Ableton co-founder and general visionary Robert Henke (also known as Monolake) gave a full-length workshop in New Zealand recently. If you’re up for 90 minutes of discussion of musical and sonic techniques in Live, plus a look at his unique Monodeck controller, the whole video is there. But that’s not the main reason the video is making its way around the Interwebs. It’s because there’s a bit of a bombshell right at the beginning of the footage:

He says, outright, you don’t need 64-bit sound to get “audio quality.” You don’t even need 16-bit all the time.

Okay, maybe that’s not such a radical thought in and of itself. Oh, yeah, except for one thing — the 64-bit summing engine he’s talking about happens to be the one in Ableton Live 7.

Video by Tom Cosm, via AudioLemon

Some people are already assuming this means Ableton has somehow betrayed them (well, in fairness, Robert does say the summing engine is just a marketing gimmick). And what about Cakewalk? Robert doesn’t mention them by name, but the only DAW that’s been trumpeting 64-bit mixing and signal processing is SONAR.

In fact, far from conflicting with Robert’s vision of sound, Ableton Live 7 really embodies it. And as for the Cakewalk thing — well, that’s complicated, because the term “64-bit” applies to a number of basically unrelated topics dealing with sound and computing. But none of that matters as much as one thing: if it sounds good, it is good.

Ableton’s Lo-Fi Voodoo — And Choice

Live 7 really does reflect Robert’s approach to sound and sound design. I say that not to try to make you go out and buy Live, but because I’ve been using Live nearly since it was released, and really enjoy the opportunity to get to use a tool that does have a point of view about sound.

Look at the other features in the program, like re-worked effects plug-ins. From the beginning, I think a lot of Live’s effects plug-ins have been badly misunderstood — a topic that came up recently on the CDM forums. The original reverb, the compressor, and some of the more recent additions are fantastic tools for coloring the sound of your work. Judge them as a conventional reverb or compressor, and frankly, they may turn you off. But forget your expectations there and look at them as timbral tools, and they’re great. (They also really do have a range — so once you know how to use them, you can make music that doesn’t necessarily sound like Monolake’s.)

Part of what Live 7 seems to be about is giving you a choice — if you need a great-sounding conventional compressor, for instance, and don’t want the pain of having to bring in a third-party plug-in, the new Live compressor is a valuable addition. But the old, odd compressor modes are there, too.

If you keep watching the video, Robert talks more about what he means — and it goes well beyond the issue of 64-bit summing. In synthesis, for instance, Robert demonstrates that using lower bit depth — bit depth below even 16-bit — can add harmonic content. A common, oft-repeated misunderstanding — one that’s been in the press a lot lately — is that lower bit depth contain “less” sound content. It’s just not true, ironically because distortion will add additional content to the sound. It’s a technique Robert uses regularly in Ableton’s Operator synth, and Ableton’s software does an excellent job of making these parts of the sonic palette available. But it goes well beyond Ableton; it’s part of why some musicians are embracing low-quality digital synths, chiptune music, and other technologies.

If you don’t feel like watching the whole video, by the way, one of Robert’s big tips is just to use this:

It’s the Spectrum tool in Live. (Naturally, there are equivalents in other programs and plug-ins.) What it gives you is a view of what’s happening in the sound itself, and couples something you can see with what you can already hear. That’s more valuable than anyone’s marketing.

What 64-bit Means

As for the 64-bit argument, we have to take at least a moment to examine what it is we’re talking about.

I’m equally dubious about just how essential having 64-bit dynamic headroom is. I’m not sure how valuable that particular feature is in Live 7; I’ll leave that to an engineer to decide. When SONAR came out with 64-bit processing, I talked to DSP engineers at AES at companies like Apogee. None could say they were really sure how useful that extra headroom is. There are a number of other sonic changes that do make the program sound “better,” so I don’t think the version 7 upgrade is snake oil, and I at least have the perspective of watching each upgrade since 1.0. And likewise, there are plenty of features in SONAR other than 64-bit mixing, so there’s no reason to get too hung up on this issue — for most people, I think it’s not even a relevant question.

