Before reading, please understand:
Digital sound works by sampling the amplitude of a waveform at given intervals (called the sample rate.) Each sample is a number - it literally means "at this point, the amplitude of the wave was at /this/ many decibels." When it gets played back, a device known as a digital to analog converter takes these numbers, and based on them, outputs voltages that correspond to the amplitude of each sample. The quickness at which it changes voltages (to depict changes in the wave) is referred to as the sample rate. This is much like how pixels on a computer screen represent a sample the color of an image at given points. Each sample is like a pixel. The more pixels you have, the more detailed the image will be. Bits are like the colors of the screen - the more different colors it can produce, the more brilliant and realistic the image will be. One might compare 16 bit audio to black and white, 24 bit to technicolor, and 32 bit being the full color range of a high dollar high-def TV.
Also, dispel your beliefs of "It all goes to 44.1/16, so why use a higher resolution"
Well, because there is only so much space for "information" about the audio. We'll call this information space "bits." The more bits you have, the more accurate of a depiction you will have of the waveform(the pure, unmolested, sound - which is literally impossible to reproduce exactly)
Think about how a story changes from person to person. The more specific the first person is about the story, the more the 2nd person has to carry on to the next, albeit with some distortion due to their own interpretations. However, when person 2 tells their story to person 3, with their own interpretations and additions included, there's even more information. Well, imagine if the story that person 2 told to person 3 was only as long as the story that person 1 told to person 2, but the story person 2 tells to person 3 has half of their interpretations and additions as well. This means that, they end up telling a story consisting of: Half of what person 1 told them, and half of what they interpreted on their own. If they had more time to tell the story, they would include even more necessary details - like more parts of person 1's story(things you /can/ hear.) If they had even /MORE/ time, they would probably give some un-necessary, insignificant information too(things that you can't really hear, and probably would not notice missing, but definitely sound better when A/B'd - like SACD/DVDA quality recordings.(many of which are bogus remasters, but nevermind that for now.)
This is exactly what happens when you run things through signal processors at inferior bit rates. There is so little room for information, that, not only does the insignificant stuff get cut off, the really important stuff does too, and you end up with something that doesn't make sense to your ears - and thus sounds unnatural and weird - or "unprofessional." Think of it like a quality bottleneck - if you have to reduce bitrates or sample rates, you're reducing the amount of information that the next person(plugin) has to work with.
NOW, for the information about signal rates and such:
Not sure if you guys were aware of this - I was not until skimming through the manual earlier, but:
Perhaps this could be the cause of some of Ableton's bad rap - people unknowingly downsample their project when they export, so it sounds bad when played back?Rendering audio from Live with a sampling rate other than the one that was used while
working on the project is also a non-neutral operation, and may result in a loss of sound
quality. It is recommended to always render using the original sampling rate, and then
convert the rendered file using a dedicated mastering application that is optimized for
these kinds of CPU-intensive, offline tasks.
I now know why my mixes always sounded so much better in the control room than in the car.
Anyway, if this is true, then it must also be true that:
When exporting in Ableton, selecting 44.1 from the export screen puts all of my plugins in 44.1 mode for the mixdown - whereas selecting 96k puts them all in 96k mode. There is a similar paragraph that states that changing the bit depth at this point is also a non-neutral operation.
The manual does not go into depth about the "In/Out Sample Rate" setting in the preferences menu, however, my tests have shown (one of my fx plugins tells me what SR and Bit depth it's operating at at any given moment - have not tested with an instrument plugin) that changing the in/out sample rate also forces all of my plugins into the SR/depth I selected. Mind you, it's labeled "In/Out Sample Rate" not "Run everything at this sample rate" - though it appears that is just what it does. "In/Out Sample Rate" - to me - would mean "I'm going to run everything as high quality as possible, then convert to this sample rate for output."
