There will be no judgment in this post. No sound snobbery. I’m simply asking the age-old question: At what bitrate should we encode MP3s? And I need your help.
This test is occurring in two parts. In part one, I’m sampling three songs chosen from vastly different genres, encoded from CD and transcoded into the various popular bitrates available for MP3s (64k, 96, 128, 160, 192, 256, and 320kbps with VBR off). I tell you what I hear, then you sample the files yourself, and tell me what you hear.
Part I – My Test
I’m sitting here with Pioneer’s brand-new VSX-1019AH-K receiver, a $US500 model that actually pulls the MP3 data off of USB drives and iProducts for decoding within the receiver itself. (According to Pioneer, this “Advanced Sound Retriever” technology restores sound lost in the MP3 conversion process, so I figure it’s the best MP3 experience I’m gonna get.)
The sound is being sent through 14 gauge Monoprice speaker wire to twin Definitive Mythos STS Supertowers ($US3,000/pair). We wanted to assemble an ideal, nice home audio system that could make MP3s sound their best. We feel that this combination of superb speakers and MP3 decoding reaches a reasonable benchmark for the reasonable enough price of around $US3,500. Since most readers including me aren’t going to run out and buy anything nicer, it represents a decent ceiling of audio quality.
Pure Prairie League – Woman
My first pass, I couldn’t hear a difference beyond 128. And it was a little worrisome. But no judgment, that’s the rule! I took another pass…things did seem to get better…but was I imagining it?
So I skipped from 128 to 192. Then I could hear an improvement as the instruments were unchanged but the vocals grew more lifelike. Songs encoded beyond 192 sounded different in terms of balance, but not necessarily any better. I wonder if, since the song was “digitally remastered,” studio technicians compressed the audio to begin with.
My conclusion: 192
Gorillaz – Feel Good Inc.
It was a total shock. I could hear the differences in bitrates, all the way to the top, the first pass through the list. I had assumed, whatever, some electronic type music. It won’t matter. But even the jump from 192 to 256 was dramatic on my system, with every enhancement giving me more detail in the laugh and a richer, wetter bass line.
My conclusion: 320
Bizet – Carmen Suite #1
During my quick first pass, I didn’t hear a difference beyond 160. Skipping intervals, I found no improvement going from 160 to 192, but a noticeable improvement from 160 to 256. The middle just feels fuller, with a far more lifelike reverb to the low to mid horn section. I’d like to say that I heard a difference up to 320, but I’m willing to chalk that up to the power of suggestion.
My conclusion: 256
Also, I compared the 320kbps recordings to their uncompressed WAV counterparts. The only difference I could hear was in the Pure Prairie League’s Woman. The vocals and high level instrumentation felt ever so less harsh. It’s a bit ironic, as that was the song I had the biggest problem distinguishing bitrates in the first place.
Back when I tested my ear in college, I found the cutoff to be 160, and have since encoded all of my music at that level (though it’s become less of an issue now that MP3s are more often downloaded than ripped from CDs). Now, however, it’s pretty apparent that with more hard drive space and a nicer audio system—my earlier testing was just on a set of decent computer speakers—it might be worth reassessing my encoding rates. In just these three songs, I found a huge fluctuation, and not in any way I intended. Honestly, I figured that Carmen would require the best bitrate to assuage my ear.
Now, I wouldn’t encode lower than 192kbps, and I’d be tempted to push the boundaries to 256kbps and 320kbps on the music I planned on listening to very closely, though my laptop’s hard drive would probably hate me for it.
Part II – Your Test
Enough with me talking, now it’s your turn. You’ll find the files you need below alongside an accompanying poll. Please don’t vote based upon past experience or my subjective impressions, and feel free to test on any system you like (as long as you note it in the survey).
Oh, and the easiest way to peruse the files quickly is to click the first audio link, let it load in your browser, then just change the bitrate number in the filename up in the address bar—fast and easy to do any side-by-side comparison you like. Well, at least on your crappy computer speakers.
TEST FILES
DOWNLOAD THEM IN ONE BIG ZIP HERE (MediaFire), or use individual links through your browser below:
Woman 64
Woman 96
Woman 128
Woman 160
Woman 192
Woman 256
Woman 320
Woman WAV
Feel Good 64
Feel Good 96
Feel Good 128
Feel Good 160
Feel Good 192
Feel Good 256
Feel Good 320
Feel Good WAV
Carmen 64
Carmen 96
Carmen 128
Carmen 160
Carmen 192
Carmen 256
Carmen 320
Carmen WAV
And here is the survey (it’s a popup):
matt
Friday, May 15, 2009 at 9:43 AMencoding at 192 or above is a good conclusion. personally, with hdd space so cheap, i just rip everything to lossless so i don’t need to worry about whether the encoding is degrading the sound.
but id say the quality of the master is the biggest factor, i’ve been amazed at the clarity of 128kbps encodes of both Bomfunk mc’s Freestyler and Hotel California live, sounded alot better than some of my lossless tracks.
simulacrum
Friday, May 15, 2009 at 12:31 PMFaulty test conditions. He should have listened to all the versions of each file without looking at bitrate, then written them down in order of preference and only then looked at the bitrates of the files and checked whether his preference correlated with the bitrate
Anon
Sunday, May 17, 2009 at 4:56 AMI’m sorry, but this is a meaningless test. Anyone who knows anhything about encoding mp3s knows that VBR+LAME is the way to go. A better test would be the various VBR settings and see if anyone can hear a difference.
Daniel
Saturday, July 4, 2009 at 9:32 AMThe test is an interesting idea, but there’s one big problem with it: everyone’s going to be testing on their own equipment and it’s all going to be very subjective.
As far as the science goes, above 112 kbps will always give a faithful reproduction up to around 15kHz. Above that frequency, the higher the bitrate, the better the reproduction until at about 320 kbps it hardly makes a difference (which is why most encoders go up to 320 I guess).
As 15 kHz is about as sensitive as most adults hearing gets, most people also quote 112 kbps as the low end of acceptable MP3 encoding.
So, if you want the best space:quality ratio, go for an average between the optimum (320) and the low end (112) which is 224 kbps.
Personally I can sometimes hear flaws in a 128 kbps encoding, very rarely in a 160 kbps and find 192 kbps the ideal bitrate.
But it all depends on what kind of system you use and also if you use DSP.
For example, a 2 channel 128 kbps MP3 going through a 5.1 system emulating 5.1 surround sound
sounds atrocious compared to when it’s played in ordinary stereo. At 5.1 surround, even 160 & 192 kbps bitrates will produce artefacts in the rear channels.
As I use a 5.1 surround system with a DSP to emulate surround I now always encode at 320 kbps to reduce the possibility of artefacts as much as possible.
For 2 channel reproduction, 224 kbps will definitely banish any chance of hearing artefacts, and 160 kbps is more than satisfactory 98% of the time. I’m mostly a fan of less complex rock/pop music though. But I also like much more complex dark ambient and this kind of music definitely needs at least 224 kbps due to the complexity of the signal. At lower bitrates you once again get artefacts when listening to complex classical or ambient music.