The No.1 Website for Pro Audio
 Search This Thread  Search This Forum  Search Reviews  Search Gear Database  Search Gear for sale  Search Gearslutz Go Advanced
DAW Summing Comparison - A new testing angle Dynamics Plugins
Old 22nd February 2008
  #1
Gear Addict
DAW Summing Comparison - A new testing angle

Hi all,

This topic has been argued hard on here, all agreed. I have read hundreds of posts for several months. So ive done my research before posting. I havent posted much. But ive logged in and read weekly for a while now. I would like the chance to comment.

I have tested three DAWs, Nuendo 3, PT LE 7.3 and Ableton 7 this week. I dragged 16 x stereo 24 bit wav files into 16 stereo tracks that total the recording...ie Track 1 = Kick. Track 2 = Clap. Track 3 - Piano. Track 4 = Strings, etc. I hear a difference in the three DAWs combined renders. And if someone on here has a big ftp site where i can upload 24 bit wav files for everyone to listen, id be happy to prove it.

I can hear a difference for two reasons.

1:/ i am rendering 16 stereo tracks together, not a lousy three or four!! There is much more "strain" on a summing buss with 16 tracks attempting to squeeze into 1 stereo wav. The tests ive read on this forum have been for very few tracks sometimes only one track!!(???)...and i feel its not a real-world test. I dont hear a difference with 3 or 4 tracks either...but when ive pumped a whole mix of 16 stereo tracks with kicks, bass, strings, piano, hats etc all filling up the spectrum and driving along into poor Mr Summing Buss...i hear the difference. I think thats the criticial difference.

2:/ I am testing with dance music...House. A big, thumping house kick, strong vocals, stomping piano, crunchy hat, popping clap, etc. Lots of constant energy to compare from second to second, all driving that old summing buss over and over with big transients to reveal its inconsistencies. I dont think i would be able to hear a difference in a meandering ballad or wall of guitar track.

These are my personal observations:

Pro Tools sounds better than Nuendo which sounds better than Ableton. Quite clearly.

For two main reasons:

1:/ Bottom end energy. Pro tools seems to keep more consistent "power" in the lower sub of the track, regardless of the information in the higher frequencies. Specificially, when the pounding dance piano kicks in for a chorus with layered strings and vocals and extra hats and a ton of other heavy stuff, the kick and bass sound more "smeared" and seem to "suck in" a bit in Nuendo, and worse in Abelton. To me it sounds like the bit-crunching is having a harder time when suddenly its thrown all this extra information. Nuendo or Ableton just dont handle all this extra dense sound as well thats being thrown at their summing buss from 16 different sides as Pro Tools. PT sounds more open and comfortable with the extra information, Nuendo sounds like its carrying a sack on its back, PT is saying "bring it on!". Another way to describe the difference would be that it almost sounds like Nuendo and Ableton drop to 16bit when thrown a heap of dense tracks to sum...of course they dont, but we all know the difference in sound between 24 bit and 16 bit. Thats similar to what i hear. I hear the difference most in choruses, when all the sounds are being thrown in...again...when all 16 tracks are beating up the Summing Buss to get thru.

2:/ Attack Transients. The driving kicks and claps in PT are punchier, snappier, crisper, cleaner. Damnit, but its true. I wish it wasnt because i use Nuendo. But sometimes, in Nuendo or Ableton, when a heap of different sounds on different tracks are louder for a split-second and "collide" by sheer chance at the same instant as a clap, the clap attack seems to "smear" and gets a bit lost. They just occasionally dont "cut thru" as sharply. But in PT, they just seem to soldier on and rise above whatever else is happening...they arent troubled by any other information on the 15 other tracks at all. ie. Better definition.

What i dont hear: I dont hear a difference in the stereo width. I dont hear a difference in the top end, it doesnt sound any brighter, warmer duller, anything. I dont hear a difference in the overall level of the renders. Other than these two points listed above, theres not a lot of difference otherwise.

Lets be civilised and respectful with your response post when calling me a "liar" ok. I wish there wasnt a difference. There shouldnt be a difference. But there is. To my 35 yr old engineering ears anyways...

Cheers,
Darren.
Old 22nd February 2008
  #2
Did you bounce the resulting mixes down to stereo so that they can be compared on a bit by bit basis? If not, I suggest doing that and then you (or I) can compare them to each other.

