Linear PCM audio: 44.1 KHz sample-rate, monaural, 22,050-bits-per-second

[Radium[_4_]]

On Jul 20, 6:08 pm, Oli Charlesworth > wrote
http://groups.google.com/group/comp.dsp/msg/0eef9d996e79f473?hl=en& :

> This makes no sense.
> How can
> you describe 44100
> samples with a single
> bit?

Well, one-bit-per-44,100-cycles is what I was originally looking for.
However, as some posters have stated, all I would hear in 1-bit-per-
cycle would resemble a square wave "tick tock". New stuff is learned
everyday.

So how about decreasing the amount of bits-per-cycle so that the bit-
rate becomes 20,000-bits-per-second? After all the human auditory
system perceives up to 20 KHz so covering the entire human audio
frequency range would require at least 20-kilobits-per-second.

In a sample rate of 44,100-cycles-per-second, this would best be done
at 1-bit-every-2-cycles. This would give a bit-rate of 22,050-bits-per-
second. That's obviously above 20kbits-per-second but only slightly.

Couldn't the bit-rate be less than the sample-rate if some information
in each sample is thrown away? Could this data-reduction be done
linearly?

With each 44,100-cycles-per-second, discard half the information, and
you get 22,050-bits-per-second. Discard 1 bit for every two cycles. I
could be incorrect though. If so, please assist me.

To all:

I have a neurological disability called Asperger's Syndrome.

I would like to give you some information about my disability. The
reason I am posting this message about Asperger's is to help avoid any
potential misunderstandings [though it's probably too late].

I have been diagnosed with Asperger's Syndrome (AS). AS is a
neurological condition that causes significant impairment in social
interactions. People with AS see the world differently and this can
often bring them in conflict with conventional ways of thinking. They
have difficulty in reading body language, and interpreting subtle
cues. In my situation, I have significant difficulty with natural
conversation, reading social cues, and maintaining eye contact. This
can lead to a great deal of misunderstanding about my intent or my
behavior. For example, I may not always know what to say in social
situations, so I may look away or may not say anything. I also may not
always respond quickly when asked direct questions, but if given time
I am able express my ideas.

On Usenet, the text-equivalent of my disability is probably noticed. I
do apologize profusely, for any inconvenience it causes.

Thank you very much in advance for your understanding, cooperation,
and assistance.


Regards,

Radium

[Jordan]

Radium wrote:
>
> So how about decreasing the amount of bits-per-cycle so that the bit-
> rate becomes 20,000-bits-per-second? After all the human auditory
> system perceives up to 20 KHz so covering the entire human audio
> frequency range would require at least 20-kilobits-per-second.
>

Hi Radium

You need double the sampling rate to reproduce an analogue waveform,
according to Nyqvist theorem.

[Steven Pigeon]

Radium wrote:
> On Jul 20, 6:08 pm, Oli Charlesworth > wrote
> http://groups.google.com/group/comp.dsp/msg/0eef9d996e79f473?hl=en& :
>
>> This makes no sense.
>> How can
>> you describe 44100
>> samples with a single
>> bit?
>
> Well, one-bit-per-44,100-cycles is what I was originally looking for.
> However, as some posters have stated, all I would hear in 1-bit-per-
> cycle would resemble a square wave "tick tock".

Not necessarily. In modern signal processing, signals are decomposed
into components by an analysis phase, say, by Fourier Transforms (while
more often in its special case of the Discrete Cosine Transform, for a
lot of good theoretical and practical reasons) or by other means like
wavelet transforms. The signal is then separated into its components
and it is the information about the components that is coded.

A complex sound wave might be the result of adding two basic frequencies
with given amplitudes. One would then code the frequency and amplitude
of these two frequencies to represent the signal. If the description is
"I have frequency F1 with amplitude A1 combined with F2 with A2, over
1 second", the number of bits per sample is very, very, low, achieving
a high compression ratio.

In the real world, however, sound waves are the results of a great
number of frequency/amplitude and they are not totally periodic.
Algorithms will decompose the signal into the component frequencies,
code each to a certain precision, thus yielding an acceptable
reconstruction at decompression. The efficiency of the compression
therefore depends on how you decompose the signal and how smart you are
about selectively destroying precision so that the signal is compressed
to a certain amount of bits.


Best,

S.

[Phil Carmody]

Radium > writes:
> On Jul 20, 6:08 pm, Oli Charlesworth > wrote
> http://groups.google.com/group/comp.dsp/msg/0eef9d996e79f473?hl=en& :
>
> > This makes no sense.
> > How can
> > you describe 44100
> > samples with a single
> > bit?
>
> Well, one-bit-per-44,100-cycles is what I was originally looking for.
> However, as some posters have stated, all I would hear in 1-bit-per-
> cycle would resemble a square wave "tick tock". New stuff is learned
> everyday.

Nonsense. You can digitise anything, such as the human voice
using a 1-bit signal wave. I forget the name of the technique
now, something like delta modulation. I did it back on my ZX
Spectrum in the early 80s. Basically, if the gradient of the
signal is negative output a 0, and if it's positive output a
1. It's crummy, but it's recognisable.

The only reason you'd get a 'tick-tock' is if you were trying
to digitise something that vaguely resembled a 'tick-tock' sound.

Phil
--
"Home taping is killing big business profits. We left this side blank
so you can help." -- Dead Kennedys, written upon the B-side of tapes of
/In God We Trust, Inc./.

