# Bit-resolution decrease for internet

Discussion in 'General Electronics' started by Radium, Dec 3, 2003.

I would like to use an audio codec based on WAVE PCM. It should be a
little different though. The bit-resolution should be set to equal
1/(sampling rate X # of channels). The bit-rate should be set to equal
1 bit per second. I would like to use this codec to transport audio
files though the internet via email.

I am looking for frequency response. In digital audio the sampling
rate must be at least twice the highest frequency in the signal. It
would like a highest frequency of at least 200 KHz. This would require
a sample rate of at least 400 KHz.

In this codec the bit-resolution is decreased to maintain a low bit
rate of 1 bit/sec. The bit-resolution is divided by the sampling rate
and the # of channels to acheive this.

2. ### Ben PopeGuest

You are clearly misguided.

Ben

3. ### Rich AndrewsGuest

1 bit per second? Wouldn't that equate to .5 hz or did I miss something?

r

Wrong. 1 Hz sampling rate would equate to .5 Hz. Sampling rate must be
at least 2x the maximum frequency.

If in a wave file, the bit-resolution is made to equal 1 /(sampling
rate X number of channels), then the bit-rate will definitely be
1-bit/second. If the sample rate is 44,100 Hz in a stereo (2-channel)
wave file of this type, the bit-resolution would be 1/(44100 x 2)-bit
or 1/88200-bit.

Bit-rate = sample-rate X bit-resolution X numbers of channels

Multiply the 44100 X 2 X 1/88200 and you get 1!

44100 Hz X 1/88200-bit X 2 channels = 1 bit per second

1 minute of this file would comsume 60 bits of disk space.

1 Hz sampling rate would equate to .5 Hz. 1-bit/sec, however would
not. Bit/time is the bit-rate. Sample rate is different from bit-rate.
It is also important to know the difference between *bit-resolution*
and *bit-rate*.

1 byte = 8 bits

If in a wave file, the bit-resolution is made to equal 1 /(sampling
rate X number of channels), then the bit-rate will definitely be
1-bit/second. If the sample rate is 44,100 Hz in a stereo (2-channel)
wave file of this type, the bit-resolution would be 1/(44100 x 2)-bit
or 1/88200-bit.

Bit-rate = sample-rate X bit-resolution X numbers of channels

Multiply the 44100 X 2 X 1/88200 and you get 1!

44100 Hz X 1/88200-bit X 2 channels = 1 bit per second

1 minute of this file would comsume only 60 bits of disk space. It
would definitely work for the internet. Unlike conventional MP3s and
WMAs, the high-frequency content of the PCM music will be restored due
to the high sample rate.

60 bits = 60/8 bytes

6. ### Clay S. TurnerGuest

At this rate, a 700MB CD will hold 177 years' worth of music!!
Why don't you compress some music to be just 1 bit per second and see if
anyone would be willing to listen to it more than once. I think you need to
check how you are handling your units. The Hz times 1/Hz units cancel out.

Clay

7. ### Ben PopeGuest

OK, so whats the maximum frequency you would like to capture? (More to the
point, what sampling rate do you propose)
Only if you could represent the samples in fractions of a bit.
And how do you propose to represent such fractional bits?
Fine, but you can't have less than one bit per sample.
Well done.
Yes, but even if it was possible, it would not really be classed as sound,
would it?

The reason the "bit resolution" is 16bits in CD Audio is because then you
can represent the varying wave with some degree of precision. If you took
that down to one bit, you would be turning sinusoids into square waves,
which would introduce just a tiny bit of odd-harmonic distortion. That
would result in bps = sampling rate * channels.

Ben

8. ### Bevan WeissGuest

Consider this logically for a second...
The lowest possible bit depth for a single sample is 1 bit, hence you must
have at least 1bit per sample.
If you want 400kHz sample rate then you must have a minimum of 400kbps data
rate.
1 minute of this data would consume 2400kb.

9. ### Paavo JumppanenGuest

Simple question: What is or how do you make a fractional bit? Digital
systems are quantized. One bit implies quantising to two levels On or
Off. What are the quantization levels of a fractional bit and how
would you represent it?

Regards,

Paavo Jumppanen
Author of AtSpec : A 2 channel PC based FFT spectrum analyzer
http://www.taquis.com

1 Hz sampling rate would equate to .5 Hz. 1-bit/sec, however would
not. Bit/time is the bit-rate. Sample rate is different from bit-rate.
It is also important to know the difference between *bit-resolution*
and *bit-rate*.

If in a wave file, the bit-resolution is made to equal 1 /(sampling
rate X number of channels), then the bit-rate will definitely be
1-bit/second. If the sample rate is 44,100 Hz in a stereo (2-channel)
wave file of this type, the bit-resolution would be 1/(44100 x 2)-bit
or 1/88200-bit.

