🚀 go-pugleaf

Don Hills wrote in message ...
>In article <bUpf2.1195$S7.1520@news.flash.net>,
>"Arny Kr�ger" <arnyk@flash.net> wrote:
>>
>>Even I've been surprised with how much trouble people have hearing the
>>difference that reducing 44/16 to 44/10 creates. I fully expected people to
>>hear clearly audible differences at 12 bits.
>
>It's not so surprising when you consider the following...
>In the UK, the nationally networked FM radio programming is carried from the
>studio to the transmission sites in digital form. The general opinion of the
>"golden ears" brigade is that it sounds quite good, and apart from the 15
>KHz bandwidth limit is quite close to CD sound if you have a good tuner
>(Sequerra etc). But what many of the listeners fail to appreciate is that
>the digital data stream is actually encoded as 35 KHz sampling rate, 10 of
>14 bit audio (NICAM). That is, only the 10 most significant bits of any
>sample are transmitted.

Well, FM, whatever! ;-) (I am aware of this, or at least I was aware that
something like this was a standard in the UK at one time)

>When you made the low-bit versions for the CD, did you simply truncate the
>bits or did you also add the correct amount of dither?


I did the rawest, crudest truncation I could do. No added dither. I wanted it
to sound NASTY!  The actual process was that I simply attenuated the program
material in -6 dB steps in a 16 bit environment, and then boosted it
correspondingly. The low order bits got truncated, the levels were unchanged
+/- about 0.01 dB and that was that. I did some technical tests, and this
procedure affected test tones in the way I expected.

Did not work out, subjectively, the way I planned! ;-)

The hidden item is the fact that most program material is pre-dithered at 7 to
10 bits  by residual noise on the recording that traces back to the
performance.

I've been isolating the noise on truely mic-open, but no-music segments of
CD's I have.

It has turned out that the Ricky Lee Jones track that Doug Stabler recommended
is one of the quieter pieces around - it has a residual noise level that is as
much as 72 dB below digital FS (if the recording normalized to digital FS in
accordance with modern standards).

Other quiet segments have shown residual noise as little as 52 dB below FS on
what are claimed to be DDD recordings!

I conclude that we need quieter HVAC and mics, before we get too excited about
the audbile benefits of long digital words for musical data. ;-)