Return to the archive index

Re: Hardware Sound Compression?

From:
Date: Mon, 2 Nov 1998 07:56:09 -0600 (CST)

On Sun, 1 Nov 1998, Rodney Arne Karlsen wrote:

> I think if you converted the audio to 1bit information like in an audio CD
> that would bring down the size. 1 bit  vs. 16 bit at same quality. I am not
> shure how the 1 bit DAC works, baybe somone in the group could explain it to
> us.

I'm not sure about how 1-bit DAC works, either, although I suspect that it
is something that decreases a manufacturer's costs more than it increases
sounds quality. I think they put it on portable CD players for purely
marketing reasons, eg. "Now with 1-bit DAC and Fritzen Jammin circuits!"
In any case, CDs still store the audio information in 16 bit per sample
per channel, 44,000 (or is it 48,000? I forget) samples per second. Sum
total: about 10MB per minute of music.

There are only two ways to decrease the storage requirements for sound that
I know of. The first is the easiest. You simply decrease the sample size
and/or rate. This has the obvious effect of decreasing quality as well.
However, in a wearable, CD-quality audio is probably not the biggest
concern, and telephone-quality audio would be acceptable for verbal alerts
and such, and is relatively low bandwidth.

The other way to decrease your storage requirements is to go the route of
mp3s (MPEG-II, layer three, to be precise), which uses perceptual encoding.
This is the process of removing the parts of the sound that we know from
experience that a human won't be able to hear. To use a slightly flawed
example, if you recorded the sound of a pin dropping on a table, you'd be
able to hear that in the playback; and if you recorded a brick dropping on
the same table, you'd be able to hear that in the playback as well; but if
you recorded the pin and the brick dropping at the same time, you probably
wouldn't be able to hear the sound of the pin over the sound of the brick.
So, why encode the sound of the pin in the first place? To achieve this, mp3
encoding breaks up the signal into 40 different frequency bands, and cuts
the sound into 70 (or is it 72?) frames a second. It performs this
pin-vs-brick comparison on all the frequency bands in each frame.
The problem with perceptual encoding is that it is processor-intensive. As
has been noted on this list, it would take a very fast processor to encode
CD-quality sound at realtime, and it takes at least a Pentium-100 or so to
decode the same in realtime. Not to mention patent problems with the
Fraunhofer Institute.

I hope I didn't bore those of you who already know this. I wanted to
get it out there for everyone, just in case.

Paul Archer

----------------------------------------------------
A key to the understanding of all religion is that
a god's idea of amusement is Snakes And Ladders with
greased rungs.   -- Terry Pratchett, "Wyrd Sisters"
----------------------------------------------------

--
Subcription/unsubscription/info requests: send e-mail with subject of
"subscribe", "unsubscribe", or "info" to 
Wear-Hard Mailing List Archive (searchable): http://wearables.ml.org

+Previous Message in Thread | Next Message in Thread

From Wear-Hard Mailing list Archive (WH)
Maintained by R. Paul McCarty

Archive created with babymail