Jul 22, 2015

Audio Formats
An audio format is a type of computer file that stores music. Music formats are either uncompressed lossless, compressed lossless, or lossy.

BITRATES
A bitrate is the number of bits conveyed or transferred in a unit of time. When talking about music formats, bitrate is used in kilobits per second (kbps). When comparing files with different bitrates (of the same song), the file with the higher bitrate has the higher quality.

For example, an MP3 320kbps (CBR) file transfers 320 kilobits per second.

UNCOMPRESSED LOSSLESS
Uncompressed lossless formats store all of the original recorded data. Since silence is given the same number of bits per second as sound is, uncompressed lossless files are huge. The main uncompressed lossless format is pulse-code modulation (PCM).

Examples

• WAV (PCM) (used on Windows)
• AIFF (PCM) (used on Mac OS)

COMPRESSED LOSSLESS
Compressed lossless formats store all of the original recorded data in less space than uncompressed lossless formats by compressing the data. By giving silence almost no bits per second and compressing sound, a compressed lossless file is usually half as big as the same song stored in an uncompressed lossless file.

Since both uncompressed lossless formats and compressed lossless formats retain all the data from the original recording, they can be transcoded between each other without a loss in quality.

Examples

• Free Lossless Audio Codec (FLAC)
• Apple Lossless Audio Codec (ALAC)
• Monkey’s Audio (APE)

LOSSY
Lossy formats are always compressed. Lossy formats have smaller file sizes than both uncompressed lossless formats and compressed lossless formats because they remove some of the original data. Usually the removed data is in the higher frequencies that humans can’t hear, however, there can be obvious audible differences between lossy formats and lossless formats.

Because lossy formats remove data during compression (and thus lose quality), lossy formats CANNOT be transcoded to lossless formats or other lossy formats without losing more quality.

Examples

• MPEG Layer 3 Audio (MP3)
• Advanced Audio Encoding (AAC)
• Windows Media Audio (WMA)
• Dolby Digital Audio Codec 3 (AC3)
• DTS Coherent Acoustics Codec (DTS)

FILE SIZE
Here’s an example of how the file size of the same song varies depending on whether the song’s format is uncompressed lossless, compressed lossless, or lossy. Let’s take the classic pop song, Sk8er Boi by Avril Lavigne. For reference, the song is 3 minutes, 24 seconds long.

Uncompressed Lossless — WAV (PCM): 34.3 MB

Compressed Lossless — FLAC: 25.75 MB (25% compressed)

Lossy — MP3 320 (CBR): 7.78 MB (78% compressed)

TRANSPARENCY
Transparency is a term used to describe the audible quality of a lossy music file. A lossy file is considered transparent if the average human cannot tell the difference between the lossy file and a lossless file of the same song by just listening to both without knowing which file is which.

For most people, MP3 192kbps (CBR) is considered transparent.

 

Jul 22, 2015

MP3
LAME
LAME Ain’t An MP3 Encoder (LAME) is an encoder that converts and compresses any input audio file and outputs an MP3 file. The resulting MP3 file can have a constant, variable, or average bitrate. LAME is recommended as an MP3 encoder because it is open source, customizable, and outputs high quality MP3 files.

CONSTANT BITRATE (CBR)
When encoding a constant bitrate (CBR) file, the user (you) chooses a preset bitrate and LAME targets that bitrate throughout the entire file. This means that every second in the file has the same number of bits, no matter how simple or complex the sound is. Every second in a CBR file has the same quality. Because silence is given the same number of bits as more complex sounds, CBR files are larger than VBR and ABR files of the same quality. This also means that CBR files have a predictable file size.

VARIABLE BITRATE (VBR)
When encoding a variable bitrate (VBR) file, the user (you) chooses a preset quality and LAME targets that quality, letting the bitrate vary throughout the entire file. This means that every second of the file has a different number of bits that depends on how complex the sound is at that second. For example, a second of silence would receive much fewer bits than a second of loud, blaring music. Since VBR files target a certain quality instead of a certain bitrate, exact VBR file sizes are more unpredictable.

LAME has certain VBR presets ranging from V0 to V9. V0 is the highest quality VBR preset and V9 is the lowest quality VBR preset. The two most common VBR presets are V0 (with a target bitrate of 245kbps) and V2 (with a target bitrate of 190kbps).