But don’t write off “64-bit” yet, either. Cakewalk pushes 64-bit a lot; it’s true. But they’re actually talking about a number of different things:

  • Other processing/synthesis: Cakewalk’s products now support a fully 64-bit signal chain. (Digidesign I’ve seen pitching its bit depth and rate, too, but while they’ve broken the 16-bit barrier they’re not at 64-bit, so hence the focus on Cakewalk.) Now, maybe 64-bit isn’t useful in a summing engine — that’s debatable. But there’s also the question of what it does for certain digital processes, like signal processing and synthesis, and in that case additional headroom could have potential. Live 7 only deals with the summing engine. Just as certain image processing techniques use higher-resolution colorspaces, even if you can’t see what’s going on, it’s possible some techniques will make use of this additional information digitally to produce things you can hear. That’s a discussion for another day, though.
  • 64-bit memory addressing: More importantly, Cakewalk is also support 64-bit computing — something Ableton doesn’t do yet. On Windows, that means the ability to access larger amounts of RAM under Windows “x64″ or “64-bit” Vista. Totally unrelated to sound, but very, very useful for people using larger sample libraries. (It’s a non-issue on Mac OS X, because the Mac already addresses memory beyond 3-4GB without any extra effort.) I wouldn’t ding Cakewalk’s competitors for not supporting this, because the migration has been very slow for the whole Windows platform. But it is there — and it is a completely different subject.
  • 64-bit processing: Running under 64-bit Windows also squeezes a little more performance out of a CPU. The gain isn’t huge, but it’s real. This isn’t yet available on Mac OS (that is, only parts of the OS currently support 64-bit processing), though it is available on Linux. (Even on Linux, though, people often choose to run 32-bit versions of the OS for compatibility.)

That’s a gross oversimplification of some of those issues, so feel free to discuss the finer points in comments, but you get the idea. There’s 64-bit in a summing engine, 64-bit in the rest of your signal chain, 64-bit as a way of getting access to more RAM, and 64-bit as a way of making use of 64-bit CPUs.

And then there’s 4-bit as a way of making blippy bleepy sounds.

Low Fidelity: Not Just For IDM

That said, the original point really isn’t marketing, new features in Live or any other program, computation, or audio theory.

It’s about a simple point: what sounds good, is good. And if you want to sound good, you need to listen. If you keep watching the video, that underlies everything Robert is saying, and I have to agree.

It’s easy to watch someone using downsampling or lower-bitrate / lower-bit depth / lower-fidelity sound sources and think, oh, sure, if you’re from Berlin and want to do IDM all the time, that’s great. But as Robert points out, some of the music we love best wasn’t even recorded and processed using 16-bit, let alone 192kHz, 64-bit digital audio.

This very issue came up as we were talking to Hank Shocklee. Part of the early hip-hop sound, the Public Enemy sound, was a function of the lower-fidelity sampling mechanisms on early digital samplers. And these sound nothing like Berlin IDM. (In fact, a lot of the electronica sound and its approach to the technology is deeply indebted to American hip-hop and African-American artists.)

That doesn’t mean less is more, either, necessarily — just that it can be part of your sonic palette, of the choices you make when making your music.

Digital technology is now old enough to have a history; it’s old enough to have classics. I’m all for exploring ways of pushing the envelope. Want to really test whether higher bit depth, higher sampling frequencies, and new technologies can make new sounds? Go for it. There’s no reason you can’t do that, and still remember the possibilities and history at the other end of the spectrum (pardon the pun).

If it sounds good, it is good. And that’s the problem with talking about “audio fidelity.” More is not always more. One thing is not always better than another. It really is about finding the sound you want to get. Engineers can build the expertise to help you realize that — but no amount of science can say one sound is superior to another, any more than it can say that G Dorian scale is “way more awesome” than C# Mixolydian.

Once you realize that, too, engineers don’t become any less useful. On the contrary, making something sound good really is an art, meaning you really do need that mixing engineer, that mastering engineer, that live sound engineer. (I sure as heck do.) And you need to spend some of your energy trying to learn more about sound, because it’s something that’s worth devoting a lifetime to understanding.