I've done a little bit of comparison testing before.

BTW, 16 stereo tracks doesn't seem excessive to me, in this day of 100+ track mixes. The summing shouldn't fold as a result of 32 tracks.
Old 22nd February 2008
  #3
Do they null?
Old 22nd February 2008
  #4
Lives for gear
 

Were all the pan law settings the same on each Daw?
were all the levels excately mateched?
Old 22nd February 2008
  #5
Gear Addict
Hi, thank you for your posts.

Yes they were certainly mixed down to stereo files, 24 bit wav. As i mentioned the wavs are what i can upload if desired somewhere.

Yes the panning laws are the same. Yes the levels are identical. They are stereo rendered files, with the panning built-into the wavs. So there was no panning in the DAWs done either. The multi-wavs are at a reasonably low level so that the resulted "summed" wav file is far from clipping, indeed it averages around -6.

re:waves nulling...no, i have not done that test. i have used my ears. If you want to do the null test thing and it comes up null so u can tell me "see it says theres no difference" then im very happy for you. But i have just had an associate play the files randomly to me with my back to the monitors and i just picked the PT mix correct 10 times in a row...and it wasnt Rocket Surgery.

I must say the difference in the gap between the quality of Nuendo and Ableton is not as easy to distuinguish blindfolded however. if PT is 10, then Nuendo is 8 and Ableton is around 7.2...if u get my drift...

Cheers,
Darren.
Old 22nd February 2008
  #6
Gear Addict
 
richardswag's Avatar
 

Quote:
Originally Posted by Blackout- View Post
re:waves nulling...no, i have not done that test.
Why not? It's not exactly difficult now that you've gone to all this trouble already.

Are you worried you might de-bunk yourself?
Old 22nd February 2008
  #7
Lives for gear
Digi hardware for all three apps?
Old 22nd February 2008
  #8
Gear Head
This really confuses me, being a programmer its hard to see how 3 programs could add up 16 numbers and get a different result. You do get rounding errors and things in floating point calculations but they ought to be the same in each program.

Certainly there shouldn't be enough of a difference to affect the dynamics or frequency content.

I suppose its possible they use different dithering algorithms (ableton 7 at least works at 64bit internally) or something. Did you leave all the levels at 0db? Is the master out at -6db without turning it or any of the tracks down?

Not saying you're wrong I just want to know why its different!
Old 22nd February 2008
  #9
Lives for gear
 

Pro Tools is fixed point. Nuendo is floating point.
Old 22nd February 2008
  #10
Gear Head
oh well that makes sense then i suppose
Old 23rd February 2008
  #11
Lives for gear
 
sam c's Avatar
 

Science is exact if this is done right. So, somewhere there is a difference. I will try and find the summing by G Anderton that has 40 tracks of classical since you seem to feel more tracks will change science.

This is getting to be goofy how some of us hang on to it sounds different (and others the need to say it isn't so....like me!) no matter what. Actually I believe you can convince yourself of anything you want and I think you want to hear a difference. Fine..........

Pick the one you think sounds best and record. Feel free to tell all your DAW sounds better than all the others in the universe. Hell, it might inspire a few good recordings..........
Old 23rd February 2008
  #12
Lives for gear
 
hmiller's Avatar
Ableton sounds different than the others because of its built in time stretching algorithm. I think that core feature of the application contributes to it's sort of smaller, smeared, almost pinched sounding master bus. But that's just a guess!
Old 23rd February 2008
  #13
Lives for gear
 
stag's Avatar
 

Which interfaces did you use to perform the test? BTW: which monitors?
Old 23rd February 2008
  #14
Gear Addict
Hi all,

Ableton might have contributed a smearing , correct, if its timestretching algorithms were on obviously. But of course they were not being used. It was straight wav file playback....i confirmed they were turned off.

Re:Levels, the output buss fader in every instance was not touched. It was left at 0db. What im describing at -6 is where the peaks in the resulting mixdown wave file tend to land around. The point im making is that there was a fair amount of headroom remaining, so it wasnt a case of the summing buss "running out of numbers".

Re:Hardware, i only used my Digi 002. And i am playing back all 3 wav files in Sound Forge, side by side, thru my Digi 002. Monitors are Dynaudio BM6A's...And a decent set of sony headphones.