At a certain VBR preset setting, the average bitrate throughout the file is usually close to thetarget bitrate. However, keep in mind that the target bitrate is just a target — the average bitrate will not necessarily end up near the target bitrate.

AVERAGE BITRATE (ABR)
When encoding an average bitrate (ABR) file, the user (you) chooses a preset bitrate and LAME allows the bitrate to vary throughout the entire file, but the average bitrate of the file will be the bitrate you preset. This means that like CBR, the file size is predictable, and like VBR, the quality and bitrate of the music varies throughout the file depending on how complex the music is each second.

 

Jul 22, 2015

Transcodes
Transcoding (verb) a file means converting from one format to another. A transcode (noun)can mean any converted file, but is usually used in a negative context (as in a bad transcode).

GOOD TRANSCODES
A good transcode means that during the transcode process, the file has either never been converted to lossy, or the file has only been converted to lossy once during the last step.

Examples of good transcodes:

• uncompressed lossless > compressed lossless
• compressed lossless > uncompressed lossless
• compressed lossless > compressed lossless
• uncompressed lossless > lossy
• compressed lossless > lossy

BAD TRANSCODES
A bad transcode means that during the transcode process, the file has either been converted to a lossy format more than once, or the file has been converted from lossy to lossless.

Examples of bad transcodes:

• higher lossy bitrate > lower lossy bitrate
• same bitrate lossy > same bitrate lossy
• lossy > lossless

 

Jul 22, 2015

Spectral Analysis
Spectral analysis is a visual way to display the data in a music file. Every music note has a specific frequency: lower notes have lower frequencies and higher notes have higher frequencies. All of the frequencies are displayed on a spectral diagram (“spectral” for short), which is a graph of all the frequencies vs. time in a music file. Frequencies are measured in hertz (Hz) and kilohertz (1,000 Hz). Humans have a hearing range from about 20 Hz — 20kHz (20,000 Hz).

Since spectrals show all the data in a file, they are helpful tools to use when you’re trying to decide whether or not a song has been transcoded. Every file has a relatively standard frequency cut-off.

Click on any of the spectrals below to view it in a higher resolution.

CD / LOSSLESS
Songs on a retail CD and lossless songs have frequencies that extend all the way to 22 kHz. Since lossless to lossless transcoding preserves all of the data in a music file, the spectral of a lossless song will look the same in FLAC, WAV (PCM), ALAC, etc.



However, different genres have different-looking spectrals. The example above was a pop song, so most of the frequencies were represented. But look at this classical piano song.



It looks much different, right? But it’s still a lossless spectral! Notice how “white noise” (the light purple) still extends to 22 kHz, even though those frequencies aren’t used.

MP3
Different types of MP3s have different frequency cut-offs. MP3s also tend to have a “shelf” at 16 kHz (you’ll see it in the spectrals).

MP3 320kbps (CBR) has a frequency cut-off at 20.5 kHz.



MP3 256kbps (CBR) has a frequency cut-off at 20 kHz.



MP3 V0 has a frequency cut-off at 19.5 kHz.



MP3 192kbps (CBR) has a frequency cut-off at 19 kHz.



MP3 V2 has a frequency cut-off at 18.5 kHz.



MP3 128kbps (CBR) has a frequency cut-off at 16 kHz.



TRANSCODES
How are spectrals helpful when trying to detect transcodes? Say you download a song in FLAC from a blog. The only way to verify that this song is truly a lossless file and not a transcoded file is by looking at its spectral. (Programs like AudioIdentifier are not reliable at detecting transcodes.)

For example, the spectral below is of a FLAC file: the file extension is .flac, it is 21.8 MB, and it sounds okay.



But whoa, does that look anything like what a regular FLAC spectral should look like? No! This file was transcoded from MP3 192kbps (CBR) to FLAC. It’s a lossy to lossless transcode, which is bad.

PROGRAMS
For spectral analysis, we recommend using either Adobe Audition (Windows or Mac OS),Audacity (Windows, Mac OS, Linux), and SoX (Windows, Mac OS, Linux — command line only). All of the spectrals that appear in this guide were viewed in Adobe Audition CS 6.

Although you should use spectral analysis to determine whether a file is a transcode or not, you will need to use another program to first determine what bitrate or encoding preset the file claims to be. For this purpose, we recommend using Audio Identifier or dbPowerAmp on Windows and dnuos or MediaInfo on Mac OS.

