As we know, AAC supports high sampling rates, for example 88.2kHz and 96kHz. What is the purpose of this? If I was to test the encoder for quality at this sampling rates, how would I do that? Is down-sampling a must since humans can't hear this high freqs? Or AAC is used by bats and whales...

Daniel

Eventhough people generally can't hear frequencies above 20-25 kHz, having higher sample rate than 44.1 kHz is supposed to have better temporal resolution which might lead to sound sounding better.

hello,
there seems to be a misunderstanding on your side: the samplingrate-frequency has nothing to do with the frequencies your ear is able to hear !!! for more info see :the wiki

Bye,
56W11

Ahem. There seems to be a misunderstanding on your side. The sampling rate is directly related to the frequencies that can be represented. Look up the Nyquist theorem.

And I don't think there is such a thing called "samplingrate-frequency".

Are we purposely misreading things here, PoisonDan?
Although 56W11 English sucks (my German is much much worse, in that I know approx 3 words ), he/she is actually speaking the truth (although not reacting to the correct question).. Nothing you can do to a sample will affect what your ears can hear

Focus: Its probably to get 'Magazine' Audiophiles interested in the format - they probably believe that their music contains 44khz energy, and although they cant hear it directly, the harmonics can make the sound better

Okay, 56W11's statement itself is correct, but I fail to see how it was relevant to dand's question (I also don't see the purpose of posting that wiki link).

Or maybe I interpreted dand's question wrongly too.

Or maybe I just need more sleep.

I agree that the samplingrate has nothing to do with psychoacoustics but is related to the future trend in audio electronics.. Higher samplingrate would be prefered because it meant that the ADC - DAC electronics will be greatly simplified... and thus lower hardware costs...

I do not think anyone can hear above 20 kHz.. In fact, most 96 kHz system will finally output a audio bandwith of 20 kHz.. and not 48 kHz..

wkwai

Just a little question

Is it good to resample CD-A to 88/96 kHz before encoding to increase temporal resolution, and BTW, avoid pre-echo ?
And I think a quality resampling shouldn't degrade the wave file.

Regards,
Nick

Could be so, but on the other hand, you will need a way higher bitrate than just encoding something without resampling.

Cool !

Do you mean at least 192/256 kbps for stereo material ?

Temporal resolution is not increased by upsampling. If non-integer, it is even decreased by truncating effects.

Algorithms cannot make up information.

Thanks both Ultranalog & Latexxx for your reply !

If I follow you, it's only good for true 24/96 sources...
But for 44.1 kHz, 88.2 kHz should be a a acceptable value, right ?

This thread is a little misinformed. Please don't upsample your redbook CDs. There is almost no reason to justify doing this. As alluded to earlier, higher sampling rates are most useful in the conversion process because the analog circuits you need can be much simpler and more accurate for the 20Hz to 20kHz range.

Yeah... cheap electronics..

>>Temporal resolution is not increased by upsampling. If non-integer, it is even
>>decreased by truncating effects.
I'm unsure, but this statement seems to be wrong.
EDIT: Sorry. My bad. This statement is true when it is used ALONE.
But this is not true for aac or mp3 or other similiar encoding algohytm.
>>Algorithms cannot make up information.
The speak is not about information, but about encoding windows, that will affect temporal resolution.
You can have good freq or good temporal resolution, but not both. It is clear. Encoding at higher freq rate will eat up bitrate, but may increase temporal resolution (aka blocks change their length in time domain).
As I understand things, this effect will be used in DSBR.
I must apologize, I did not investigate how DSBR affects encoding results, thus information I provided can be somehow incorrect... I think Garf or Menno or Ivan should know it better
I hope we will see the DSBR implemetation soon and check it by ourseves