Reply
 
Thread Tools Display Modes
  #1   Report Post  
Posted to rec.audio.tech
 
Posts: n/a
Default Interesting technical question

Let's say you have an mp3 file at a bit rate of 128kb/s and a sample rate of
44.1k samples/second, and you want to cut the size of the file in half. You
have 2 choices: 64kb/s at the same sample rate, or throwing away half the
samples, to 22.05k samples/second. Which choice would the average person
prefer? Assume normal hearing.

Thanks,

Norm Strong


  #2   Report Post  
Posted to rec.audio.tech
Richard Crowley Richard Crowley is offline
external usenet poster
 
Posts: 4,172
Default Interesting technical question

wrote ...
Let's say you have an mp3 file at a bit rate of 128kb/s and a sample rate
of 44.1k samples/second, and you want to cut the size of the file in half.
You have 2 choices: 64kb/s at the same sample rate, or throwing away half
the samples, to 22.05k samples/second. Which choice would the average
person prefer? Assume normal hearing.


Do you think there is a single answer for all types of content
(music, speech, etc.) and for all situations? If there were,
why would we need separate control over all those parameters
(and even others)?


  #3   Report Post  
Posted to rec.audio.tech
Todd H. Todd H. is offline
external usenet poster
 
Posts: 116
Default Interesting technical question

writes:

Let's say you have an mp3 file at a bit rate of 128kb/s and a sample rate of
44.1k samples/second, and you want to cut the size of the file in half. You
have 2 choices: 64kb/s at the same sample rate, or throwing away half the
samples, to 22.05k samples/second.


But it's not like that, actually. You appear to be applying linear
PCM thinking here to a concept that involves lossy compression.
Your question isn't terribly well constained so it's difficult to
answer.

One thing worth noting though, if you reduce the front end sample rate
to 22k, Nyquist demands that you have to say goodbye to anthing near
11kHz or above straight away by low pass filtering the source.
Listeners will have a big problem with that if we're talking about a
general musical source, for instance.

Also, dropping samples isn't a good way to do it. If you just
casually throw away half the samples yet your analog input wasn't low
pass filtered adequate to ensure the Nyquist criteria was met, you are
nearly guaranteed to be introducing new frequencies into your sampled
material through aliasing. Listeners hate that too.

There are lots of ways to get a file size in half. Choice of codec,
parameters provided to the codec, variable bit rate techniques,
encoding mono vs stereo... No one size fits all surely.

What problem are you trying to solve?

Best Regards,
--
/"\ ASCII Ribbon Campaign | Todd H
\ / | http://www.toddh.net/
X Promoting good netiquette | http://triplethreatband.com/
/ \ http://www.toddh.net/netiquette/ | http://myspace.com/mytriplethreatband
  #4   Report Post  
Posted to rec.audio.tech
Schöön Martin Schöön Martin is offline
external usenet poster
 
Posts: 9
Default Interesting technical question

writes:

Let's say you have an mp3 file at a bit rate of 128kb/s and a sample rate of
44.1k samples/second, and you want to cut the size of the file in half. You
have 2 choices: 64kb/s at the same sample rate, or throwing away half the
samples, to 22.05k samples/second. Which choice would the average person
prefer? Assume normal hearing.

Thanks,

Norm Strong


Stereo - Mono

--
Martin Schöön

"Problems worthy of attack
prove their worth by hitting back"
Piet Hein
  #5   Report Post  
Posted to rec.audio.tech
Richard Crowley Richard Crowley is offline
external usenet poster
 
Posts: 4,172
Default Interesting technical question

""Schöön Martin"" wrote ...
writes:

Let's say you have an mp3 file at a bit rate of 128kb/s and a sample rate
of
44.1k samples/second, and you want to cut the size of the file in half.
You
have 2 choices: 64kb/s at the same sample rate, or throwing away half
the
samples, to 22.05k samples/second. Which choice would the average person
prefer? Assume normal hearing.

