Kara-Moon Forum

This is a subject discussed in depth everywhere digital music is made - but I want a definitive thread here at Kara-Moon.
(please only post information you know to be accurate, so we can all use this thread as a reliable resource in our recordings and collaborations ;))

Marc JX8P and I have discussed collaborating with 24 bit / 48 kilohertz files. Great fidelity with a manageable file size.

Both Audacity and Ardour default to a 32 bit (floating-point) depth and an internal sampling rate of 44.1 kilohertz, but can be set to any standard bit depth and sampling rate up to 24/192.


My questions:

                Would there be any advantage in changing the internal bit depth and sampling rate in Audacity and Ardour to 24/48?

                Are projects destined for CD, MP3, or Ogg Vorbis files with a sampling rate of 44.1 kilohertz best recorded and mixed in my digital audio workstation at 44.1 kilohertz?             

24 bits instead of 16 bit is giving you more headroom to work with, 48khz opposed to 44.1 does not give much extra. Advantage of using 88.2khz (or 96Hz) is less aliasing. Also consider that not all plugins support 88.2 khz or even 48 khz.
Sample rates of 192 kHz only makes sence when recording samples, not realy for making music.

I'm using 24 bit and 44.1 Khz as a standard, 44.1 being the sample rates that complies the CD format (easier for mastering).

Note, this is my humble opinion. Since I miss the added value of working with 48 khz, I've stuck with 44.1khz which is more than enough to record decent music. In the past I did use 96khz, solely for the reason of bringing the latency of my audio card lower. But know with fast cards, cpu's and asio support, this is no longer an issue.

Moon

Moon,
I have been using a 32 bit (floating-point) depth, with the 44.1 kilohertz (Linux default)sample rate, and been very happy with the results. My new computer is fast enough to handle 96 kilohertz processing, so upgrading my audio file quality is now a practical option.

From my research, most experts recommend working at the highest bit-depth and and sampling rate possible to preserve the nuances of the source material, and allow audio processing without losing/corrupting digital information. Apparently there are frequencies humans cannot hear, but can feel, that affect the listening experience. Likewise, audio detail which we may not consciously notice, subtly enhances the sonic landscape. Why not take advantage of the increased processing power a new computer offers to reproduce the musical performance as faithfully as possible?

The other argument is that the intended playback format (CD, Ogg Vorbis, MP3) cannot reproduce the full spectrum of sound resulting from 32/96 audio processing - and many playback devices have severely limited dynamic range and frequency response.
My feeling: Make the best audio possible; then tailor it for each playback situation - big, bold, and beautiful for DVD Audio played through a premium sound system, or compressed and down-converted for lesser formats on low-fidelity systems.

For the next few projects, I'm going to keep the Audacity/Ardour default 32 bit floating-point setting, but increase the sample rate to 96 kilohertz, just to see (hear :D?) how the computer performance is affected. Wish me luck... ;D

I agree with the comments here.
I always work with 24 bit audio for higher resolution (yeah more headroom)

for sample rate it depends on the project. I like to work with 48 khz but if the project has many tracks and many effects my CPU could have problems handling it so if the final result will be CD or MP3 I stay with 44.1 Khz.

Cheers

the attention is decisive on integer multipliers  8) ...cya

<2°TRo²³

Right on topic and a good read : http://mixonline.com/recording/mixing/audio_emperors_new_sampling/

k

Good article, Kara!

It is becoming evident that regardless of file size, bit depth, and sample rate,...the actual playback equipment and listening environment are critical to the quality of sound.

Also, the care with which an audio file is prepared will, to a large extent, determine the quality of the listening experience.
A 160 kilobit/second MP3 from RHarv will often sound better than a full 16bit .wav file, just because he knows which frequencies to emphasize, and how far to push the compression to get maximum impact while still preserving dynamics.

Very true and, even without research, a bit of a common sense issue. There's also, of course, the issue of diminishing returns. There's no doubt that 48 kHz sounds better than 11.025 kHz. But 96 kHz towards 48 kHz will be much more subtle and I'm not even speaking about 192 kHz. Considering bit depth, I think that is more of an issue (when I listen to a DVD movie with music playing at 48 kHz/24 bit there's a significant greater depth to it than similar music on a cd) but ONLY WITH THE RIGHT MUSIC! If the music doesn't have a lot of dynamic range, then you could probably well suffice with 16 bit.

So, after talking with some very knowledgeable audio enthusiasts and reading as much as possible over the last few years from periodicals and the web, here's how it looks.....

16 bit / 44.1 kilohertz is just fine for preserving audio recordings and exchanging audio files over the web during internet collaborations. The format can be made even more serviceable by compressing the file with MonkeysAudio or FLAC (Free Lossless Audio Codec) for storage or transmission via FTP or e-mail.

Currently, a new audio standard for DVDvideo has been introduced - 24bit / 48kilohertz. For those who want to store digital audio or send audio project files, but at a higher fidelity than the standard WAVE format, 24/48 looks like a very workable choice. Audio file size is kept manageable, dynamic range is significantly enhanced, and those elusive frequencies occurring around the upper limit of human hearing will be more completely preserved.