Software companies still need bullet points on marketing sheets; that’s fine. But as for the fact that sound doesn’t easily fit into bullet points? I think that’s a good thing.

  • http://www.milezero.org Thomas

    You mean bit depth, not bit rate, don't you?

  • drumwell

    definitely "bit depth," or "bit resolution." bit rate refers to data transfer rate, ie, 256kbps for mp3.

  • drumwell

    great post, nonetheless. :D

  • Chris

    Peter, extra points if that Duke Ellington reference was actually a Peter Schickele reference.

  • Ashton

    I think it's funny that high bit rate is such a huge topic when the average dynamic range on most albums is less that 3db. I now it make a huge difference in the tracking and mix stages, but I've done recordings at 24 bit (PT) that always sound better(in my subjective opinion) than things my friend does in his 64bit cubase. I think a lot of times engineers tend to look for a piece of technology that is going to solve their audio problems and have them creating music that sound incredible. more and more lately I've been hearing a bit depth and sample rate debate and it seems to be one of those things where some engineers think "my stuff doesn't sound good because im not at 192khz 64 bit" when really it's because they don't know how to mic an acoustic guitar or because they don't know how to eq a kick drum. Higher fidelity is a great advancement in digital audio, but when will we be satisfied with digital audio? maybe with 384khz sampling freq? http://www.digitalaudio.dk/

    all this to say, (like what was said at the end) super high bit depth or the newest plugins can't replace good old engineering.

  • http://www.createdigitalmusic.com Peter Kirn

    Sorry, hit 'publish' a little prematurely and made a couple of corrections. That's why I'm a writer… tend to get on a roll to make massive mistakes in my first draft. Although this wouldn't disqualify from political talk radio…

    @Chris: let's say Ellington and Shickele, for some real musical inspiration.

  • http://myspace.com/themysterioush the mysterious H

    Audio fidelity has allways been a bit of a meaningless buzzword. Even in the day of all analog signal chains. The good thing then is that the science was less well understood, pretty much everyone knows what going from 44.1khz to 11.025khz sounds like and does. But not many people directly understand how chaninging the slew rate of an op-amp from .2ms to .25ms will change things. So back then people didn't bait with these statistics, they just said "we use High quality components" and left it to the buyer to decide based on all the unscientific garbage we use to decide things like this like whether the other guys are using it, if it looks cool, or the important one if it sounds good.

  • subbasshead

    As I understand it, the reason for a mix bus higher than 16 bit (or 24bit) is to cope with summing channels without clipping…

    There is a PDF on the digidesign site that explains their 48bit mix bus here:
    http://akmedia.digidesign.com/support/docs/48_Bit

    Presuming you are working with a 24bit session;

    "The extra 24 bits in the system are used to provide channel

    faders with additional dynamic range above and below the

    original 24-bit word, and it guarantees that the same fidelity

    is maintained when adding more inputs to the mix bus. In the

    Pro Tools +12 mixer, 9 bits are reserved for levels above 0 dBFS,

    providing 54 dB of headroom. This is enough headroom to

    allow 128 tracks of full code, correlated audio (imagine sample-

    aligned sine wave source files) to be summed with all faders at

    +12 without clipping the “input side” of the bus. It also provides

    enough bits below the 24-bit word to allow channel faders to

    be placed at nearly -90 dB before they stop contributing a full

    24 bits to the mix."

    The effect of clipping the mix bus is essentially the same as limiting

    - when you run out of headroom on a mix bus its like having a

    brickwall limiter across the outputs…

    & maybe thats why its not such an issue for many people – much

    of the music they listen to (& produce) has minimal dynamic range

    anyway…

  • http://www.createdigitalmusic.com Peter Kirn

    Thanks for that, subbasshead.

    But does that explain why you'd want to go to a full 64-bits and not just stop at 48?

    I guess there are two ways to look at what Robert (and others) are saying, as well — is there a reason for, say, 64-bit summing? Perhaps. Is it the reason people think it is? Probably not. (The sort of "it goes to eleven" non-logic, in other words.)

    But you could still have *some* dynamic range … quite a lot, in fact … and not worry about this, so I'm not sure it's just because everyone is squashing everything to within an inch of its life.

