Should Compressed Hi Def Sound Better?, 96kHz,24bit in 128kbps x 44.1kHz,16bit in 128kbps Options

[fcmts]

Would a 128kbps file (A'), compressed from a 96kHz,24bit sample (A), that also plays in 96kHz,24bit sound more accurate than a 128kbps file (B'), compressed from a 44.1kHz,16bit sample (B), that also plays in 44.1kHz,16bit ?
I'm assuming that A and B are the same, except for the above specs and that both compressions have the same algorithm.

[jmartis]

If both use the same encoding algorithm, and both preserve all frequencies, then the one compressed from 96k 24bit will sound worse, because it will encode more frequencies we cannot hear, leaving less bandwidth to audible ones.

J.M.

This post has been edited by jmartis: May 29 2006, 06:56

[Axon]

And if they have the exact same compression algorithm, they will give equal weight to supersonic frequencies - none. So they would decode to the same quality.

[Firon]

At 128kbps it would be lowpassed, so the supersonic content just plain wouldn't be there.

[menno]

And if they have the exact same compression algorithm, they will give equal weight to supersonic frequencies - none. So they would decode to the same quality.

Don't forget that you still have to encode more than 2 times more samples, the overhead will be much higher. The frequency resolution is also different. So no, the quality will not be the same.

[kjoonlee]

Is 96 kHz MP3 possible at all? Without resampling, that is.

[menno]

No, but the TS didn't mention MP3

[Firon]

Theoretically, the quality from a 96KHz file encoded at 128kbps would be lower than one made at 44KHz, wouldn't it?

[Mercurio]

Please can someone explain this to me? Does Hi-Def audio really contain supersonic frequencies? Do they use mics that can record them?

I thought 96Khz were for noise-shaping and easy lowpass antialiasing filter.

[Firon]

Generally they do contain content >22KHz. Some are brickwall filtered below the frequencies the sampling rate can support but still higher than 22KHz, and others yet are reportedly resampled versions of the CDDA version.
They almost certainly use mics that can record supersonic frequencies, for the content that actually has supersonic frequencies (and not just watermarks in that range).

This post has been edited by Firon: May 29 2006, 10:04

[menno]

Theoretically, the quality from a 96KHz file encoded at 128kbps would be lower than one made at 44KHz, wouldn't it?

In general I would say yes, but the increased time resolution might be a benefit in some cases.

[fcmts]

Theoretically, the quality from a 96KHz file encoded at 128kbps would be lower than one made at 44KHz, wouldn't it?

In general I would say yes, but the increased time resolution might be a benefit in some cases.

That's not reasonable. It's the same as stating that human statistics made from fewer people would be more accurate than other made from a large sample of a population. Isn't it?

[menno]

I don't really see the connection with statistics. I'm stating that I think that the file with the lower samplerate will sound better, but, also depending on the codec, an increased time resolution might be of benefit. Since I never tested this, I can't see if this is really true for any case.

[GeSomeone]

It's the same as stating that human statistics made from fewer people would be more accurate than other made from a large sample of a population. Isn't it?

Following this (flawed) reasoning, you would gather the same amount of data from the small group as from the large group, so less info from each. But as Menno says, you can't tell the effect on quality, as (psycho-acoustic) sound compression is not statistics.

[saratoga]

Theoretically, the quality from a 96KHz file encoded at 128kbps would be lower than one made at 44KHz, wouldn't it?

In general I would say yes, but the increased time resolution might be a benefit in some cases.

That's not reasonable. It's the same as stating that human statistics made from fewer people would be more accurate than other made from a large sample of a population. Isn't it?

YOu're assuming all of the population is equally interesting. In reality, we don't really care about information above 20k, particularly not at 128k. So while theres more information, its probably not useful. In your analogy, it would be like having a 2x larger sample size, but then half of the responses are blank.

[Firon]

The lowpass would probably be lower, maybe 16KHz, at 128k.