 

Jul 22, 2015

Spectral Analysis
Spectral analysis is a visual way to display the data in a music file. Every music note has a specific frequency: lower notes have lower frequencies and higher notes have higher frequencies. All of the frequencies are displayed on a spectral diagram (“spectral” for short), which is a graph of all the frequencies vs. time in a music file. Frequencies are measured in hertz (Hz) and kilohertz (1,000 Hz). Humans have a hearing range from about 20 Hz — 20kHz (20,000 Hz).

Since spectrals show all the data in a file, they are helpful tools to use when you’re trying to decide whether or not a song has been transcoded. Every file has a relatively standard frequency cut-off.

Click on any of the spectrals below to view it in a higher resolution.

CD / LOSSLESS
Songs on a retail CD and lossless songs have frequencies that extend all the way to 22 kHz. Since lossless to lossless transcoding preserves all of the data in a music file, the spectral of a lossless song will look the same in FLAC, WAV (PCM), ALAC, etc.



However, different genres have different-looking spectrals. The example above was a pop song, so most of the frequencies were represented. But look at this classical piano song.



It looks much different, right? But it’s still a lossless spectral! Notice how “white noise” (the light purple) still extends to 22 kHz, even though those frequencies aren’t used.

MP3
Different types of MP3s have different frequency cut-offs. MP3s also tend to have a “shelf” at 16 kHz (you’ll see it in the spectrals).

MP3 320kbps (CBR) has a frequency cut-off at 20.5 kHz.



MP3 256kbps (CBR) has a frequency cut-off at 20 kHz.



MP3 V0 has a frequency cut-off at 19.5 kHz.



MP3 192kbps (CBR) has a frequency cut-off at 19 kHz.



MP3 V2 has a frequency cut-off at 18.5 kHz.



MP3 128kbps (CBR) has a frequency cut-off at 16 kHz.



TRANSCODES
How are spectrals helpful when trying to detect transcodes? Say you download a song in FLAC from a blog. The only way to verify that this song is truly a lossless file and not a transcoded file is by looking at its spectral. (Programs like AudioIdentifier are not reliable at detecting transcodes.)

For example, the spectral below is of a FLAC file: the file extension is .flac, it is 21.8 MB, and it sounds okay.



But whoa, does that look anything like what a regular FLAC spectral should look like? No! This file was transcoded from MP3 192kbps (CBR) to FLAC. It’s a lossy to lossless transcode, which is bad.

PROGRAMS
For spectral analysis, we recommend using either Adobe Audition (Windows or Mac OS),Audacity (Windows, Mac OS, Linux), and SoX (Windows, Mac OS, Linux — command line only). All of the spectrals that appear in this guide were viewed in Adobe Audition CS 6.

Although you should use spectral analysis to determine whether a file is a transcode or not, you will need to use another program to first determine what bitrate or encoding preset the file claims to be. For this purpose, we recommend using Audio Identifier or dbPowerAmp on Windows and dnuos or MediaInfo on Mac OS.

 

Jul 22, 2015

My man

Nice to see somebody who shares my near OCD interest of audio

 

Jul 22, 2015

neat read m8

 

Jul 22, 2015

i didnt know where to put this thread tbh

 

Jul 22, 2015

Gaming/Tech is fine

Anywhere but Off Topic

 

Jul 22, 2015

i think it deserves a sticky tbfh


too many noobs when it comes to audio

 

Jul 22, 2015

Has anyone taken this 128kbps vs. 320kbps vs. flac test before?

http://www.npr.org/sections/therecord/2015/06/02/411473508/how-well-can-you-hear-audio-quality

And what were/are your scores?

 

Jul 22, 2015

I thought this said the Pedophile thread

 

Jul 22, 2015

6/6 with headphones, thats a great test, i loved it

btw its not flac :L its wav

uncompressed

 

Jul 22, 2015

i thought i would do great with my KRK monitors.. got like 2/6 but wasn't truly focusing my mind tbh..

 

Jul 22, 2015

the hardest for me was jayz, because of the d--- claps and electric sounds f---in with frequencies

 

Jul 22, 2015

 

Jul 22, 2015

man i hate that i can't redo the test.. if only there was another one just like it.

 

Jul 22, 2015

hahaha

 

Jul 22, 2015

When you want some HQ file, @detroitskills is your savior
haha

 

Jul 22, 2015

me