As for bit depth and sampling rate within your digital audio workstation, the highest values your computer can manage will get you the most accurate processing. Using the Linux workstation, Ardour, in a personal computer with 4 gigabytes of RAM and a dual processor AMD chip clocked at 3 gigahertz, I can comfortably use a 32 bit(floating point) depth, and a sampling rate of 96 kilohertz(96,000 Hz).

More bit depth and higher sampling rates are not absolutely necessary for excellent quality audio, but if you're going for the gusto...
.....that's my story ;)

Comments and criticisms welcome....

The 24 bit / 48khz format can be considered as the standard. I was stull puzzling why most of my gear takes this setting as default. The new standard makes it starting to make sense...

Personally, I don't think that 96khz or 32 floating bit brings additional value, unless you're in the business of sampling.

24 bit / 48khz is giving us enough room to make pro-quality sounding recordings. Why spoiling more cpu anyway ? Is the globe not warm enough ?

Moon

Moon,

Agreed; important points.... ;)

Personally, I think a 16bit/44.1khz audio file is quite sufficient. Carefully engineered digital audio mastered to this format can sound spectacular.

Most audio processing software designed since Windoze XP  processes audio (internally) at 32bit floating point. It is the default internal bit depth, and seems to work well.
96,000 hertz is, however, more than we need to achieve detailed, spacious sound. My choice would be 44,100Hz. A lot of excellent audio has been processed at 32bit(floating point) / 44,100Hz.

(however, since I handle other people's audio mixing and mastering, and that can involve a lot of intensive digital processing, the higher numbers discussed in the last post assure that they are getting the most comprehensive audio performance available)

Excellent point..


..and you have certainly demonstrated that.

Wyatt

 :-*.............................................................. :D.................................................... ;)

+1   8)

Thanks, L.R.

It was the quality of audio production on the forum posts that attracted me to this site. A lot of our members have advanced production skills and refined musical abilities (look in the mirror, fella ;)). Perhaps more so than anywhere else on the web.... :;

Back to the origins of the thread, and the conclusions (based on some excellent advice, both here on this thread and via e-mail):

Most modern "digital audio workstations" process sound internally at a 32-bit floating point bit-depth. This is good, and should not be altered.
The internal sampling rate of a "digital audio workstation" will usually default to 44.1KHz. For all but the most demanding and specialized work, 44.1KHz is adequate to produce world-class audio.(I have experimented with sampling rates up to 96KHz, and cannot hear the difference :-[)
___________________________________________________________________________________________________________

An accessory issue which is often confused with my original question is:
           
 What choice of bit-depth and sampling rate is best for the "digital audio files"(wave files) we use? Although many of us who are striving to produce an exceptional audio experience are working with 24bit/48KHz wave files, most listeners and producers(including me) find that 16bit/44.1KHz wave files are entirely adequate to produce big, rich, detailed sound.

Thanks to everyone who contributed information, opinions, and advice  :;

Oren.

I read once (in a reputable forum) that an audio cable is capable of less dynamic range than 24bit when considering the noise levels..  ie the noise level inherent in an audio cable is above the noise floor of 24bit.. at line levels...

just thought I'd thro a little more mud into it

Ah noise. Now there's an interesting subject ::)

What do you use to measure noise when the measuring equipment generates its own noise?

Same for distortion.

Rharv:"...an audio cable is capable of less dynamic range than 24bit when considering the noise levels..       ie the noise level inherent in an audio cable is above the noise floor of 24bit.. at line levels..."

Folderol:"...What do you use to measure noise when the measuring equipment generates its own noise?
Same for distortion..."

Precisely the purpose of this thread.
How much of this theory, and how many of these numbers, are actually applicable in a real-world situation? How much is pointless conjecture?

The goal is premium audio. My intention is to de-mystify the subject, and answer the question "What works, and what doesn't?".

Just so as we can say we really drove this sucker right into the ground:

http://mixonline.com/recording/mixing/audio_emperors_new_sampling/

The Emperor's New Sampling Rate.

8)

Wyatt

Nice article. I you consider yourself an audio purist, read this one and be surprised !  ;)

Moon

Coming from a highly hands-on electronics background I'm more pleased than surprised  :D

... almost to the point of being insufferably smug ::)

Interesting that time/phase differences should get a mention. I'm firmly of the opinion that many stereo panning systems are less that ideal simply because the don't make the appropriate phase corrections.

At 1kHz a wavelength is roughly 300mm. In a typical listening triangle the path difference from one speaker to both ears is around 100mm. That will result in a very significant phase shift, and I know from practical headphone tests, people can hear phase difference at that frequency.

I have read this study before, which inspired me to engage in some creative listening situations of my own.

Playback equipment and listening environment had a dramatic effect on the quality of sound. Audio file format, sampling rate, and bit depth made little significant difference.
The care with which the audio file was created, however, rendered astonishing differences. Low bit-rate MP3s and Ogg Vorbis files prepared with extra attention to detail sounded as good to me as many commercially engineered 16/44.1 CD tracks.

My mother told me not admit to such things in public.... :-[ :'(

I enjoyed it.  Lots to think about in little space.