    Still, I would like to know what the merits of going 64-bit versus 48-bit might be, even theoretically…

  • http://fallsastar.com foosnark

    I've always thought it ironic that analog "errors" (saturation, various other kinds of distortion, out-of-tune oscillators, whatever) are considered "good" but digital ones are all bad.

    Yes, I'm aware that some of the "bad" is that aliasing is inharmonic, hard clipping tends toward brutality, etc. But these can be just as useful as a stomp box or mic'ing a guitar cabinet, and are easier to control.

    The same people who worry about jitter and dithering and so forth will go ahead and running stuff through old tape decks to make it sound "more analog."

  • efluon

    resolution in the summing bus is mainly an issue if the summing is done with fixed point, where you get clipping. this is only an issue for protools as all other DAWs use floating point.

  • http://gmm.ugress.com GMM

    Brilliant post Peter, thanks

    I find myself more and more often reaching for my old Protracker ST sample libraries, to use in Logic/Kontakt because of the delicious 8 bit 11 khz grittiness in the samples.

    It is a peculiar situation indeed, we finally have supergreat gear and equpiment, and we use this quality to enjoy and benefit the inherent problems we DIDN'T hear before :)

  • http://www.nicolasfournel.com NickSonic

    All this is very misleading. Using extra bits to have more room in the mixer is mosly a thing of the past, at least in the software world (i.e. not depending on a fixed-point DSP chip).

    Most recent software (if not all) is using floating-point arithmetic, in which case using 64-bit means using doubles instead of floats (32-bit). And you would have more than enough dynamic with 32-bit floating point already.

  • Richard Lawler

    "It’s a non-issue on Mac OS X, because the Mac already addresses memory beyond 3-4GB without any extra effort."

    64-bit address space support on the Mac is hardly a "non-issue". While it is true that Apple has chosen not to split the 32-bit and 64-bit versions of the OS into different products as MS has, the issues surrounding 64-bit addressing are very complex. Notice how few applications actually support a 64-bit address space. Also Apple completely backtracked on 64-bit toolbox support for non-Cocoa apps.

    Here's a blog post by an Adobe Photoshop developer on the issue of a 64-bit Photoshop.

  • Richard Lawler

    Sorry the URL didn't make it. Try this:

    http://blogs.adobe.com/scottbyer/2006/12/64_bitsw

  • ctx

    64 instead of 48 because that's what is natively supported by the processor. There are instructions for multiplying 32 bit floats, and 64 bit floats, but not 48 bit floats. Protools uses fixed-point DSP chips instead of floating point on Intel CPUs so the situation is different for them.

    Dynamic range is not an issue with 32 bit floats. There is no need for any more, or even as much as it has got. Your main source of distortion is that floats have a very wide range but relatively limited resolution; you might be able to represent 1234567 and 0.1234567 but not 1234567.1234567. If you try to add the two you might get 1234567.1 or even just 1234567 (the specific numbers are just for illustrative purposes).

    64-bit float improves that, but by definition the distortion is very low in level in the first place. Someone more knowledgable can correct me if I am wrong, but I would guess it might only turn the last bit or two of the fractional part into noise, so in the neighborhood of -130 to -140 db for 32 bit floats?

  • http://www.myspace.com/archiveofeverything Chris Blundell

    I think all of this is down to preference.

    I started using 24 bit and it just sounded different and lost something from the recordings. kudos for Ableton to allowing the choice to leave 64 bit mixing off (not sure on the others).

    I think its great Ableton has given the option to go 64bit if you're into that. But its great to be able to still have a strong amount of people still making music without falling into the "everyones doing it so if I don't I'm left behind."

    Hell, I know people that like to render stuff to 192khz mp3 and keep it in that format, even if they burn to CD (seriously!). Mp3's have their own sound too and it can work on the right tracks.

    Trying to get everything so clean and pristine can take so much away from music and people who still use tapes, vinyl and rerecording stuff through crappy pc speakers make me happy. Lofi can still have a place in a world of hifi.

