@holster asked in a post if there were any preferences as to 128kbps vs 320kbps and I thought that it might be worth discussing on its own.
Apart from the resultant file size I have no real problem either way. I want to say that I can hear the difference. But I guess I can’t.
I’ve been rendering my mp3s at 128 for so long (default in reaper) that I never even considered that there was a better alternative until I read the posting instructions for the manic mix contest.
24bit wav, 16bit wav, 44.1… etc. I always figured that if my mix can’t cut it in an mp3(128), that it’s my problem, not the codec. But out of interest, I threw 3 renders of the same mix into my daw and put them to the test. A 44.1 24 bit wav, a 128 mp3 and a 320 mp3.
First off… none of them will null with eachother when phase-opposed. The 320 against the wav leaves a fairly clear resultant that is consistent and just sounds like a band limited, quiet mix. The 320 against the 128 leaves a phasey, spikey mess that is all high end and modulates wildly.
What does this tell us? I don’t really know and I don’t know if I care… But since the null resultant is what we would be missing, chop logic wants me to believe that I’m glad that I don’t have that mess in the 320… I know it’s wrong on so many levels. But you can draw your own conclusions.
Even more interesting (if you can believe it)… I put an analyzer on all of them
The 128 mp3 has a distinct freq cutoff at ~17K the 320 at ~21K and the wav may or may not have any. I low passed this mix at 22.5K. Can I hear it? btw the wav and mp3, yes. btwn the 128 and 320… not so much.
Last thing (really) I did notice that even though I brickwalled my mix to -1db… the 320 mp3 did hit -0.8db inside the daw whereas the wav and 128 mp3 stayed at -1. Did the 320 have 0.2db of ISPs? Are 320 mp3s more prone to ISP? Don’t know.
Bottom line is I will probably never worry about it and will probably go back to 128kbps for my mp3 renders. If I can’t get an mp3 to sound “good”, I need to work on my mix and not this nerdy stuff…
have fun
rich
I think I can hear a difference between a 120 and 320 mp3…I never rendered directly an mp3 from my DAW. But any Wav-Mp3 converter I used made the difference quite obvious.
I don’t know if a DAW create a better 128 that any converter…maybe…adaptive encoding?
PS: I’m very sensitive to high frequencies and this is mostly in the upper range that I can hear a difference…
Sounds about right (no pun intended). The .WAV format should not truncate any data. The .MP3 format is “lossy” and is truncating data, resulting in lower (perceived) quality. It looks like you found the big difference between 128/320; it’s doing more low pass, cutting off (truncating) the top end of hearing. Most anyone older than a teenager has begun to have some hearing loss in the upper register, or wouldn’t notice the “missing” information except in a comparison - and perhaps not even then as you found out. So that’s probably where the algorithm cuts the most non-essential data.
I believe there is something in the MP3 conversion algorithm that “oversamples” or results in “overs”. I didn’t know that might depend on your bit-rate for the MP3, but I know it’s recommended that when converting to MP3 to give a fair amount of headroom to compensate for this tendency to go “over” the source peak. It shouldn’t happen with .WAV IIRC.
Yes, some people with high-end hearing still intact and trained ears may be able to tell the difference. Just make sure you are testing it ‘blind’ to avoid “confirmation bias”.
That’s a good question if it matters converting the MP3 from the DAW source or an exported WAV file. I think the MP3 encoder is pretty standard, and I’d guess the results should be pretty consistent overall. However, I know at least two encoders are Fraunhofer and LAME, and they may work differently.
I can absolutely hear the diff between a 128 and 320 mp3 render, and my high-freq hearing is shot-- my sensitivity above about 10K drops like a rock. But even I can absolutely tell the difference in just a few seconds.
I’m a Mixcraft user, and it does the mp3 render natively, I don’t have to render a wav and then convert. It uses the venerable L.A.M.E. mp3 codec.
Yup… mp3s can have intersample peaks, hence the recommended precautionary extra allowance.
But oddly enough, the 320 version was the only one that showed any… strange.
Thanks for the input guys…
Listening back to some other material 128 vs 320, I’m starting to hear a bit of a difference.
Maybe I’ll set the default in my daw to 320 and be done with it.
But I still view an mp3 as “disposable” or just a quick and dirty medium. If the mp3 sounds good, then the mix is good.
My main render is always a 44.1 24 bit wav unless there is some other requirement.
Have fun
rich
You know it is real interesting I’ve always did my rendering in 320k knowing it compresses the least but not until I got ozone 7 and heard the artifacts it induces at 128k VS 320k really cemented my ears to what I am hearing in fact it put unwanted noises and sounds into your mix that I find very unappealing.
OK… New data here. Take it for what it’s worth.
I dumped a new 44.1 24 bit wav, a 320 mp3 and and 128 mp3 into the daw again and ran phase opposed, null tests and here are the resultants:
These are all NASTY sounds. So don’t crank your speakers!
320 mp3 against 128
WAV against 320
WAV against 128
Interestingly enough… the 320 mp3 had overs again whereas the WAV and 128 did not. I made sure that all the mixes engaged my limiter. None of them should have gone over -1db.
But the 320 hit -0.8 again. Go figure…
I’m figuring it’s not what it puts in but what it leaves behind. If you listen to my nulls in an above post, you’ll see what I mean… The 320 vs Wav seems to indicate the compression ( stuff it discards) works smoother and more uniform, so what’s left is smoother.
The 128 is a jigsaw blade in comparison…
Trying to null these files against each other doesn’t really tell you much. If there is a phase shift (time shift) of any amount due to conversion, when nulled against the original it will be a disgusting mess since the the waveform is literally not in the same time even by fractions of a ms. You can take the same original file and shift it 1 ms then null and you will get a bunch noise. One difference is in where the cutoff filter is applied and as you noted the 128 k is at around 17k in your example which is lower than with a 320 k file as you again confirmed. The biggest difference will be in the aliasing caused by the extra compression in the 128 k. This is very noticeable, some songs more than others. Another thing is although there is no cut off filter on the low end (or if there is one it is very close to 20hz) the extra compression really cause aliasing artifacts which affect other octaves above the compressed bass frequency. This also happens in the 320 version but much less as the compression is applied a little gentler. I use 320 mp3 for pretty much everything, except in a mixing session where I use 24 bit wav, since the extra bit rate is important when doing all the calculations from plugs. Once the mix is done a 320k mp3 holds up very well for most music.
Yes… I understand how it works.
But it was pretty surprising how “recognizable” and stable the wav to 320 was… ie: less phase distortion. That’s enough for me to start to care about mp3 “quality” if that’s really a thing… even if the “science” is wrong
have fun
rich
in ozone 7 there is a codec tab that lets you here your 2bus in mp3 or aac format and you can listen from 96k to 320k you can hear the artifacts or noise each one adds
ah that makes sense now!! like I said I never really cared about it till I got ozone 7 and stumbled on to this I was really surprised at the difference could be a useful tool if you find your adding or in what it leaves behind you could make correction to bring less of that into your final mix.
But since the null resultant is what we would be missing, chop logic wants me to believe that I’m glad that I don’t have that mess in the 320… 