Nuendo and Ableton seem to sound very similar. Ableton has slightly more "graininess" about it to me, which is the only way i can pick them. The definition and transient problem thing seems to be very similar, but Abletons transients are slightly further effected by the graininess. So thats how i tell them apart. And i guess incorrectly in a blindfold test quite a bit between them. But PT stands out a mile. Perhaps its the fixed-point vs floating point thats doing it....would make sense as both Nuendo and Ableton are floating....hmmm

Pulling out even a 50 piece orchestral session to test is not going to cut it im afraid. You wont hear a difference. Thats the whole point im making. Its in the big transients like thumping kicks and claps. If you tell me youre going to use a 50 track session of an african drumming band, then id get excited and say go ahead..and be prepared to hear the difference...
Old 23rd February 2008
  #15
Gear Head
 
ipressrecord's Avatar
 

You really have to null them and listen to the results. If you have silence, then there is no difference.

Correct?

Jeff
Old 23rd February 2008
  #16
Lives for gear
 

panda21 .
your a programmer....
thus i'll share a little theory of mine..cos youll understand
the array manipulation technique.
all daws do essentially the same thing from what i see.
use look ahead disk techniques to bring trak time portions into memory arrays.
before they are needed.
this explains why when some people say they hit play.there is a small lag.
this is cos the first look aheads are being created.
(some daws i suspect are doing it at idle so everything is prefilled.)

thus one has a memory array allocated per trak.
memarray1 ..trak 1.
memarray2...trak 2.
memarray3...trak 3.
etc etc...
memarraynn..traknn.
prolly the daws actually have several memarrays for each trak.
so as one is being played next ones being filled etc.
(look ahead again.)

due to the severeity of folks claims they "can" hear a diff tween different daws....
i suspect the clue lies in those memory arrays rather than at the summing stage.
prior to summing useing look ahead a "canny" programmer could look at the trak ,
and do some dsp sample value manipulation on the samples prior to summing.
this is what is foxing folks out i believe , and why they are scratching there heads
on the null aspect. see what i mean....????? signal conditioning prior to summing to
the stereo master memarray that gets pumped out to the sound device.

as so many people have commented very severely they can hear a diff tween different daws,this is the only logical area i can see where a canny proggie could "improve" the sound.
ie.signal analysis/low level artifact n noise clean up prior to summing.
we will never know the whole story of course cos the issue will never be solved unless a forensic programmer is allowed access to the source code of each daw.

as a matter of interest..how i came to this concept was... one particular daw i tested a few years back i purposely fed it a really bad input audio.....and...on playback i definitely noticed a sound improvement. which led me to conjecture that somehow the memarray concept i outline above was being used to do dsp signal conditioning.

any comments ??
(pm me if u wish n we can get into some tech aspects n
debate further...if u find it interesting.)
Old 23rd February 2008
  #17
Lives for gear
 
hmiller's Avatar
Quote:
Originally Posted by Blackout- View Post
Hi all,
Ableton might have contributed a smearing , correct, if its timestretching algorithms were on obviously. But of course they were not being used. It was straight wav file playback....i confirmed they were turned off.
Right and right. I was just saying that since they are built into the core of the app, it affects it's sound even when you're not stretching. Just my best guess.
Old 23rd February 2008
  #18
Lives for gear
 
dannygold's Avatar
 

I just did a test similar to the OP... I'm not saying that's the best way to test DAW summing, but since he didn't seem interested in a null test I thought I'd try one. I do think they're valuable. If you get complete cancellation they ARE the same... anyone who says otherwise doesn't understand digital audio theory very well. Anyway, I don't have (all) the same DAWs as him, but it's still interesting to compare DAWs.

I took 16 stereo and 2 mono files and put them in DP 5.13 on tracks with no FX. I left everything panned up the middle. No tracks were clipping. I made a master fader and turned it down 12 dB to prevent clipping there. Yeah yeah, I know you all think you should turn down the individual tracks not the master fader, but whatever, it's my test! I've never thought turning down the master fader was a bad thing to do, in DP at least.

I did the same thing in PT LE 7.3. Same SDII files, setup the same way, same master fader down 12 dB.

Created 24 bit bounces from both.

