All kinds of things too precious not to share, in short articles. Constantly under construction or destruction. Note that code snippets are usually working minimum examples (often stripped of extra features). If you plan to use and of the code for anything serious, you may want to contact me. It's likely that I can provide you with some updates and good (?) advice. fabian@fkurz.net
Fabian Kurz → Ham Radio → Stuff
Some examples (and a small program to generate such files) of different CW filter widths applied to the same received signal, at different signal levels.
| File | Description | SNR/dB (for each bandwidth) |
|---|---|---|
| sweep009.wav | Very weak signal (constant), bandwidth changing 100 → 500 → 1000 →2100Hz | 1.8, -0.4, -3.4, -6.4 |
| sweep011.wav | Weak signal (constant), bandwidths as above | 3.7, 1.3, -1.6, -4.8 |
| sweep020.wav | Moderate signal (constant), bandwidths as above | 8.8, 6.5, 3.5, 0.36 |
| sweep060.wav | Loud signal (constant), bandwidths as above | 18.3, 16, 13, 9.8 |
| 400-800-400.wav | Weak signal, 2.1kHz bandwidth, pitch changing between 400 and 800Hz. |
noise.tar.gz - C source code (12. February: Added SNR calculation)
Psychoacoustics suggest that at lower frequencies you can hear weak signals better.
Why is that? The ear works like a filter bank, of 24 frequency bands. At low frequencies,
those bands are of equal width and about 100Hz wide each. With increasing frequency, their
width increases. (See e.g. Bark scale)
If you now hear a weak signal within wideband noise (no other signals present), what your ear/brain signal processor actually has to figure out is only the signal vs. noise in the particular frequency band. All other frequencies are pretty much irrelevant.
Now it seems logical, that it's easier to discern a signal at the same strength in a frequency band that is only 100Hz wide than in a wider frequency band.
That also explains, why it hardly matters whether you listen to a weak signal in a 100Hz filter or a 2kHz wide filter (with, mathematically, very different SNRs, as can be seen in the table). Your ear/brain does all the work. The filter bandwidth only matters in the presence of other signals, QRM etc.
Another point worth considering is, if you are listening to a desired signal, plus a jammer signal 100Hz above. At 800Hz, this means the signals are 12.5% separated, at 600Hz it is already 16.6%, etc.
There are many benefits from chosing a low BFO frequency, few benefits from spending a lot of money for super narrow filters. Note that (real) narrow filters also have a higher insertion loss than wide filters.
2011-06-25: A few slides about this and similar topics are covered in my presentation "Morse Code — Alive and kicking!"
Added: 06-Feb-2008. Last modified: 31-Aug-2020.