This leads me to believe that the same thing happens at rendering time if you select an option lower than the (SR/depth at which you listened) OR (SR/depth of the highest sr/depth capable vst you are running) - depending on how you look at it from Ableton's "neutral operation" standpoint. Either way, this poses huge problems for noobs who don't even understand that each sample is a measurement of amplitude - not a "little snippet of audio." Even worse, it doesn't appear to be properly documented(correct me if I'm wrong) - so this could cause problems for the pros unknowingly as well.
Bear in mind, Ableton will still SUM at 32 bit, but you have to go to a lower bit depth in the end anyway(how to do that is a whole 'nother debate.) Notice the difference between:
VSTi outputs 16 bit, Ableton sums @ 32
VSTi outputs 32 bit, Ableton sums @ 32.
Throwing any signal processor (Vst effect) in there will exhibit the same results. What you'd be doing is literally reducing the quality of the sound so much before it even enters the plugin, that by the time it's on the way out, it's so mangled (mind you, even more so, because it's still 16 bit) that it causes (semi-audible) distortion. By semi-audible, I mean, you probably wouldn't immediately spot it if it were there, but you'd definitely like it better if it weren't, and would undoubtedly know in an a/b comparison.
Also keep in mind, each plugin that you add to the chain degrades the signal quality in some way(normally un-noticeable and minuscule, or nonexistent if it's not actually changing the sound (like a disengaged EQ)) - though the real culprit for killing your sound here is downsampling(which it sounds like Ableton takes care of for you - you're stuck at a low sample rate) or BIT REDUCTION! If you're forcing them all to 16 bit, it would stand to reason that you're degrading the signal quality exponentially each time you add another one, because there's so much less room for information.
Remember, 24 bit contains 256 TIMES MORE information than 16 bit. Imagine being able to see in 5120/20 vision - which means, what the average person can clearly make out at 20 feet, you can clearly see at 5120 feet! 5120 is 256 x 20 - remember, we're saying, MORE clarity than 16 bit, not an absolute amount. Now, imagine if the image were to change just a little bit - but you're only 20 feet away now - with this super perfect vision - (realistically, it'd probably be out of focus due to a oddly shapen eye, but this is an analogy so hey) - you'd be able to see it 256 times more clearly.
So all this got me thinking - To achieve the best results(quality wise) I should either:
Render @ 96/32, then dither in another application to 41.1/16 (Redbook/CD format) or
Render @ 96/24, (with dither?), load it up into Ableton, and run the output of my interface into the input of my other interface, and re-digitize the analog stream at 41.1/16, rather than the original rate.
Ableton mentions in their manual that you should export at the highest SR/depth possible, and use an offline, dedicated application for these "post-export" dithering tasks. This is another sign to me that Ableton, when you adjust the sample rate, pushes everything into that SR. That being said, why would it not be a good idea to export from Ableton at 96/32, then put the file back in Ableton and dither from there?
Actually, what sparked this whole investigation is what happened to me in the studio the other day - I had some spare time, so I pulled up a track in Ableton that I'd been working on (at 96/24 listening,) routed that through the board, and went to the CD burner with it(not even exporting first - just hitting play and record) The CD was operating at 41.1/16 of course. When I popped it in the car later, I noticed immediately that it sounded worlds different from anything I'd ever created with Ableton - it was nothing specific to the song - I just was hearing details that I never before could. I immediately attributed this to the D/A > A/D conversion I did - which I normally do whenever I mix down a track(normally I do rock bands with multitrack tape or 96/24 hard drive though)
<EDIT> : Sorry for the ambiguity - I was routing the 2mix from ableton (Ableton did 100% of the summing/processing) into the board - not each individual track and summing them with analog. What I'm pointing out is that, rather than using dithering or downsampling to get to cd format, the analog waveform itself was "re sampled" at the lower bit depth.
A/Bing it with other tracks I'd made in Ableton made even more of an impact on the difference I could point out. I still need to try exporting it in Ableton and A/Bing it like that, but I haven't yet.
Have you guys had similar findings? Does this support your experiences with Ableton, or am I totally confused here?
)