When you null them out in a new project (via phase inversion) nothing is AUDIBLE, but there's something happening at about -98 dB down below full scale on the meters. Now, that's very quiet. Arguably meaningless. However, I cranked it up with trim plugs to see what it sounded like. It was mostly noise, dither I assume (although I didn't add any... maybe there's some auto dithering going on in the 32 bit float to 24 fixed conversion). I could also hear a faint ghost of one of the louder tracks. But again, keep in mind this was happening at -98 dB. Hardly what I'd consider a major difference... but to be fair, they were different! When I nulled a dup of one of the bounces it was totally canceled out, so my nulling system was accurate.
Old 23rd February 2008
  #19
Lives for gear
 
feck's Avatar
If there is any difference, what does it amount to? .01%? Who really cares....it would be so miniscule as to not matter. Jeez, talk about hyper detailed!
Old 23rd February 2008
  #20
Gear Nut
 

Quote:
Originally Posted by cepia View Post
Right and right. I was just saying that since they are built into the core of the app, it affects it's sound even when you're not stretching. Just my best guess.
Why guess when you can get the facts, straight from the source?

Ableton - Ableton Live 7

Check out the Audio Engine Fact Sheet found there.

The long and the short of it is that yes, Ableton Live is capable of transparent reproduction of audio files provided that none of the pitchshifting/timestretching mechanisms are engaged ... however, this is beside the point as by Ableton's own admission, Live is not capable of rendering it's output to 24 bit or dithered 16 bit transparently - they recommend rendering to 32 bit and doing bit-depth reduction in an offline application. Given these facts, and the fact that the test was performed with audio files rendered to 24 bit, I would not be the least bit surprised to find that the bounce from Live will not Null with the other two - whether or not the difference is particularly audible in an A/B comparison is another question, but I'd expect to hear more of a difference (if any) between Live's rendered output and the other files compared to the difference (or more likely the lack of difference assuming the tests were done correctly) between the outputs from ProTools and Nuendo. If the test was done correctly and the rendered outputs from PT and Nuendo do not in fact null, then it implies that one of the two programs is broken, and should join Ableton Live in the category of "not a professional DAW/not fit for use as a primary recording medium in a professional studio environment."

In my view the sole prerequisite for something to be called a professional DAW should be that it has the capacity to sum/render it's contents to a 24 bit stereo WAV file transparently, given that this is the closest thing that exists to a standard for multi-platform interoperability and archiving of digital audio content.

G.
Old 23rd February 2008
  #21
Lives for gear
 
dannygold's Avatar
 

Quote:
Originally Posted by dannygold View Post
Yeah yeah, I know you all think you should turn down the individual tracks not the master fader, but whatever, it's my test! I've never thought turning down the master fader was a bad thing to do, in DP at least.
OK, did a null test with master fader in DP turned down 12 dB and no master fade vs. tracks turned down 12 dB. THAT nulled out completely. So there.
Old 23rd February 2008
  #22
For anyone who is interested, a similar and very rigorous testing of 19 DAW platforms was conducted 5 years ago by members of my forum (almost to the day, 2/24/03). You can read dozens of threads with hundreds of posts in a forum dedicated solely to this topic.

It is here:

Awesome DAW-SUM Comparison - 3dB

You can even listen to the resulting soundfiles. They are available online here:

Awesome DAWSUM Samples - 3dB
Old 23rd February 2008
  #23
Quote:
Originally Posted by dannygold View Post
OK, did a null test with master fader in DP turned down 12 dB and no master fader, tracks turned down 12 dB. THAT nulled out completely. So there.
By nulled completely, I assume you mean the summed result with polarity inverted files was -infinity?
Old 23rd February 2008
  #24
Lives for gear
 
dannygold's Avatar
 

Quote:
Originally Posted by Lynn Fuston View Post
By nulled completely, I assume you mean the summed result with polarity inverted files was -infinity?
Yeah... they nulled to infinity. My conclusion is that turning down the master fader is just as effective as turning down individual tracks... unless of course the individual tracks are clipping which is another issue.
Old 23rd February 2008
  #25
Gear Head
Quote:
Originally Posted by manning1 View Post
panda21 .
your a programmer....
thus i'll share a little theory of mine..cos youll understand
the array manipulation technique.
all daws do essentially the same thing from what i see.
use look ahead disk techniques to bring trak time portions into memory arrays.
before they are needed.
this explains why when some people say they hit play.there is a small lag.
this is cos the first look aheads are being created.
(some daws i suspect are doing it at idle so everything is prefilled.)