  • efluon

    yep, as ctx said. and 24 bit-resolution is already beyond the hearing threshold (IIRC that's at about 20 bit), so a higher floating-point resolution is good for recursive processing only (like eg in filters), where errors get amplified. so 64 bit processing might really be an improvement for digitally processed fx, but for summing it is a myth. which is important, too. i recently read that more expensive placebos do indeed cure better than cheap placebos.

  • ifthe21stcentury

    I'm new to all this and I have question. I just got back from Justice and Diplo (great great performance from both). I heard a friend say in passing that Justice uses way too much compression, and in the pitchfork review of the new Clark album, Turning Dragon, they mention 'Justice-level compression'.

    Does any of this have anything to do with what is being discussed here (most of which is going over my head completely, 13 wikipedia articles open and all)?

  • efluon

    @ ifthe21stcentury: yes. this is related. insofar, as when you do compression levels like in recent pop/rock-music, when you do the conversion to integer, 8 bits would really be enough, as all the dynamics has been removed..

  • Craig

    @Chris Blundell – where is the option to switch off 64 bit in Live?

    A bit of history may be relevant here as Ableton have had a lot of bad press about audio quality, mostly due to users not fully understanding how the warping works. There have been countless debates on the Ableton forum with people performing phase tests and so on and on….so I think it may be important to realise that this was the backdrop for their decision to add 64 bit – it's not snake oil, just providing users with more than they need now and for the future, with the additional aim of trying to counter the persistent audio quality debates.

  • Doug Rouxel

    Any ideas on how the Monodeck II is scanning the session view to check which slots have clips loaded and which do not? or is it set up in advance via the Max Patch…

  • http://myspace.com/alxzndr Alex Laing

    Nicksonic ftw. As schenke's pointing out, a 64-bit bus is nice, but a summing engine is really not rocket science, even if it does have 64 lovely bits.

    The main advantage of anything over 16-bit for me is it gives me a bit more headroom for math/quantisation errors in vsts that don't get introduced as perceivable noise.

  • vvvoid

    <blockquote cite="Chris Blundell">Hell, I know people that like to render stuff to 192khz mp3 and keep it in that format, even if they burn to CD (seriously!). Mp3’s have their own sound too and it can work on the right tracks.

    I actually really like the way mp3 compression treats some material, particularly low frequency stuff and some transients – seems to "mush" it together in a pleasing manner.

    <blockquote cite="ifthe21stcentury">I heard a friend say in passing that Justice uses way too much compression

    Well that's a foolish thing for your friend to say – obviously Justice's slamming/pumping compression is an integral part of their sound.

    Objectivity is impossible when talking about sound "quality".

  • Craig

    @doug – set in advance

  • Sjoerd

    "the only DAW that’s been trumpeting 64-bit mixing and signal processing is SONAR"

    = wrong. REAPER was the first to do that, and still beats SONAR at it, with resolutions up to 64-bit floating point. It uses 64-bit native plug-ins too. Check http://reaper.fm.

  • http://www.createdigitalmusic.com Peter Kirn

    @Sjoerd: Yeah, but Reaper isn't going around making such a big deal of it. ;) I'm pretty sure it was SONAR Robert was describing. Reaper is excellent, too, absolutely. I like SONAR for its own reasons, even if it's more expensive… I think they're both worth a look.

  • mode

    Haven't had a chance to watch the video yet, but I just wanted to point out that DJ Shadow's classic first record, Entroducing, was done on an early, 12 bit version of the MPC.

  • http://jackit.sf.net/ Paul Davis

    @Sjoerd: actually, Ardour had a 64 bit mixer and signal path several years before Reaper was even started. We dropped it because on the CPUs of the time, the extra 8-15% cycle hit from handling doubles rather than floats was too much of a hit for too little sonic gain.

    @Chris Randall: "24 bit samples" and "64 bit processing" are not really related to each other. "24 bit samples" describes the format that your audio interface provides/receives. There is no h/w that provides even 32 bit samples at this time, and there almost certainly never will. Ask if you need to understand why, the answer is physics not technology. As Peter sort of indicated, the most important areas for "64 bit processing" are the bus/signal routing pathways used between elements of the processing chain, and the actual summing mixer itself.

