AUDIO IN ACTION

Introduction:

Given the standardization of CD-Recordable drives in today’s computers, combined with prevalent desktop music software (e.g.: iTunes, etc…) which default to low KPS perceptual audio encoders, it has become increasingly common to find individuals creating/copying Audio CD’s that have gone through one (or more) lossy encoding stages, often without the user realizing that their audio data has been compromised.

This paper aims to raise awareness of the issue, provide information, and links to further resources in the interest of empowering the reader to determine if their Audio CD’s are utilizing the full potential of quality within the Audio CD “Red Book” standard (Linear PCM 16-bit / 44.1kHz sr audio).

What is Perceptual Audio Coding?

Encoding:

A Perceptual Audio Encoder is a software program that takes full-bandwidth audio data as an input signal (.WAV, .AIF, etc…) and processes it into a “lossy” audio format (.MP3, .AAC, etc…), where “lossy” means that is resulting encoded file is not a replica of the original.  The goals of a perceptual audio encoder are:

1)  Create as small a file size as possible
2)  Retain as much fidelity of the original source data as possible given the applied perceptual model and chosen encoding parameters

Encoders use “Perceptual Models” which are algorithmic constructs designed to accept a filtered (Band-split) audio stream and apply psychoacoustic rules (defined in the model) to discard audio information that is deemed to be un-heard (information above the human hearing threshold, information being “masked” by more prominent frequencies, etc. etc.).

A sample workflow of a Perceptual Audio Encoder is as follows:

Decoding:

On the receiving end, a Perceptual Audio Decoder is used to “un-pack” the encoded bitstream back into audio.
The decoder used must be matched in protocol to the input bitstream (i.e.: An .MP3 decoder will not decode .AAC files, etc.).

A sample workflow of a Perceptual Audio Decoder is as follows:

Together, you have a “Codec” (Encoder/Decoder).

Note: The above two flowcharts are taken from representations in the AES paper #17-009 “High Quality Audio Encoding”.
See link in “Additional Information” section for more information. Copyright information for Karlheinz Brandenburg remains with respective owner(s).

Commonly Encountered Pitfalls – A Real-World Use Case:

iTunes and other common desktop music catalogers/organizers allow anyone to create an Audio CD with a few clicks. However are you really creating an audio cd at the bandwidth possible to the CD Red Book Standard, that being 16-bit, 44.1kHz sr?  Here we explore the impact of compounded fidelity degradation using iTunes in its default settings state.

Within the Preferences dialog, under the “Import Settings” dialog, you can find where to set encoder rules. Here is a screenshot of the default encoder values of iTunes 9:

Starting with version 9,  iTunes defaults to .AAC, 256kps. If you wish to copy an audio CD (CD-DA) then you would want to change the file format within this dialog to .WAV or .AIFF (full bandwidth file formats), and then perform the copy. If you leave this dialog to a lossy format (.AAC, .MP3, etc…) and rip/burn an audio CD, you have incurred generation loss via the encoder, and the resulting audio CD is not a replica of your source.  Additionally, it is worth noting that audio cd copies made via the playlist rip/burn method will strip copies of TOC (Table of Contents) information/metadata and alter the resultsing PQ subcodes so that metadata recognitions may not work as desired.  For more information please see articles:
——–metadata link
Making Reference Copies:  http://mikewellsmastering.com/blog/2010/03/08/making-reference-copies/

Q)  So what is to be learned from this?
A)  If you are creating/copying cd’s using the iTunes application in its default settings state, you are incurring a generation loss in fidelity.

Summary:
If you are not aware of the workflow/audio algorithms of the application you are using to create/copy audio CD’s, then you run the risk of compounded fidelity degradation.

How to determine if a cd you have received has been processed through a perceptual codec:

1) Listen:

Nothing beats a pair of trained ears.  Get to know what the various codecs “sound” like by performing your own listening tests.
Additionally, due to the nature of lossy encoding, working solely from analysis charts is not enough.  Different programme material reacts differently per codec, so understanding how the parameters being fed into the source is affecting the end result is key.  ABX testing is a great way to train your ears to learning different codecs.  For more information on ABX testing, see the provided links below.

2) Analyze:

A wealth of audio analysis tools exist in the marketplace today.  Choose a few and get to know them, and use them as supporting information to what your ears are telling you.  Here are some tools to assist you in determining if your audio cd is full-bandwidth or if it has been compromised.