Thanks,

Norm Strong


Stereo - Mono


Brilliant! :-)
I've done it myself. Especially for speech.




  #6   Report Post  
Posted to rec.audio.tech
Ian Bell Ian Bell is offline
external usenet poster
 
Posts: 442
Default Interesting technical question

Richard Crowley wrote:

""Schöön Martin"" wrote ...
writes:

Let's say you have an mp3 file at a bit rate of 128kb/s and a sample
rate of
44.1k samples/second, and you want to cut the size of the file in half.
You
have 2 choices: 64kb/s at the same sample rate, or throwing away half
the
samples, to 22.05k samples/second. Which choice would the average
person
prefer? Assume normal hearing.

Thanks,

Norm Strong


Stereo - Mono


Brilliant! :-)
I've done it myself. Especially for speech.


In practice, mp3 stereo files are mainly mono (sum of L and R) and a small
amount of difference (L -R), so switching to mono will typically only give
a 25% reduction in file size.

Ian
  #7   Report Post  
Posted to rec.audio.tech
 
Posts: n/a
Default Interesting technical question


"Todd H." wrote in message
...
writes:

Let's say you have an mp3 file at a bit rate of 128kb/s and a sample rate
of
44.1k samples/second, and you want to cut the size of the file in half.
You
have 2 choices: 64kb/s at the same sample rate, or throwing away half
the
samples, to 22.05k samples/second.


But it's not like that, actually. You appear to be applying linear
PCM thinking here to a concept that involves lossy compression.
Your question isn't terribly well constained so it's difficult to
answer.

One thing worth noting though, if you reduce the front end sample rate
to 22k, Nyquist demands that you have to say goodbye to anthing near
11kHz or above straight away by low pass filtering the source.
Listeners will have a big problem with that if we're talking about a
general musical source, for instance.

Also, dropping samples isn't a good way to do it. If you just
casually throw away half the samples yet your analog input wasn't low
pass filtered adequate to ensure the Nyquist criteria was met, you are
nearly guaranteed to be introducing new frequencies into your sampled
material through aliasing. Listeners hate that too.

There are lots of ways to get a file size in half. Choice of codec,
parameters provided to the codec, variable bit rate techniques,
encoding mono vs stereo... No one size fits all surely.

What problem are you trying to solve?


The file already exists. 128kb/s, 44.1k. The question is how to cut this
existing file in half with the least effect on sound quality. Going from
stereo to mono would constitute a big sacrifice. Cutting the upper
frequency response to 10kHz would be much less bothersome. But is there a
better scheme?

Another question: If I convert the mp3 above to a wav file, and then
convert it right back to the same mp3 I started with, will the file undergo
any deterioration?

Norm


  #8   Report Post  
Posted to rec.audio.tech
Arny Krueger Arny Krueger is offline
external usenet poster
 
Posts: 17,262
Default Interesting technical question

wrote in message


The file already exists. 128kb/s, 44.1k. The question
is how to cut this existing file in half with the least
effect on sound quality. Going from stereo to mono would
constitute a big sacrifice.


IME, not really.

Cutting the upper frequency
response to 10kHz would be much less bothersome.


Most MP3 encoders implement a soft high frequency cutoff as the bitrate goes
down. Dropping the sample rate to 22 KHz is obviously more like a brick
wall.

But is there a better scheme?


Don't use MP3. Other formats do better at lower bitrates.

Another question: If I convert the mp3 above to a wav
file, and then convert it right back to the same mp3 I
started with, will the file undergo any deterioration?


Yes, but probably not as bad as the drops to 22 KHz sampling or mono.


  #9   Report Post  
Posted to rec.audio.tech
Barry Mann Barry Mann is offline
external usenet poster
 
Posts: 41
Default Interesting technical question

In , on 02/06/07
at 10:26 AM, said:




[ ... ]
Another question: If I convert the mp3 above to a wav file, and then
convert it right back to the same mp3 I started with, will the file
undergo any deterioration?


