mp3's : 44.1KHz or 48KHz ?

[Vulcan]

hey all...

say ia have an audio source from a DVD , meaning a 5.1 dolby digital (ac3 , sample rate of 48KHz) track and i want to convert it to a stereo mp3 file with 192kbps.

should the mp3 also be 48KHz or 44.1KHz??

i'm asking because i'm confused :
48KHz means more compliance to nyquist's rule and there for easier filtration, but on the other hand, since it is the same bitrate (192kbps) , it will be devided to more samples (compared to 44.1KHz) meaning less bits for sample or less accuracy to a sample.

what say you...?

[setarip]

The only thing that counts is your ear - so why not simply use the same source to create one at a frequency of 48,000Hz and another at 44,100Hz, and decide which sounds better to you...

[Vulcan]

i would have done that , if i had good ears...

as it seams..i'm not as sensitive to artifacts in sound as my family and friends...

so i need a scientific solution to my question..

[setarip]

The higher the sampling frequency, the higher the quality...

[Enchanter]

Vulcan,

Keep the original sampling rate of 48KHz, for the following reasons:

1. 192Kbps is more than enough for 48KHz.

2. Downsampling (to 44.1KHz) will logically reduce quality.

3. Most computers use AC97-based chips for their soundcards (for example, the Turtle Beach Santa Cruz, all AC97 on-board soundcards and all Creative Labs Soundblaster products), which means that all sound inputs having sampling rates lower than 48KHz will be upsampled to the said sampling rate. There is a catch; None will upsample perfectly without causing some upsampling artifacts. Hence it is best to prevent this upsampling scenario, that is to keep the original 48KHz sampling rate.


Don't consider upsampling to higher frequencies (>48KHz) either, since it brings no audible benefits and, as said above, will cause upsampling artifacts. I hope that is scientific enough for you.

[setarip]

To Vulcan

You might find the following scientific paper interesting - especially the brief section on "Frequency):

http://www.csuchico.edu/~nwylde/pdfs/DigitalAudio.pdf

[Enchanter]

What is so interesting in the frequency section, setarip?

If you are just trying to stress the use of 'frequency' rather than 'sampling rate', I can tell you that the latter is used far more prominently in other forums and sites; in fact, I find intermixing the use of 'frequency of sound' with 'frequency rates of sound' a tad confusing. Frequency and sampling rates would be better digested by most people.

Regards.

[setarip]

To Enchanter

If YOU must know (or, if you've changed your name to Vulcan) it's the reference to human hearing range...


("If you are just trying to stress the use of 'frequency' rather than 'sampling rate'"

- And you were all prepared to "get your panties in a bunch" [means to get upset] for no reason ;>})

[Enchanter]

setarip,

"it's the reference to human hearing range"

That is an obvious fact. Hence what is so interesting about the "frequency section" that you had to recommend reading it? Afraid the original poster does not understand its meaning (of which he has not shown the lack of it)?


"And you were all prepared to "get your panties in a bunch" [means to get upset] for no reason ;>}"

I was wanting for a discussion and you answer me with this ridiculous statement? You make me laugh.

[Vulcan]

trust me....the author does understand a little about the sample theorem...

Enchanter..according to what you say, when i play an audio CD the 44.1KHz sample rate is upconverted to 48KHz and then it is played ???

anyway...setting the mp3 to 192kbps means a constant bitrate no matter the sampling frequency.
the bits in an mp3 file are mostly used to set the order and the mantissa of each frequency.higher sample rates with the same bitrate forces fewer bits for the order and the mantissa, which as i see it means less accuracy and poorer quality mp3 than the 44.1KHz equivalent.

your thoughts...please.

[Enchanter]

Vulcan,

"according to what you say, when i play an audio CD the 44.1KHz sample rate is upconverted to 48KHz and then it is played ???"

