The effects of so-called 'lossless compression'
Over the years there have been many audio formats created and within them a great number of options.
Understandbly it can cause confusion as to what type of master to provide us with when producing a bulk run of CDs.
Most PC digital recoding systems will either work in WAV file format or a proprietory format. Most Apple based recorders will work with AIFF's.
Within these formats can be a wide range of "Sample Rates" and "Bit Depths", however for the sake of brevity a CD Audio disc is replicated at 16Bit, 44.1Khz Stereo. As long as the final format is compatible with that then there should be little problem in producing your CDs.
Where we encounter problems is when listening copies have been produced as Mp3's or M4a's (Itunes) files. Often we are asked if they are "good enough" to make cd copies from. The simple answer is yes you could, but we'd prefer not to! Even though people tell us they are "CD Quality Mp3s".
The following is an explanation and practical demonstration of why we want to work with the highest quality recording you have of your music.
What happens to your music with Mp3 and m4a Conversions?:
Many years ago during the infancy of the "Internet" when 28k & 33kbps modems were common, large files were difficult to distribute, Audio compression techniques were developed to reduce the size of files to make them easier to send. Typically by converting a Wav file to an Mp3 the file was reduced to about 1/10th the size of the original. So a 4 minute song of approximately 40Mb size was reduced to about 4Mb and therefore a much more manageable size.
The methods for doing this "compression" work is housed in "Mystical Maths" that I'm sure you can research on Wikipedia if you feel the need!
Over time the maths has got smarter and the impact on the original audio has got less noticeable, BUT it still affects the original audio and hence the reasons why we don't like to use them for your product masters.
The essence of this article is to try and explain the effect that producing an Mp3 or M4a of your music has on your original audio. It does require some pretty tricky technical stuff so I hope you can stay with me. I'll accompany the article with some audio & visuals so that it helps explain techy stuff.
I've used a track from Dave Gilmour to highlight what happens when downconverting music to other formats, no particular reason - I just happened to have the CD at hand and know that he always produces music with excellent production values.
I've used NCH's 'Switch' programme to do the necessary conversions and then imported them in to my Pyramix system so that I could do the technical & analysis work.
Step 1
I've produced a "direct" copy of the CD as a 44.1Khz 16 Bit Wav File. Download & play the Wav File Sample
I then produced a 128Kbps Mp3 (from the CD) Download & Play the Mp3 Sample
I then produced a 320Kbps M4a (from the CD) more comment on this later*
Note: the audio files may take a while to start.
Observations:
At first listen all versions are reasonably similar, with perhaps a small reduction in the top end and less spatial imagery with the down converted files. All in all the kind of audio that could be expected for a "convenient" small audio file.
Step 2
These 3 tracks were left in their native formats and imported in to the Pyramix Native Version 8 as stereo pairs on the timeline. See image below.
WAV file in tracks 1&2
MP3 file in tracks 3&4
M4a file in tracks 5&6
Each compressed version of the song was lined up to be frame accurate to the original WAV file and all tracks are normalised at -3db, basically to give me headroom while working on the project and to ensure that minor differences in levels caused by the conversion work are eliminated as much as possible.
Pyramix capture layout (Click for larger image)
Step 3
Now for the tricky stuff!
In order to work out what effect "data compression" does to a WAV file we have to subtract one file from the other. So once the files are lined up precisely, the Mp3 file is 'reverse phased' compared to the Original WAV file. When the two versions are played at the same time this process 'subtracts' the Mp3 audio from the audio WAV file. The audio is "captured" within Pyramix and produces the resultant audio content that the MP3 processing takes out of the original track. Confused?? you'd better read that again!
The same technique was used with the m4a files and the difference audio captured. The audio files have been left at the "captured" levels so you can hear the relative difference of the processes. For those that prefer to 'visualise information' waveforms of the difference files can be compared in the images below.
The captured difference files can be heard here:
Audio result of Mp3 processing
Audio result of m4a processing
Mp3 processing - difference file
m4a processing - difference file
Step 4
As a matter of interest I also normalised the difference files to the same level as the Original Wav file for a couple of reasons. (They can be seen in the Pyramix image in tracks 1&2 following the main wav file).
1/ It's easier to hear the affect of the audio processing.
2/ It gave me a qualitative indication of the amount of signal removed from the original audio.
In this case the Mp3 processing removed approximately 7.6db LESS audio than the m4a. (that's roughly a quarter). You can do the same thing in a simple audio editor if you've downloaded the audio samples.
Observations:
As I mentioned in Step 1 the m4a processing was done at a higher sample rate (320kbps) as opposed to the Mp3 which was done at 128kbps. In theory the higher sampling rate should have had less impact on the original audio, so from the results produced it appears that this particular form of encoding is not as good as Mp3.
When listening to the samples it is evident that the m4a processing takes out more of the original content in general, both level wise and frequency wise. It certainly affects the mid range more, where it takes out a "chunk" of the guitar solo! There also seems to be more reverb or spatial imagery in the m4a difference file.
The result of Mp3 processing appears to be more rhythmical and "gated". So the processing is confined to smaller 'windows' or areas of the original audio. The overall amount of the difference signal is about 7.6db less than the m4a processing.
It would certainly seem that overall the Mp3 processing algorythm is more music friendly than the m4a (At least with the software I'm using). The other practical consideration is when you do the audio mastering stage. Do you add more reverb because you know you are taking some out if you distribute "online" or do you produce content for the CD listener?
Would I use either for a master or even a decent listening copy NO! I just don't like to be aware that I'm missing out on something that the Musicians, Sound Engineer or Mastering Engineer have taken hours or weeks and lots of dollars to carefully put together.
I hope this helps explain what downconverting does to your music and please don't hesitate to contact us if you have further questions.
If you'd like to discuss a CD project you can contact us easily by calling 1300 4 PROCOPY or 08 9375 3902 or clcik a link below:
Get a Quote :: Place an Order :: Ask a Question :: Contact Us