Setting the standards: audio capture, release, comparison

[spoRv]

I guess we need to set some standards here, about audio capture, release, comparison.

There are my guidelines - please let me know if you agree or if not, state reason and alternative.

AUDIO CAPTURE

PCM tracks:

• if a bit-perfect capture card is available, capture it at 44.1kHz/16bit
  
• if a bit-perfect capture card or digital input is not available, capture it at 48kHz/24bit from analog outputs to avoid second conversion
  

Analog tracks: capture at 48kHz/24bit

AC-3 tracks: bit-perfect capture card with digital input and AC-3 demodulator are needed; capture at 48kHz

DTS tracks: bit-perfect capture card with digital input; capture at 44.1kHz/16bit

AUDIO RELEASE (captured tracks)

PCM tracks:

• if captured with a bit-perfect capture card, release as 44.1kHz/16bit is mandatory; a 48kHz/16bit conversion is allowed
  
• if captured through analog outputs, release as 48kHz/16bit
  

Analog tracks: release as 48kHz/16bit

AC-3 tracks: release as 48kHz/XXbit - where XX is the number of bit of the AC-3 track when de-encapsulated

DTS tracks: release as 44.1kHz/16bit is mandatory; a 48kHz/16bit conversion is allowed

Release formats: WAV, FLAC; DD ONLY for untouched AC-3 captures; DTS ONLY for untouched DTS capture

Note: releasing captures at more than 16bit is useless, while capturing at 96kHz is not only useless, but can also be harmful, introducing unwanted high frequencies and forcing the capture card to use slightly less dynamic range than 48kHz.

AUDIO CONVERSION

Question: what about captured track levels? Set the maximum at 0dB? Set the maximum according to other tracks if in sync?

AUDIO RELEASE (other tracks)

The following guidelines could be applied only if the audio is edited; if the track is untouched, release it "as is".

Dolby Digital: while professional encoders have good quality, the free ones are noticeable less refined; hence if you decide to release tracks in Dolby Digital using a free encoder, use a higher bitrate than the source

Examples:

DD source → DD release (minimum bitrate):

• 192kbps → 256kbps
  
• 256kbps → 384kbps
  
• 384kbps → 448kbps
  
• 448kbps → 640kbps
  
• 640kbps → release as DTS or lossless
  
• DD EX → release as DTS-ES or lossless
  

DD+ source → DD release (minimum bitrate):

• 192kbps → 384kbps
  
• 384kbps → 640kbps
  
• 448kbps or higher → release as DTS or lossless
  
• 6.1ch or more → release as DTS-ES or lossless
  

DTS: avoid to use half rate, as it has a rolloff at high frequencies; if you need to have small size track, use DD 640kbps instead.

Examples:

source → DD release (minimum bitrate):

• DD/DD+ 448kbps or higher → DTS 1536kbps
  
• DTS 1536kbps → release lossless
  

Lossless: PCM, Dolby TrueHD and DTS-HD MA sources should be released as well as lossless; anyone of the three main format could be used - PCM, TrueHD, DTS-HD MA - even if different from the source; FLAC is accepted as added option, or only when an audio-only track is released

Bit depth and frequency: I continue to think that releasing track at more than 16bit is useless and more than 96kHz could be harmful; even if someone could theoretically manage to hear the dynamic difference of 98dB that 16bit allowed - maybe using great headphones and risking deafness - I bet it is not possible, using any speaker system, listening to a movie audio track.

AUDIO COMPARISONS

To show audio tracks features graphically, it's possible to post waveforms and spectrums; we have only to find a standard way to do it.

As not everyone could have access to paid tools, I guess we should stick to free software; I'd go with Audacity for waveforms and Spek for spectrums; so we have comparable screenshots.