    My own, decidedly non-double blind experience (you may therefore discount it entirely) is that computation with double precision floats *can* benefit audio synthesis, but has no detectable effect on summing mixers and even less on data pathways (compared to 32 bit ("single precision") floating point).

    Finally, I'd like to remind everyone that CPUs have provided double precision floats, which is really what is being discussed here, for a long time. All that is new on modern CPUs is their ability to more efficiently use 64 bit *integer and pointer* values. Great, except that almost no audio software uses 64 bit integer values for any part of the signal processing pathway. As a result, the real win here, other than some relatively minor improvements in the system memory bus bandwidths that benefits dbl-precision floats as a side effect, is the ability to address more memory. File sizes larger than 4GB no longer need fun and games (though they do need something other than RIFF/WAV format :) and a boatload of samples cached in 8GB of RAM are all addressed without MS-style segmented addressing. Sweet for programmers, and perhaps even manna for users.

  • http://www.audiodamage.com Chris Randall

    When did I post in this thread prior to this, Paul? Oh, wait, I didn't. I think you mean "Chris Blundell," although it's nice that you thought of me. But correcting me for the mistakes of others won't get me to say nice things about you. ;-)

    My own opinion on the matter is generally the same as yours; there is no audible difference between 32-bit and 64-bit summing busses, all else being equal.

    The problem is that hardly anything else is ever equal.

    -CR

  • http://www.digitallofi.com/words digital lofi

    Just thought I'd point out, nothing in Sonar forces you to use the 64-bit signal chain; it's easy enough to turn off. And honestly I think it's more of selling point to audio production house and film scoring studios, where high-definition audio is actually important. i.e. Going more for the Nuendo market than the Ableton market. Though I guess the lines are more blurry than ever.

    But I heartily agree with you, Peter.

  • http://jackit.sf.net/ Paul Davis

    And honestly I think it’s more of selling point to audio production house and film scoring studios, where high-definition audio is actually important.

    Man, these companies have technology users eating out their hand :) What makes you think that 64 bit versus 32 bit floating point is about high definition audio?

  • Pingback: digital lofi » BeatBurner: A Mini Review

  • http://www.digitallofi.com/words digital lofi

    @Paul Davis: Ignorance?

    I wasn't trying to make any sort of pro/con argument, and fully admit I'm sorta out of my depth here. I was merely making a comment on who Cakewalk marketing is likely targeting with this boast, other than power-geeks who just want the latest and greatest and the most perceived bang for their buck.

    I watched a video online somewhere that showed the inside of a large pro film scoring studio, where they had a server room of PCs networked, and a huge 128(?) channel board/controller. It was enlightening about who these companies are targeting with some these more high-spec features.

    But what do I know? I work at 44.1/24. I keep the 64bit signal chain engaged but more from inertia – really what harm can it do?

  • sph9000

    "You don’t even need 16-bit all the time"

    I totally agree!

    I actually really like low-fi sounds, and in these last times I do my best trying to sample old gear instead of using soft synths as VSTs in Live. I find the 80's electronic music recordings so raw, and sounding much better than modern electronic music.

    I understood the reason about this reading about recording techniques, and equipment used those times:

    8 or 12bit samplers, FM and analogue synths, analogue mixers, tape recorders…

    Is it possible to recreate a similar sound in Ableton now?

    Redux doesn't really match the sound I'm intended to get, which should be like an Emulator II (8bit 27Khz), or an SP1200 (12bit 22Khz), or an AKAI Mpc2000…

  • Pingback: Disquiet » tangents / Video Streams (Shocklee, Monolake, machinima …)

  • Pingback: Create Digital Music » Albeton Live’s Non-Existent "Secret" Vocoder; No One Needs a Vocoder

  • http://nein.com Phil Bajenski

    I really wish ableton would make live a x64 bit application because that way i can load many more of my live pa samples to ram.

  • doh

    "(In fact, a lot of the electronica sound and its approach to the technology is deeply indebted to American hip-hop and African-American artists.)"

    Fact? Deeply indebted? nigga please. electronica sound is completely INDEBTED to folks such as Bob Moog, Dave Smith, Miller Puckette, John Chowning…

  • Trance Pacific

    @doh

    Werd.