Bit-rate = sample-rate X bit-resolution X numbers of channels

Multiply the 44100 X 2 X 1/88200 and you get 1!

44100 Hz X 1/88200-bit X 2 channels = 1 bit per second

1 minute of this file would comsume only 60 bits of disk space. It
would definitely work for the internet. Unlike conventional MP3s and
WMAs, the high-frequency content of the PCM music will be restored due
to the high sample rate.

From the responses to my above message I found it is impossible to
have less than 1-bit/sample. After the bit is the quantum of digital
info.

What about sampling rate? Is it possible to have less than 1
sample/bit?

Lets say a codec with:

Sample rate = 1/(bit resolution X number of channels)

Since:

bit-rate = sample rate X bit-resolution X number of channels,

The sample rate would cancel with the bit-resolution and # of
channels.

After all AB X 1/AB = 1

Plug in CD quality bit resolution (16-bit) and # of channels (2),

sample rate = 1/32 = 0.03125 Hz

I know this codec would be impractical but is it possible?

I'm just in it for the science. No application.

12. ### Robert BaerGuest

In the very remote past, "Ug" took a sample of the La Brea tar pits,
and died in the muck.
He never took another sample, and that was thousands of years ago.
How is *that* for a sample rate???

Actually, in real life, photographers have used one sample per day
with a movie camera, to show the growth of a plant.
Also, a sample rate of once every 15 minutes (i am guessing) to show a
flower opening in the morning, turning to follow the sun, and closing at
night.
Another real case: to show cloud movement over a day or many days.

13. ### Clay S. TurnerGuest

At this rate, a 700MB CD will hold 177 years' worth of music!!
Why don't you compress some music to be just 1 bit per second and see if
anyone would be willing to listen to it more than once. I think you need to
check how you are handling your units. The Hz times 1/Hz units cancel out.

Clay

14. ### Ben PopeGuest

OK, so whats the maximum frequency you would like to capture? (More to the
point, what sampling rate do you propose)
Only if you could represent the samples in fractions of a bit.
And how do you propose to represent such fractional bits?
Fine, but you can't have less than one bit per sample.
Well done.
Yes, but even if it was possible, it would not really be classed as sound,
would it?

The reason the "bit resolution" is 16bits in CD Audio is because then you
can represent the varying wave with some degree of precision. If you took
that down to one bit, you would be turning sinusoids into square waves,
which would introduce just a tiny bit of odd-harmonic distortion. That
would result in bps = sampling rate * channels.

Ben

15. ### Bevan WeissGuest

Consider this logically for a second...
The lowest possible bit depth for a single sample is 1 bit, hence you must
have at least 1bit per sample.
If you want 400kHz sample rate then you must have a minimum of 400kbps data
rate.
1 minute of this data would consume 2400kb.

16. ### Paavo JumppanenGuest

Simple question: What is or how do you make a fractional bit? Digital
systems are quantized. One bit implies quantising to two levels On or
Off. What are the quantization levels of a fractional bit and how
would you represent it?

Regards,

Paavo Jumppanen
Author of AtSpec : A 2 channel PC based FFT spectrum analyzer
http://www.taquis.com

17. ### Bevan WeissGuest

Ask any audiofile and they'll say that the sound is much better with a
higher sample rate and bit resolution....
Clearly the answer is to reduce the number of channels of audio.

If you have the number of channels as
num channels = bit-rate / (bit-resolution X sample-rate)
then you can still have a 1bit per second for the bit rate, but you can now
have a huge sample rate and an awesome bit resolution. I mean you could
easily get 192bits per sample, and some 6MSPS with this system. You would
just have to only use 868.056x10^(-12) channels.

I'm pretty sure that everyone else in this newsgroup will agree with me that
this is more the approach that you should utilise radium...

lol

18. ### Randy YatesGuest

construct a CODEC that will blow the socks off WMA, MP3, etc. and show us
all up.
--
%% Fuquay-Varina, NC % 'cause no one knows which side
%%% 919-577-9882 % the coin will fall."
%%%% <> % 'Big Wheels', *Out of the Blue*, ELO

19. ### Rich AndrewsGuest

Methinks something is wrong with the math and/or definitions.

The way I understand it, if one samples at 44.1kc that means that 44
thousand times per second you have a 16 bit word. Thus, in one second,
you have 44,100 16 bit words of data. Put another way, in one second you
have 44,100*16*2(for stereo) or 1,411,200 bits of data.

1,411,200 / 8 = 176400 8 bit bytes per second.

Now if you are wanting to change the sampling resolution or depth from 16
bits to something like 8 or even 4, the result would not be very good. I
believe that there are examples of 4 amd 8 bit on the net.

r

20. ### Dick PierceGuest

Kind people, stand up and give Radium a hearty Huzzah!, for he
has volunteered for the role of Village Idiot, a job for which
has has proven himself truly qualified.