True Audio – Tau Analyzer:
The Tau Analyzer (Freeware) runs against a CD-DA disc you have in your computer drive, and performs an analysis to determine if the source is a true CD-DA stream or if it has first been processed through a lossy codec.  Results are displayed textually, and graphically as a time-based sonogram.  To aid in educating yourself to the “sound” of codecs, take a look.

For this test, a 16 bit, 44.1kHz .WAV file was used, along with the same file encoded into .MP3 format at a popular setting of 128kps.

Source file (16/44.1 .WAV file):

Encoded file (128kps .MP3 file):

Sony Sound Forge – Spectral Analysis:

Sony’s Sound Forge application supports a spectral analysis “summary view”, meaning that the resulting graph is representative of the frequency spectrum averaged across the length of the track. This can assist in “seeing” if a codec was present at one point.

Source file (16/44.1 .WAV file):

Encoded file (128kps .MP3 file):

Potential hazards, downstream effects, and why you should care:

If you are participating in the recording process, and you encounter this situation while moving final mixes from the recording studio to the mastering facility (as an example), you run the risk of the final mixes being rejected by the mastering engineer (best-case scenario). Alternatively, you have the potential of going through the mastering process using these compromised data sets, resulting in a less-than-cd-quality audio master to be replicated into a final product (worst-case scenario).
Additionally, recipients of the replicated product will most likely re-encode the files into their desktop music software, thus introducing yet another (and probably different) lossy encoder into the equation, reducing the fidelity further.

How do you solve this problem?

1)  Be aware of how you are delivering/receiving files to/from a recording studio/engineer/mastering facility/artist/etc…
2)  Listen  & Analyze the programme material. If you hear or see suspicious results, investigate further.

If you are part of the modern-day recording world; now more than ever it is the responsibility of the audio engineering community to pro-actively engage and educate end-users & clients of the risk of compounded lossy encoding recordings.

Further considerations:

This article demonstrates a couple of simple methods for determining fidelity of incoming material delivered via optical media containing a CDDA stream.  As an audio engineer, you owe it to your clients to help educate them of this process and the potential pitfalls they may be incurring against themselves without even knowing it.

Additionally, while updating this article I found that with iTunes 9, the app now defaults to 256kps encoding, which while a great step in the right direction, is still applying a lossy encoding stage, period.  Furthermore, it is more important to note that we have a decade of low-res encodings floating around the netsphere, people’s hard drives, etc…  so if a rough mix from 5 years ago wants to be added as a bonus-track to an upcoming release during a mastering session, you’ll always have this issue to regard going forward.

Be aware and be on top of your game.

Lastly, this article does not attempt to tackle the issue of compounded lossy encoding stages (transcoding), and/or the introduction of stems within a stereo mix that may be lossy encoded (or transcoded) due to collaboration (ie:  someone new to audio recording doing guest vocals on a track supplying an .MP3 as their stem to place in the mix, happens all the time as .WAV files are too big to distribute through email and people start looking for quick solutions to solve problems, etc..).  This subject is an ever-growing phenomena and will be discussed seperately.

Looking Ahead and Action you can take today:

Within the last few years we have seen a wealth of free and low-cost file-sharing applications come to market that are a wonderful gift to audio engineers in sharing files, collaboration, submitting mixes for mastering, etc.  These tools come in the form of YouSendIt.com, DropBox.com, SoundCloud.com, etc.  And of course there is the mother of all file-sharing techniques:  FTP.

So as an audio engineer, engage your clients as to how they are preparing files for collaboration, mixing, mastering, etc. and help them navigate their DAW to deliver the highest fidelity from their end to yours, etc.  Everyone wins.

Additional Information:

True Audio Tau Analyzer:  http://www.true-audio.com
Sony Sound Forge:  http://www.sonycreativesoftware.com/products/soundforgefamily.asp
MP3 Information:  http://en.wikipedia.org/wiki/Mp3
AAC Information:  http://en.wikipedia.org/wiki/Advanced_Audio_Coding
ABX Testing Info & Tools:  http://www.hydrogenaudio.org/forums/index.php?showtopic=16295&
Audio Codec Comparisons:  http://en.wikipedia.org/wiki/Comparison_of_audio_codecs
AES White Paper 17-009:  http://www.aes.org/e-lib/browse.cfm?elib=8079