Norm


Try it! It's not expensive in terms of time or $$, and it's a much
better lesson than reading posts from a bunch of grumpy know-it-alls.

Every time a WAVE file goes through the MP3 CODEC, things will be
thrown away and the file will get smaller. The first trip through for
the original file will probably yield the best compression on a
percentage of the original file size basis.

Your tolerance to the audio degradation will tell you when to stop.

In general I think that you will find that the best result will be from
establishing the compression ratio at the outset rather than fuss with
a chain of MP3-WAV-MP3 ..., but this is something that you should play
with.

A a general rule, if the original compression scheme was any good (and
appropriate for the type of data involved), compressing a compressed
file does not result in much, if any additional space saving.

-----------------------------------------------------------
spam:
wordgame:123(abc):14 9 20 5 2 9 18 4 at 22 15 9 3 5 14 5 20 dot 3 15
13 (Barry Mann)
[sorry about the puzzle, spammers are ruining my mailbox]
-----------------------------------------------------------

  #10   Report Post  
Posted to rec.audio.tech
Schöön Martin Schöön Martin is offline
external usenet poster
 
Posts: 9
Default Interesting technical question

Ian Bell writes:

Richard Crowley wrote:

""Schöön Martin"" wrote ...

Stereo - Mono


Brilliant! :-)
I've done it myself. Especially for speech.


In practice, mp3 stereo files are mainly mono (sum of L and R) and a small
amount of difference (L -R), so switching to mono will typically only give
a 25% reduction in file size.

Not too bad in other words.

--
Martin Schöön

"Problems worthy of attack
prove their worth by hitting back"
Piet Hein


  #11   Report Post  
Posted to rec.audio.tech
Karl Uppiano Karl Uppiano is offline
external usenet poster
 
Posts: 232
Default Interesting technical question


wrote in message
. ..
Let's say you have an mp3 file at a bit rate of 128kb/s and a sample rate
of 44.1k samples/second, and you want to cut the size of the file in half.
You have 2 choices: 64kb/s at the same sample rate, or throwing away half
the samples, to 22.05k samples/second. Which choice would the average
person prefer? Assume normal hearing.


The L+R / L-R MP3 encoding automatically switches to mono if there is no
stereo content. If you don't want the directional cues, you can save some
space by switching to true mono. However, switching to mono or dropping the
sample rate will produce obvious perceptual losses. Switching to a lower
encoded bitrate will produce more subtle perceptual losses. By choosing
mono, or lowering the sample rate, you choose a fixed -- and non-negotiable
loss in quality. By lowering the encoded bitrate, you let the MP3 algorithm
make the tradeoffs adaptively, in real-time.


  #12   Report Post  
Posted to rec.audio.tech
 
Posts: n/a
Default Interesting technical question

Arny Krueger wrote:

The file already exists. 128kb/s, 44.1k. The question
is how to cut this existing file in half with the least
effect on sound quality. Going from stereo to mono would
constitute a big sacrifice.


IME, not really.

Cutting the upper frequency
response to 10kHz would be much less bothersome.


Most MP3 encoders implement a soft high frequency cutoff as the bitrate goes
down. Dropping the sample rate to 22 KHz is obviously more like a brick
wall.


The perceptual coder doesn't assign the same importance to the octave
between 11 kHz and 22 kHz as it does to the range between 0 Hz and 11 kHz,
anyway, so dropping the sample rate to 22 kHz wouldn't produce the desired
halving of file size.


Francois.

  #13   Report Post  
Posted to rec.audio.tech
Arny Krueger Arny Krueger is offline
external usenet poster
 
Posts: 17,262
Default Interesting technical question

"(null)" wrote in message
news:1170880604.289039@smirk
Arny Krueger wrote:

The file already exists. 128kb/s, 44.1k. The question
is how to cut this existing file in half with the least
effect on sound quality. Going from stereo to mono
would constitute a big sacrifice.