Upsampling of non-48KHz audio source on AC97-chips will occur on almost all sound formats, including Wave, MP3, OGG and such. On Audio CDs (that is playing straight off the CD) the same should apply as well, at least with digital playback (CD -> IDE cable for digital transfer -> DAC on soundcard -> Speakers). On non-digital playback (CD -> DAC on CD drive -> CD Audio cable from drive to soundcard -> Speakers), I cannot say for sure, but I would say there is no resampling since the soundcard is only acting as a pass-through to the speakers (i.e. no further processing occurs on the soundcard chip, hence no resampling).

Now before you start thinking that non-digital CD playback is better, consider this. Most DACs found on typical CD drives are of the cheap and lousy type and you would be better served by the (generally) better DACs found on most soundcards, resampling or not.


"setting the mp3 to 192kbps means a constant bitrate no matter the sampling frequency.
the bits in an mp3 file are mostly used to set the order and the mantissa of each frequency.higher sample rates with the same bitrate forces fewer bits for the order and the mantissa, which as i see it means less accuracy and poorer quality mp3 than the 44.1KHz equivalent."

As I previously put it, 192kbps is sufficient for all but the most demanding tracks. 192kbps is probably just enough for detailed music tracks (where the audience, you, has the attention completely on the aural nuances of the track). In movie tracks however, the audience's attention will be split between the video and audio, and for the audio, it is usually the vocals (do you need to hear the emotion of the actors with every word? Not in the movies!) and explosions (just simple kabooms. What details do you need?) that are paid most attention to. Whether they turn out any worse than the original or not, the audience will very unlikely detect it in the exhilarations of watching the movie.

As for the virtues of downsampling, unless you have a set of revealing and detailed studio monitors (multimedia and hi-fi speakers do not count) for your computer, I doubt you can hear a difference (downsampling effects aside). In fact, I am personally against this, especially if you are using an AC97-compliant soundcard (and most multimedia PC users do!). Downsampling results in reduced quality; upsampling from the soundcard results in a degree of artifacts. Both combined together result in an audio track that will undoubtedly sound worse than the original, non-resampled (unadulterated) track, don't you agree?

Cheers!

Edit: mis-types.

[Vulcan]

i get you point...

if i do the downsampling i reduce quality twice.
first the removal of samples , and second the upsampling of the soundcard.

and if 192kbps is enough for a 48KHz sample rate, there is no point in downsampling.

thanks a lot....especially for the detailed answer...!!

[Enchanter]

You're welcome.

[benderman]

Even if you use 44.1 kHz it would make no noticable difference. MP3s psychoacustic model cuts off most frequencies higher than 16 kHz which would not need more than a samplerate of 32 kHz.
In fact most people are not able to notice sounds with a frequency above 16 kHz.

[Enchanter]

benderman,

"MP3s psychoacustic model cuts off most frequencies higher than 16 kHz which would not need more than a samplerate of 32 kHz"

This applies for FhG and a lot of older MP3 compression algorithms only. LAME does not adhere to this 16KHz roll-off rule and, in fact, automatically determines the high frequency roll-off value according to the bitrate used, or to whatever custom (manually input) value is desired.

On 44.1KHz vs 48KHz, I merely suggested 48KHz as this is the original sampling rate of the source. I see no point in downsampling (thus reducing quality, audible or not), especially in light of the fact that a compression bitrate of 192kbps will be more than sufficient for a movie soundtrack.


"In fact most people are not able to notice sounds with a frequency above 16 kHz"

Let me go a bit off-topic here.

For movie soundtracks, I completely agree. However, on more complex music soundtracks, having sounds above the 16KHz frequency range will add (much) details to the music. It's true that many people cannot hear anything above 16KHz, some can do better until 18KHz, and even far fewer can hear somewhere beyond the normally accepted human hearing limit of 20KHz. Regardless of whether one can hear the HF sounds, it is generally accepted that musics that include sounds beyond the 16KHz sound much fuller or dynamic or detailed (however you can describe better sounds with words), compared to musics that don't. Feeling the music becomes an altogether different thing than hearing the music.

Cheers!