64-bit summing engine - I'm gladly eating my words

[ilia]

kb420 - not a bad idea

rosko - until Ableton posts their white sheet we can't say for sure


in Live 6 I took a dozen of those waveforms, dropped them all by -15dB and recorded them into a clip. I saved the file and re-opened it in L7 and did the same. then I inverted the L6 clip and recorded the output. it's very small, you can hear the original in it. with this the dithering process is out of the equation. all I can say is that there's a difference, better or worse by how much, dunno.

below is the file I used, it's a Live 7 file and can't be opened in Live 6.
http://www.4shared.com/file/25907447/63 ... _test.html

Nice test.
You are doing something for which 32-bit summing is adequate.
The difference you are seeing is likely an indication that beyond simply changing precision, they took apart their summing/dither algorithms and overhauled them.

I'd also check to see if record via resample makes a difference wrt render.

[Hardtoe]

Well, from using my ears (outrageous I know...) I clearly agree that Live 7 sounds much better - it is obvious even on my Event 20/20 Bas's with a motu Ultralite sound card. I sure it will be even more obvious on higher-end pro setups. There is really a BIG difference (like when I upgraded the 828 mkI to the ultralite - I wasn't expecting it, but it was like night & day))

Cheers

Orion

[Tone Deft]

I'd also check to see if record via resample makes a difference wrt render.

render will make a difference, it's shaped random noise added to the signal. the test I did was track routing, dunno if that's the same as the resample feature.

[ilia]

the test I did was track routing, dunno if that's the same as the resample feature.

probably (hopefully) same summing.

render will make a difference, it's shaped random noise added to the signal.

why would render impose a particular dither/noise shaping strategy? -- either way you have to come down to the lower res of the rendered/recorded file, so as long as you are not doing two passes (normalize on) or explicitly requesting dither, the mix should be the same, I think. And if it's not, we should know about it.

[Timur]

I'd also check to see if record via resample makes a difference wrt render.

render will make a difference, it's shaped random noise added to the signal. the test I did was track routing, dunno if that's the same as the resample feature.

Could you please zoom into the waveform at one of those points where the the inverted signal still produces clearly audible sound? That should reveal a spot like the one I have posted above where you could check wether there are time shifts or not.

I also had a look at the beginning of the two files I checked where most of the signal is nulified by the inversion. There I noticed what I would expect from a simple bit-depth change, namely that the L7 signal shows a higher resolution aka still tiny steps where the L6 signal is already flat.

But as you can see in my above sample of the snare drum in this range even the synchronous parts of L7 look quite different from the ones in L6.

I'm not sure yet if L7 sound so much better than L6 when doing an a/b of the latest samples by Nebulae (thanks for that, pal! ). But my ears are not exactly trained for such a task as I am no audio-engineer, but it's not like a night/day difference which strikes at you "clearly". Also I did not spend much time on that but took a "visual" approach instead.

A simple way of doing an a/b is to load both tracks into Live and mute them, then switch from one to the other via solo buttons.

[philipc]

Wouldn't a blind audio test reveal more practically speaking than a summing test? After all, if a summing test reveals differences it doesn't necessarily mean one is in fact better than the other..

[Timur]

You're surely right. Although I don't need to be blind to stay neutral in my judgement when doing a/b, because it's not that I would want any of the two to be better than the other (like some probably do) but just listen and judge on that behalf.

[Angstrom]

Live 6's engine was working fine for me and now we know that Live 7's is technically better than the old one (we'll see how in the paper)

It is now 'one better' , and we can all agree that it certainly doesn't sound worse. Of course it's up to you if you want to spend time straining your ears to hear summing errors / or not, at the -96db marker

it is one better, and it is not worse.
There is a paper coming which will tell you what you need to know, it seems silly to waste valuable lifespan trying to deduct the contents of that paper from phase inversions.

[philipc]

You're surely right. Although I don't need to be blind to stay neutral in my judgement when doing a/b, because it's not that I would want any of the two to be better than the other (like some probably do) but just listen and judge on that behalf.

True, it'd just be interesting to get multiple opinions on such a test given that my listening environment is pretty poor at the moment.

[silicon1138]

Ok, so i've been following this thread very closely. I hold my hand up and claim to know nothing technically about the whole phase conversion, different dither options etc . What i can say is that i've recently finished a new album that' i've been working on for the last 5 or 6 months. I'd just compiled the final cd. All the tracks were written 100% in Live 6, no hardware other than a a few boxes for recording some vocals in Live.

This morning i loaded up one of the songs into Live 7.
The difference was very noticeable to me immediately. There is a clarity and definition pretty much right across the board. the synth bass was clearer, so were the hi hats. The overall mix felt like it had more presence because of the definition. It felt more punchy too, i guess again because of the perceived spacial difference.
All the plugs (in this song) i have used are not from Ableton. Things like TC limiters, CamelSpace, and on the master out i used PSP mix saturator.

Just in normal play back it sounded better. I also rendered using triangle and pow-r3 dither, both renders sounded better than the Live 6 render (i know 6 doesn't have dither).

