Collapse
Torrentech Radio is currently offline.
The radio is currently offline · No scheduled sets...
Schedule · DJ-s · Stats

> Help Topic
Here is the topic you chose to view
Enter keywords to search for   

> Transcoding & Spectral Comparisons
Here is the topic you chose to view
What is a Transcode?

Put simply, a transcode is any audio file that wasn't encoded from an original lossless source, such as a retail CD (not created from lossy files) or a properly made lossless rip.



  1. Introduction: lossless and lossy formats

    This is not a guide. This section is here in order to visually show you the difference between the different kinds of files treated in the EAC guides. It ends with a brief explanation of what transcoding is and why transcoding normally is a bad thing to do.

    Our point of departure is the wave file: the .wav file that EAC produces as you rip an audio CD track. To be sure, .wav is really a container format that can hold different kinds of audio; in this case the content of the .wav file is "really" PCM audio.

    If you have ripped the CD properly, in accordance with the guides, the .wav file contains exactly the same audio data as the original CD. Wave files are rather large, though, which makes them awkward for transfer and/or storage. This is where FLAC enters the picture. It stands for "Free Lossless Audio Codec". It compresses the audio data without removing any of it, just like Zip and RAR compress data without removing any of it. This is why FLAC is properly called a lossless format. With the right kind of cue sheet, you can re-create a copy of your CD from your archived FLAC files any time.

    Lossy formats can save space by removing some audio data, for example by cutting down on the number of kilobits used per second of audio and/or removing "unnecessary" audio data according to some algorithm, in addition to compressing the files. This way the files can be made much smaller, but there is always a trade-off between audio quality and file size.

    A common standard for "acceptable" audio quality for mp3s is 128 kbps. It is a constant bitrate (CBR) setting that you see, for example, with most iTunes downloads (though they make use of the FhG encoder, not the superior LAME encoder for mp3). Ogg Vorbis has been developed to focus on quality measures rather than bitrates, but a rough equivalent to mp3 at 128 kbps would be the q3 setting.

    Others claim that the threshold for "transparency" (where you can't tell the difference from the original) would be at 192 kbps for mp3s, and q6 (or even q5) for Ogg Vorbis. For LAME mp3, there is also the variable bitrate (VBR) preset V2 (target bitrate ~190) that would be the "transparent" VBR setting. (Variable bitrate, used in both the Ogg Vorbis -q presets and the LAME mp3 -V presets, means that the encoder analyses the audio data and only uses as high bitrate as is needed, i. e. lower bitrate for less complex (or silent) parts of the audio data.)

    The pictures below are intended to show you the differences between different formats and bitrate settings, from "best possible" to "transparent" to "acceptable".




  2. Spectral view

    The pictures are screenshots taken while using the "spectral view" in Adobe Audition (v. 1.5). The spectral view is a way of making some audio properties visible. The screenshots show you the two stereo channels with the left channel on top. The horisontal axis is for time, the vertical axis shows the frequency of the audio data for each channel. The colours indicate the amplitude: from dark blue for the lowest decibel (softest) sounds to bright yellow for the highest decibel (loudest) sounds.

    These screenshots show a maximum of 22 kHz for each channel. That is because we are looking at CD rips with the standard frequency of 44.1 kHz (combined for both channels). Other recordings, such as some soundboard tapes or vinyl rips, may go much higher.

    All the screenshots, except two of the FLAC ones, are for the same track: "I Feel You" from Depeche Mode's Songs of Faith and Devotion. That is a reasonably complex audio track that serves to illustrate the differences between the different encodings. The LAME mp3 and Ogg Vorbis files were created from the FLAC file using dBpoweramp.

    An easier way to get the spectral of an audio file is using this program: http://spectro.enpts.com/index.php
    It's a small tool that shows the spectrogram and the power density of the file it loads. It also has the advantage that it opens all kinds of files.

    here is a screenshot of the results of Spectro:

     
     
     
  3. FLAC

    First out is the original FLAC file - if it were decompressed to WAV (PCM) it would still look exactly the same. (In fact, any file you open in Audition is "converted" to WAV in order to show the waveform.) As you can see, it goes all the way up to 22 kHz. The file size is 31.2 MB; the decompressed .wav file would be 46.3 MB.


