Saturday, February 16, 2008

The reason not to use low MP3 compressions!

Notice : this aricle is my 2nd arcticle on mp3 tech and appeared on 31-5-7 on multiply site . The main reason writing that article is that many  people  are using  low level MP3 compressions (as 192 adn 256 )  that i do not think  are  so necessary  since per  my measurements  128 kbps are more than enough. To my year  wma's  of the level of 48 kbps  have neatly the  same level .

--------- Given that many users on Multiply rip or covert their music in quite big files of over 7.5 MB fr a 4 minute song ,i decided to write this article to explain why it is not necessary to use very high sampling rates.

At same time I will notice to you why I use the WMA format in 48 kbps to include mass material into a small MP3 player as shown in my previous reviews

It is important to advise you that I was in the audio technology and HiFi's since 1980 , including also the MP3 tech form ca 1997 and my first ever MP3 player was with CD on 2000.Also i used several professional equipment the older days and I know very much about it.

This is a abridged version of my blog in with a little more explanation in the pictorials If someone is still very boring to read this whole message ,please check the flies on the for his personal evaluation , though it is very worth of reading

And here is the basic background for the above theory .

1.Make a file with white noise . This noise is used primarily for testing purposes in audio laboratories. it can be used to measure the linearity of audio systems including the audio part of radio transmitters. The easiest way to make this white noise is with the Cooledit program.

For best measurements i made a file of 60 secs making a file of 10M336B [10336 Kb]I also deliberately use the linear frequency response in order to find more easily the cutoff frequencies than with the standard logarithmic freq response

click to view the original file This is our original WAV file's frequency response , designed from a white noise of 0 db level. As you may notice it is a totally linear, and will be used as reference to the results from various compressions shown later

2.Convert the files toMP3 and /or WMA in the most popular conversions . I choose the MP3 format on 128 [standard]192 [broadcast mode ] and 256kb [store mode] while for WMA the 44 [FM] ,. 48[I say just CD] 64 [default near CD ]and 96 [CD ] kb.

I always used the Musicmatch Jukebox in its 6th version which is very light comparing to the newest versions and has very good results especially if converting from CDs and MPs to WMAs . And though Cooledit has MP3 save option

3.View the resulted files back to Coledit's frequency analysis screen


256 kb
As you will see the 256 kb sampling rate offers a very high audio bandwidth of ca 21 kHz filtering out everything above this frequency ,with more than 60 db cut off .But these cut off frequencies are inaudible to the ear ..
Tthe file size is 1876 KBytes
click to view the original file

The curve here goes to the 20 kHz , and about 60 db cutoff above these frequencies. This is the maximum the ear can listen The file size is 1407 Kb
click to view the original file
128 kb
And here is the standard MP3 with strong cutoff above 15.900 Hz . I do not see any significant difference between 128 and 192 in my ears though still 112 with a cut off at 14000 and 96 kb at 12000 sound nearly the same
The file size is 938 kb on the 128kb

WMA s in discussion

AS Cooledit cannot open WMA files , i used first a third party program to convert the WMA files iinto WAV and then opened with Cooledit , and here are the results :
click to view the original file

48 kbps

48 64 and 96 kb

Here all the WMA files have the same audio bandwidth at 20 kHz with the only difference on the dynamic level. ,
click to view the original file

64 kbps

The dynamic level is ca 36db , for the 48 kb , ca 40 db for the 64 kb and ca 55 db for the 96 kb compression. The 40 db are the audio levels used by most music groups. More than 60 db are used only by classical music And the file sizes are 367,484 and 718 kB respectively . IN practice that means that all these modes can offfer just high quality audio though very low audio levels are 'removed' .
click to view the original file

96 kbps

In a very careful listening you can see some minor artifacts on the 48 kbps especially in drums and high trebled audio.

click to view the original file

WMA 44

And here a funny case of the 44 kb WMA. Instead of having a smooth cuve , that reminds a triple 'treble control' , a part of the bandwidth is somehow cut out

However this happens with the MMJB 6 . Using another converter as for example the MP3WM OGG converter, the results are also diferent . The results in the MP3 conversion showed me a flat bandwidth for 128 to 256 to 22000 with 96kb having -6db on 16000max .while the WMA compressions rates were showed as a bandstop filter analogous to the compression rate ...

Furthermore these are the results with the over mentioned program . I mean , using a different program to convert files gives a different result
AS for example Cooleditwith its MP3 plugin ca offer linear response to the maximum of the bandwidth.

At another example i was also using the ACE mult iformat converter . The results once opened or converted to WAV (for WMA) have made several interesting results . Please see the picture below, it is an animation!


From the above results , it is clear that ACE can technically offer better audio than MMJB on the MP3 format , but worse than MMJB in te WMA format

No comments:

Post a Comment