Set Audacity fullscreen, tracks as wide and high as window, waveforms in dB, meter dB range at (I guess 60dB is the minimum; 96dB is already too much, while 36dB is not enough; sadly someone is posting screenshots from other software with range set at only 24dB; in that way, most of the track seems completely silent!!!); set Spek fullscreen too; release image at 1920px width (scale it if your resolution is bigger or smaller); height is dependant from the number or tracks - for example if you want to show more details in an 8-channel track, you could take two (or more) screenshot and glue them together as a single one.

Note: is there any Nyquist programmer? Would be nice to modify a plugin to get stats similar to professional softwares - nothing fancy, only highest and lowest peaks, noise floor, track length etc.

EDIT: written an article about waveform comparisons - contains different dynamic range from 18dB to 24bit full scale 145dB:
http://blog.sporv.com/comparing-tracks-waveforms/

[spoRv]

Added link to waveform comparison article

[BusterD]

Analog tracks: capture at 48kHz/24bit

Analog tracks: release as 48kHz/16bit

I think it's generally agreed upon that 24-bit is best for preservation and future editing purposes, but converting to 16-bit probably doesn't save that much bandwidth, does it?  Also might be a good idea to share it at 24-bit in case the original capture is lost or the project maker goes offline, etc.

96kHz might be overkill, but I similarly like to capture at that rate and release it that way for peace of mind and because I don't like to mess with audio conversion unless absolutely necessary.   I personally don't really need to save room on storage nowadays, at least.
 
Also, we should probably discuss best practices for sample rate/bit depth conversion before recommending a different format from the one used upon capture.

[spoRv]

I think it's generally agreed upon that 24-bit is best for preservation and future editing purposes, but converting to 16-bit probably doesn't save that much bandwidth, does it?  Also might be a good idea to share it at 24-bit in case the original capture is lost or the project maker goes offline, etc.

96kHz might be overkill, but I similarly like to capture at that rate and release it that way for peace of mind and because I don't like to mess with audio conversion unless absolutely necessary.   I personally don't really need to save room on storage nowadays, at least.

Personally I think 48kHz/16bit is good for 99.99% of movie tracks and/or people, but I can understand and endorse the use of 24bit for editing and/or capturing; actually I'd go with 20bit, as there is no ADC with much more effective resolution of 20bit, but there are no 20bit ADC anymore, I think, so... let it be 24bit! And 96kHz to me is completely futile.

But let's pretend that anyone could hear up to 48kHz and bare 144dB of dynamic without having eardrums blowed instantly when a 144dB loud sound would come, and we listen to the movies with amps and speakers able to reproduce 48kHz and 144dB peaks without exploding; let's pretend for a moment that we are all rich superheroes owning imaginary hi-end hardware, can we?

Even in that case, what would be the reason to capture at 96kHz a signal that in the best-case scenario may have the highest frequency response of 20kHz, or release it at 24bit when it contains less than 14bit of information - apart your peace of mind?

Is not about storage - even if 96/24 is three times 48/16... - it's about facing reality: marketing is bombarding us with something new every day, and the most advanced tech of yesterday is s**t immediately. But fact is, this is not true: they are forcing us to believe that we "need" this and that, when a 10 or 20 years old "stuff" could perfectly fit the job.

One example I found myself lately; I thought to get a supa-dupa multichannel capture card, and I face the problem "would be my 'old' 16core CPU be powerful enough to capture, let'say, 16ch analog audio?" - answer: a 15+yrs old PC can capture more than 100ch 96/24 without a fuss!

So, to allow future edit of a captured track, even 16bit is more than enough; 96kHz is completely useless.

I'd prefer any time a 48/16 lossless version instead of 96/24 lossy one, whichever format and encoder it used!

Also, we should probably discuss best practices for sample rate/bit depth conversion before recommending a different format from the one used upon capture.

Well, I recommended to capture at 48/24 - adjusting the levels and release at 48/16; this would shave only 8 bits of LSB (read: noise), nothing else - and format would be the same - capture as WAV, release as WAV (or FLAC).

Unless you were talking about AC-3/DTS captures - true, I did not state to release it as untouched AC-3/DTS; edited now!