    FLAC (I Feel You, Depeche Mode) 31.2 MB

    The following two pictures are here in order to illustrate that CD rips will look different depending on the source and the kind of music. The first one is a recording originally made in 1937, digitally remastered from the analogue master in 2000. It's the small jazz group Benny Goodman Quartet (clarinet, piano, vibraphone and drums) playing "Tea For Two". This is what old and most new acoustic recordings will look like. In other words, there is not necessarily anything wrong with a CD rip that does not show audio data all the way up to 22 kHz per channel. The second picture drives this fact home: it is Mitsuko Uchida playing the rondo from Mozart's KV. 545, recorded digitally in 1984 by Philips' finest.


    FLAC (small group jazz, 12.9 MB)



    FLAC (classical solo piano, 6.6 MB)

     


      
  4. LAME mp3

    LAME mp3, starting with the highest CBR setting, 320 kbps. The frequency has been cut off at 20.5 kHz - that is one reason for the comparatively small size. Another reason is the removal of some "unnecessary" audio data, that you can see in the stripey character of these pictures as compared to the original FLAC. The V0 that follows should be equal in sound quality to the 320 kbps CBR, but it is smaller in size because lower bitrates have been used for less complex parts of the audio, and the frequency maximum is at 19.5 kHz. Also, in the V0 you can see a "shelf" cut-off at 16 kHz, that is more pronounced in the "transparent" V2 (max frequency at 18.5 kHz), and prominent as a clear cut-off for the "acceptable" 128 kbps CBR.


    LAME mp3 320 kbps (I Feel You, Depeche Mode) 10.8 MB



    LAME mp3 V0  (I Feel You, Depeche Mode) 8.8 MB



    LAME mp3 V2  (I Feel You, Depeche Mode) 6.6 MB




    LAME mp3 128 kbps  (I Feel You, Depeche Mode) 4.3 MB

     
     
     
  5. Ogg Vorbis

    Ogg Vorbis can go up to 500 kbps, and it thus has a higher "best possible" quality than mp3, illustrated here by the q10 setting. Ogg Vorbis uses a different algorithm from mp3, and doesn't cut off the frequency for the higher quality settings: both the q10 and the "transparent" q6 shown here have more or less audio data preserved at 22 kHz, though much less for the q6. The "acceptable" quality q4, still reaches roughly 19 kHz, but you can clearly see that some audio data is missing, compared to the original FLAC.


    Ogg Vorbis q10  (I Feel You, Depeche Mode) 15.4 MB



    Ogg Vorbis q6 (I Feel You, Depeche Mode) 5.8 MB


    Ogg Vorbis q4 (I Feel You, Depeche Mode) 3.9 MB

     
     
     
  6. Transcodes

    So, encoding to lossy formats gives you the advantage of small file size at the price of losing some audio data. This simple fact should be enough to explain why you can not encode such files again, and get higher quality audio files. You simply can not regain lost audio data, not by increasing the bitrate, and not by converting to another format.

    The quality will actually decline in lossy to lossy encoding, even if you try to "increase" the quality by encoding to a higher bitrate, since the algorithm for removing "unnecessary" audio data is applied once again. If you encode from lossy to lossless, the quality will stay exactly the same, only the file size gets bigger. See the screenshots below. Yes, they look very much like the mp3 128 and the V0 above, respectively, but these really are screenshots of transcodes.

    Do not ever do it.


    128 kbps LAME mp3 transcoded to 320 kbps LAME mp3  (I Feel You, Depeche Mode) 10.8 MB




    V0 LAME mp3 transcoded to FLAC  (I Feel You, Depeche Mode) 30.4 MB
Lo-Fi Version Time is now: 23.04.2019 - 17:52