top of the line
Kenwood expensive crystal filters. What a disappointment they were.
The adjacent channel monkey chatter still came in as the filters had
poor skirts and were too shallow. A good filter has to have a depth of
at least 80 dB, and those Kenwood filters were only 60 dB, they turned
out to be a total waste of money.

Filters have various characteristics: their shape factor and their
depth. Most analogue crystal and mechanical filters have a bell shaped
characteristic, allowing the possibility of letting in powerful
adjacent channel break through. Then there is the depth of a filter,
most are bell shaped and not more than 60 dB in depth, again allowing
in adjacent channel stations to interfere with your target station. An
ideal filter has near vertical skirts and reaches to a depth of 80 dB
or better, hence my love affair with my 756Pro3 which has superb DSP
filters.

Lastly just a note to say that adding an expensive high quality
Collins mechanical filter to a poor radio like a Yaesu FRG-7 just will
not work. Everything has to come together for it to work: good
circuitry with a top class roofing filter, high dynamic range etc.
Generally you get what you pay for = the more expensive the radio the
better the filters and overall performance are going to be. And when
you are in very tough DX situations, that is what is going to yield
the results.

Enjoy your listening and good DX.

John Plimmer, Montagu, Western Cape Province, South Africa
South 33 d 47 m 32 s, East 20 d 07 m 32 s
RX Icom IC-756 PRO III with MW mods
Drake SW8 & ERGO software
Sony 7600D, GE SRIII, Redsun RP2100
BW XCR 30, Sangean 803A.
Antenna's RF Systems DX 1 Pro Mk II, Datong AD-270
Kiwa MW Loop, PAORDT Roelof mini-whip
http://www.dxing.info/about/dxers/plimmer.dx