IME, not really.

Cutting the upper frequency
response to 10kHz would be much less bothersome.


Most MP3 encoders implement a soft high frequency cutoff
as the bitrate goes down. Dropping the sample rate to 22
KHz is obviously more like a brick wall.


The perceptual coder doesn't assign the same importance
to the octave between 11 kHz and 22 kHz as it does to the
range between 0 Hz and 11 kHz, anyway, so dropping the
sample rate to 22 kHz wouldn't produce the desired
halving of file size.


Agreed. But, halving the sample rate should provide more size reduction than
leaving the decision up to the coder.


  #14   Report Post  
Posted to rec.audio.tech
Federico Federico is offline
external usenet poster
 
Posts: 378
Default Interesting technical question


Let's say you have an mp3 file at a bit rate of 128kb/s and a sample rate
of
44.1k samples/second, and you want to cut the size of the file in half.


Cut the song in half and keep the first part. :-)
For example I see that on myspace.com-music people tend to listen to the
first minute of the songs.
If it's for a marketing thing maybe you don't need the second half of the
audio track.
F.



  #15   Report Post  
Posted to rec.audio.tech
Peter Larsen Peter Larsen is offline
external usenet poster
 
Posts: 498
Default Interesting technical question

wrote:

Let's say you have an mp3 file at a bit rate of 128kb/s and a sample rate of
44.1k samples/second, and you want to cut the size of the file in half.


I wouldn't want to start there, let us take it from a 32 bit file.

You have 2 choices: 64kb/s at the same sample rate, or throwing
away half the samples, to 22.05k samples/second.


I would not go as low as 22.05 kHz sample rate, but I do tend to go to
32 kHz for mp3's because doing that gives the encoder more room and
seems to reduce splattyness, and I also go to 16 bit with suitable
dither, whatever that might be, even though I will set the decode to be
to 32-bit audio, simply because it tends to sound best like that and
because the playback device may be 16 bit only.

Which choice would the average person prefer?


I don't know, but 96 kbit/sec - or rather 48 kbit pr. second pr. channel
- is as low as I like to go, and I have upped my standard mp3 choice to
max available variable bandwidth setting at 48 kHz, otherwise as per
above.

As a direct reply to your question I submit that the optimum choice is
likely to be to go to 32 kHz sampling and to 96 kbit pr second stereo.

Assume normal hearing.


Then we have to forget about mpeg encoding, because all of that is like
trying to listen immediately after having been exposed to unsafe spl so
that the inner ear bones have decoupled, all the detail is missing and
fails to come out even if you turn it up to clipping. For normal hearing
.... don't use mp3.

The masking vs. threshold issues and in my understanding of this
temporal resolution issues are complex, and there is no reason to assume
that bad audio will be less noticeable to those with a speech
comprehension hearing issue.

Norm Strong



Regards

Peter Larsen
Reply
Thread Tools
Display Modes

Posting Rules

Smilies are On
[IMG] code is On
HTML code is Off


Similar Threads
Thread Thread Starter Forum Replies Last Post
I have a question Clete W. Audio Opinions 7 November 11th 04 12:33 PM
Question regarding Phantom Power Neil Pro Audio 110 September 27th 04 02:30 PM
Question regarding Phantom Power Neil Pro Audio 0 September 24th 04 06:44 PM
newbie question - aardvark q10 + external mixer? alex Pro Audio 1 August 14th 04 07:29 PM
How To Write A Technical Paper By Eddie Runner The Lizard Car Audio 7 July 20th 03 11:29 PM


All times are GMT +1. The time now is 04:16 PM.

Powered by: vBulletin
Copyright ©2000 - 2024, Jelsoft Enterprises Ltd.
Copyright ©2004-2024 AudioBanter.com.
The comments are property of their posters.
 

About Us

"It's about Audio and hi-fi"