[Don Pearce]

On 23 Jul 2007 12:51:07 +0300, Phil Carmody
> wrote:

>Nonsense. You can digitise anything, such as the human voice
>using a 1-bit signal wave. I forget the name of the technique
>now, something like delta modulation. I did it back on my ZX
>Spectrum in the early 80s. Basically, if the gradient of the
>signal is negative output a 0, and if it's positive output a
>1. It's crummy, but it's recognisable.

Sort of right. What happens is this. The 1/0 digital signal is used to
charge or discharge a capacitor. At each time tick, the voltage on
that capacitor is compared to the level of the audio signal. If it is
too low, a 1 will charge it up a bit, and if it is too high a 0 will
drain a little charge from it, keeping it matched exactly to the
audio. The resulting stream of 1s and 0s describes the audio signal.

d

--
Pearce Consulting
http://www.pearce.uk.com

[Pete Fraser]

"Phil Carmody" > wrote in message
...

>
> Nonsense. You can digitise anything, such as the human voice
> using a 1-bit signal wave. I forget the name of the technique
> now, something like delta modulation. I did it back on my ZX
> Spectrum in the early 80s. Basically, if the gradient of the
> signal is negative output a 0, and if it's positive output a
> 1. It's crummy, but it's recognisable.

It's only crummy if you don't sample fast enough.
Many audio A/D converters work like this.

[Karl Uppiano]

"Pete Fraser" > wrote in message
...
> "Phil Carmody" > wrote in message
> ...
>
>>
>> Nonsense. You can digitise anything, such as the human voice
>> using a 1-bit signal wave. I forget the name of the technique
>> now, something like delta modulation. I did it back on my ZX
>> Spectrum in the early 80s. Basically, if the gradient of the
>> signal is negative output a 0, and if it's positive output a
>> 1. It's crummy, but it's recognisable.
>
> It's only crummy if you don't sample fast enough.
> Many audio A/D converters work like this.

Practically all of the current Sigma/Delta (or Delta/Sigma) converters do
this. They are sometimes called MASH converters or 1-bit converters. If the
1-bit converter sample fast enough (in the MHz range) that resolution is not
lost at the maximum slew-rate (which can be calculated from the slope of the
desired maximum input frequency at full scale amplitude), then it is
theoretically capable of arbitrarily high resolution (i.e., quality). The
output of this converter is then numerically transformed into a 16-bit or
24-bit PCM datastream at the selected sample rate (e.g., 44.1KHz or
whatever).

A well-made Sigma/Delta converter can be made more reliable than the older
successive approximation register (SAR) converters, because of the
difficulty of trimming SAR resistors or capacitors to the exacting precision
needed for 16 or 24 bit audio (all 24 voltage references must be accurate to
about 3 millionths of a percent for 24 bit accuracy to 1/2 LSB - impossible,
even if hand trimmed - and it wouldn't stay that way for long!). A 1-bit
converter, on the other hand, only requires one coarsely-trimmed voltage
reference, and accurate timing, which is much easier to achieve with careful
circuit layout.

http://www.maxim-ic.com/appnotes.cfm/appnote_number/1870

[Jasen Betts]

On 2007-07-23, Phil Carmody > wrote:
> Radium > writes:
>> On Jul 20, 6:08 pm, Oli Charlesworth > wrote
>> http://groups.google.com/group/comp.dsp/msg/0eef9d996e79f473?hl=en& :
>>
>> > This makes no sense.
>> > How can
>> > you describe 44100
>> > samples with a single
>> > bit?
>>
>> Well, one-bit-per-44,100-cycles is what I was originally looking for.
>> However, as some posters have stated, all I would hear in 1-bit-per-
>> cycle would resemble a square wave "tick tock". New stuff is learned
>> everyday.
>
> Nonsense. You can digitise anything, such as the human voice
> using a 1-bit signal wave. I forget the name of the technique
> now,

delta-sigma.

> something like delta modulation. I did it back on my ZX
> Spectrum in the early 80s. Basically, if the gradient of the
> signal is negative output a 0, and if it's positive output a
> 1. It's crummy, but it's recognisable.

I recall hearing speech from the games "Ghost Busters" and "Freedom
Fighter"

In the 90s modplay would overlay wav files an play them out the PC
speaker (which is also one-bit).

Bye.
Jasen

[Willem]

Jasen wrote:
) On 2007-07-23, Phil Carmody > wrote:
)> Nonsense. You can digitise anything, such as the human voice
)> using a 1-bit signal wave. I forget the name of the technique
)> now,
)
) delta-sigma.
)
)> something like delta modulation. I did it back on my ZX
)> Spectrum in the early 80s. Basically, if the gradient of the
)> signal is negative output a 0, and if it's positive output a
)> 1. It's crummy, but it's recognisable.
)
) I recall hearing speech from the games "Ghost Busters" and "Freedom
) Fighter"
)
) In the 90s modplay would overlay wav files an play them out the PC
) speaker (which is also one-bit).

But that's a different technique. You can program the PC speaker to putput
a square wave with a programmed phase width and a high frequency, and then
vary the phase width you get different samples.


SaSW, Willem
--
Disclaimer: I am in no way responsible for any of the statements
made in the above text. For all I know I might be
drugged or something..
No I'm not paranoid. You all think I'm paranoid, don't you !
#EOT

[[email protected]]

On Jul 24, 9:01 am, Willem > wrote:

>You can program the PC speaker to putput

LOL. Willem, I wanna 'putput' you into a jar, display it to the masses
and say "This, folks, is why we can't have nice things!"