View Full Version : MP3 file size: how many MB/minute @ 320kbps?

06-02-2005, 21:55
Hi everyone...as the subject line says, I'm curious to know what the "rule of thumb" is when in comes to encoding MP3s @320kbps.

I know it will vary, but the rule seems to be about ~1MB/minute @128kbps, so would ~2.5MB/minute @320kbps seem reasonable?

06-02-2005, 23:17
The short answer: As you suspected, the correlation is linear.

The long answer follows:
Let's say X is the kbps of the file in question, which in your example is 320. Realize that kpbs stands for "kilobits per second". Your goal is to convert this number to "megabytes per minute".

There are 60 seconds in a minute, so we multiply by 60 to get X*60 kilobits per minute.
There are 1024 kilobits in a megabit, so we divide X*60 by 1024 to get X*60/1024.
There are 8 megabits in a megabyte, so we divide X*60/1024 by 8 to get X*60/1024/8.
Simplifying, we get X*60/1024/8 = X*60/(1024*8) = X*60/8192 = X*15/2048.

So, a file that is X kbps is X*15/2048 megabytes per minute. Plugging in your example of 320kbps, you get 320*15/2048=4800/2048=2.34375 megabytes per minute. For reference, 128kbps would yield 128*15/2048=1920/2048=0.9375 megabytes per minute.

I realize that I could have made the reasoning behind those calculations clearer, but I don't have enough time to explain it correctly right now. If you want further explanation, let me know, and I'll see if I can explain it further when I have the time.

06-02-2005, 23:21
Thank you Tempus, that was really helpful. I appreciate it! 8)

06-05-2005, 12:04
OK, just curious now, how would these files sizes and bitrates compare to OGG?

For example, how many MB does one minute of music take up at OGG Q8 or Q9?

06-05-2005, 20:12
That is a LOT harder to say with any kind of certainty, and here's why. Ogg Vorbis is capable of encoding at a CBR or VBR bit rate you specify, just like MP3, but it is recommended that you instead use a Q (quality) level. The reason is that a Q level goes one step beyond VBR. I will try to explain the difference between CBR, VBR, and Q levels as I understand them.

In the beginning, there was only Constant Bit Rate, or CBR. This encoded each piece of a song such that it was the same size, as determined by the bit rate specified at the time of encoding. But what if you're listening to a song that has some parts that are easy to encode, such as a momentary silence, and some parts of such complexity that the loss inherant in any lossy compression format becomes obvious?

Enter Variable Bit Rate, or VBR. This set out not to make each piece of the song be at the specified bit rate, but rather to let each piece be at a different bit rate, provided that the average bit rate for the entire song is the one specified at the time of encoding. This means that the encoder has to be able to judge which pieces of the song are easy to encode, and which are difficult, and therefore likely to reveal the loss inherant in lossy compression. It then encodes those parts of the song at a higher bit rate, and encodes the simpler parts at a lower bit rate, resulting in a file that sounds better at the same average bit rate, and therefore the same file size. It's a more intelligent way of allocating resources.

Q levels take this principle of resource allocation to the next level. VBR only lets you shift resources around within one song, but what if you're encoding more than one song? Let's say you're encoding two songs. The first one is mostly simple, and easy to compress, but the second one is entirely complex. To compess these songs with VBR still wastes resources because the second song needs a higher bit rate than the first, but they are still both assigned the same average bit rate, which makes the second song sound more lossy than the first, with more obvious compression artifacts.

But by using a Q level instead, you are not speficying a bit rate at all, but rather a level of artifacting that you find acceptable. This means that in the previous example, a lower average bit rate can be used for the first song, and a higher average bit rate can be used for the second one, because the second one needs more bits to be faithfully reproduced.

This is a better way to allocate resources than even VBR. The only down side is that when someone asks you how many megabytes per minute is taken up by Ogg at Q8 or Q9, you don't have a good answer. :) But I just did a little test, using the aoTuV b3 Ogg Vorbis encoder from rarewares.org, and encoded an album at Q8, and again at Q9. Q8 averaged about 1.757 megabytes per minute, whereas Q9 averaged about 2.254 megabyes per minute, but as I believe I've already made clear, it will vary from song to song, and album to album. It all depends on what you are encoding.

Your best bet to get something approaching a firm number is to run the tests yourself, using your music.

06-05-2005, 20:20
Wow, thanks a lot! That's a lot of info, thanks for taking the time for putting it together for me. I'm definitely bookmarking this thread.

Five gold stars for you! :)

06-09-2005, 18:59
...when in comes to encoding MP3s @320kbps...
If you're encoding at such high bitrates you might be interested in FLAC.