It's a shame i can't let you hear the what i'm on about. The album is not far from release and still under wraps, but i just wanted to give my two cents, using nothing but my ears. Like i said at the start, no real science, just based on a track i've been hearing over and over for months.
This morning in Live 7 it sounds 'better'.
I've now going to re render the whole album.

[Timur]

Live 6's engine was working fine for me and now we know that Live 7's is technically better than the old one (we'll see how in the paper)

It is now 'one better' , and we can all agree that it certainly doesn't sound worse. Of course it's up to you if you want to spend time straining your ears to hear summing errors / or not, at the -96db marker

it is one better, and it is not worse.
There is a paper coming which will tell you what you need to know, it seems silly to waste valuable lifespan trying to deduct the contents of that paper from phase inversions.

I am with you on that, but I do worry about the time shifts I found in the comparison. Does it mean that L6 was off time or does it mean that L7 is off time? Did it happen due to warping or due to some other effect? I find it especially creepy because it is not a general delay but there are regions which are perfectly aligned while others are off.

As you can see in the graph parts of the L7 curve look stretched compared to the L6 one, which makes sense since the L7 curve needs to "catch up" to get into sync with the L6 one again. I suspect that at another point either the L7 curve looks compressed compared the L6 one or the L6 curve would look stretched, else misalignment could not be happening at all.

Since this is a timing-difference on the horizontal axis and not just a value/resolution difference on the vertical axis I really want to know if the origin lies in the samples or in the engines. If it lies in the engines I want to know: Which engines can I trust more when it comes to timing?

[Heinz Graaf]

The new 64-bit audio engine in Live 7 is really good. I just did a side-by-side comparison of a very cluttered and muddy Live 6 song in Live 7 beta, and the new audio engine is showing off the depth and clarity of 64-bit summing bus. Very nice, very clear, all in all, EXCELLENT.

I stated in another post that it's not something we really needed...I'm eating my words, partially...we didn't *need* it to make great music, but I'm pretty sure after L7 is released, I could never go back and live without it.

I have been saying this for awhile and like EVERY FUKKIN BODY inhere disagreed with me. My mixes sound MUDDY and now I AM right?... because the Abletonners said it by giving you a new updated program??
Youve got to be kidding me. This confirms you folks cant hear for yourself. This also confirms Robert Henke is FULL of it.
Im out...

[Heinz Graaf]

The new 64-bit audio engine in Live 7 is really good. I just did a side-by-side comparison of a very cluttered and muddy Live 6 song in Live 7 beta, and the new audio engine is showing off the depth and clarity of 64-bit summing bus. Very nice, very clear, all in all, EXCELLENT.

I stated in another post that it's not something we really needed...I'm eating my words, partially...we didn't *need* it to make great music, but I'm pretty sure after L7 is released, I could never go back and live without it.

I have been saying this for awhile and like EVERY FUKKIN BODY inhere disagreed with me. My mixes sound MUDDY and now I AM right?... because the Abletonners said it by giving you a new updated program??
Youve got to be kidding me. This confirms you folks cant hear for yourself. This also confirms Robert Henke is FULL of it.
Im out...

[Timur]

This morning i loaded up one of the songs into Live 7.
The difference was very noticeable to me immediately. There is a clarity and definition pretty much right across the board. the synth bass was clearer, so were the hi hats. The overall mix felt like it had more presence because of the definition. It felt more punchy too, i guess again because of the perceived spacial difference.

Problem is, the same "phenomen" applies if you run your mix through a Hi-Fi system or use some kind of EQ (like Winamp's EQ on Rock/Techno settings). A mix through a flat monitoring system seldom sound "grand" but rather "neutral". Putting it through Hi-Fi or EQ will make it sound "better" for most people, but that doesn't mean it's more precise.

Now the questions I do have in mind is: Does your mix sound better - with more punch and presence - because of the more precise calculation of 64-bit and therefor possibly better dynamics? Or does it sound better because the new mixing engine shapes the sound different to the old engine as in getting different frequencies out, and is this wanted behavior? Have you been unhappy with your already "finished and polished" mix that you got out of Live 6? And if so, why did you finish it that way, because the old 32-bit engines did not allow for any better mix or because your ears and mood told you it was good and now after months of work your ears and mood embrace the fresh sound of a different mix?

Don't get me wrong, I'm not ranting against the new engine or any statement here, I'm just asking questions to keep the discussion reasonable and stay clear of what we're talking about.

[Angstrom]

I am with you on that, but I do worry about the time shifts I found in the comparison. Does it mean that L6 was off time or does it mean that L7 is off time? Did it happen due to warping or due to some other effect? I find it especially creepy because it is not a general delay but there are regions which are perfectly aligned while others are off.

well, you could post a succinctly phrased question in the beta forum, although don't say I told you to do it !

As this is a new feature I'm pretty sure the engine guys will be watching to see if there are any issues. There's a chance you will get an authoritative answer there, or a chance to send your files to the team so they can comment.
It's pretty much impossible for us (on the outside) to comment or guess, but it may be that an Engine coder will take time to answer you in there.
If there is an issue, they will want to address it.