[Elecraft] K3 2.7kHz filter DSP FC

[Paul Christensen]

> Billions of people are quite happy with the 300kHz cutoff used by the 
> public telephone system (a total of 3.1kHz between 300 and 3.4kHz).

Presumably, you mean 300 Hz, and not 300 kHz.

> The critical frequency for speech communications is the lowest formant 
> frequency; it is the resonances in the vocal cavity that convey speech, 
> not the fundamental.

You're referring to audio passband in which minimum bandwidth is used as a 
function of maximum articulation.  Bell Labs and other acoustical 
researchers in the late 19th and early 20th centuries found that reasonable 
bandwidth for voice communication was in the area you describe.  The concept 
was further carried on with the advent of HF SSB transmissions in the late 
'40s, although AT&T Long Lines was already economizing SSB bandwidth well 
before then.

As a practical matter, analog circuits in which crystal filters are used 
during SSB generation (e.g., Collins) precluded audio transmission below 300 
Hz.  But we're not talking about land-line voice circuits here and 
technology has marched-on well past the usage of  SSB-generation techniques 
that rely on crystal filtering for economy.  Through DSP, we easily have the 
ability to extend low frequency content down well below 300 Hz and a 
significant list of other tranceiver manufacturers are accommodating this 
demand in the amateur radio community.

> The lowest formant in normal speech (formant 1 for "u") spans about 200 to 
> 400Hz, so 250Hz probably is a reasonable compromise; it will include the 
> peak and only slightly distort the lower edge.

It may be a compromise, but a considerable amount of voice detail occurs 
below 250 Hz, even though it offer little or nothing to intelligibility and 
articulation.  Using a microphone, a sound card, and FFT software (e.g., 
SpectraPlus) observe the extent of low frequency content of you own voice. 
If you believe that relevant energy is limited to 250 Hz, you need to 
research this in better detail.  We already have reasonably good 
articulation in the bandwidth between 300 Hz and ~ 2.8 kHz but little added 
bandwidth is required to add nearly two acoustical octaves to the listening 
experience.

Incidentally, an argument can be made that to maximize intelligibility, the 
upper passband should actually be closer to 3.5 kHz or even 4.0 kHz (see a 
graph of the now-classic Fletcher-Munson family of loudness curves and 
re-determined in later years by modern researchers).  Beyond that limit, 
diminishing returns on intelligibility occur.  The reason the ear is most 
sensitive in this area is that the ear canal forms a closed pipe against the 
tympanic membrane.  The resulting fundamental frequency and harmonics can 
then be calculated with precision by the width of the ear canal and its 
length.  Of further interest is that maximum energy in the human voice is 
also matched-up against the curves.  So, either millions of years of 
evolution or God made this association possible.  Take your pick.

The key point is that compromises must be made againt the current 
state-of-the-art.  It is my belief that the benefit of adding a Rx/Tx range 
down to ~ 70 Hz greatly outweighs the relatively little extra bandwidth 
needed (less than 200 Hz of B/W) to make a profound difference in the 
listening experience.

Paul, W9AC