thus one has a memory array allocated per trak.
memarray1 ..trak 1.
memarray2...trak 2.
memarray3...trak 3.
etc etc...
memarraynn..traknn.
prolly the daws actually have several memarrays for each trak.
so as one is being played next ones being filled etc.
(look ahead again.)

due to the severeity of folks claims they "can" hear a diff tween different daws....
i suspect the clue lies in those memory arrays rather than at the summing stage.
prior to summing useing look ahead a "canny" programmer could look at the trak ,
and do some dsp sample value manipulation on the samples prior to summing.
this is what is foxing folks out i believe , and why they are scratching there heads
on the null aspect. see what i mean....????? signal conditioning prior to summing to
the stereo master memarray that gets pumped out to the sound device.

as so many people have commented very severely they can hear a diff tween different daws,this is the only logical area i can see where a canny proggie could "improve" the sound.
ie.signal analysis/low level artifact n noise clean up prior to summing.
we will never know the whole story of course cos the issue will never be solved unless a forensic programmer is allowed access to the source code of each daw.

as a matter of interest..how i came to this concept was... one particular daw i tested a few years back i purposely fed it a really bad input audio.....and...on playback i definitely noticed a sound improvement. which led me to conjecture that somehow the memarray concept i outline above was being used to do dsp signal conditioning.

any comments ??
(pm me if u wish n we can get into some tech aspects n
debate further...if u find it interesting.)
i suppose that could be possible, to be honest I think a lot of audio software is far more simple than people realise, especially having come from an analog/external gear background. i doubt any of the daws do anything like that because it would take up quite a lot of processor power, and its hard to see what they could do that would be useful without colouring the sound to an unacceptable degree without giving you any control over it.

if theres a difference between things like nuendo/cubase/logic/live i would imagine that comes down to processes like dithering and so on rather than the summing

i'd also expect 64bit floating point to sound better than 24bit fixed point, so long as you don't go over 0db, as there is more dynamic range and you can represent quiet sounds better. going over 0db i think you could argue that floats start to lose precision if i remember how they work correctly.

but if protools works internally at 24 bit fixed that would remove any upsampling/dithering from things and perhaps thats why it sounds better?

interestingly though in another forum i read someone posted that they were having problems with logic 7, loading in an audio clip of some vinyl noise. when they went to play it back it had been changed quite drastically without them doing anything to it. but if that is logic doing something behind the scenes though i would imagine its on loading the audio rather than in real-time.
Old 23rd February 2008
  #26
Lives for gear
 
DeadPoet's Avatar
To the OP: you say you don't have to take the panning laws into account because the mixed tracks are stereo tracks.
I always thought that the moment you use the stereo field (= whenever not doing mono) you have to check the panning laws.

Someone correct me if I'm wrong.

I've just acquired a lightbridge from a fellow gearslut and will buy M-powered soon so I'll be testing this as well.


Herwig
Old 23rd February 2008
  #27
Lives for gear
 

panda.
thnks for the comments n seeing what i was getting at..
i needed a check on my thinking.
that logic example is interesting.

re real time resource consumption.
well....i'm not sure mate.
useing low level v efficient asm dsp techniques n look aheads
in a multi threaded app. it might be possible.
one could build in a detector for low level artifacts etc etc running under the audio signal prolly.
i really wonder if the null tests arent somehow foxing people out.
Old 23rd February 2008
  #28
Quote:
Originally Posted by DeadPoet View Post
To the OP: you say you don't have to take the panning laws into account because the mixed tracks are stereo tracks.
I always thought that the moment you use the stereo field (= whenever not doing mono) you have to check the panning laws.

Someone correct me if I'm wrong.
The pan law only comes into play when something is panned somewhere OTHER THAN hard left or hard right. If the signals are panned 100%L or 100%R, then the pan law has no effect.
Old 23rd February 2008
  #29
Gear Addict
 

Hi,

Slightly on a different angle, but are people saying that: If i use cubase SX to record a track in, then mix down my 100+ tracks to 30 sub group stems, then load them up in PT LE7.4 and use this program to mixdown in, it will sound better than if i just mixdown straight from Cubase? Is the difference REALLY worth the effort? and is it just the pro tools software that makes it sound great (!) (as the mbox hardware is just for my monitors).

OR, has the damage already been done by using cubase in the first place?

t
Old 23rd February 2008
  #30
Gear Addict
tritace...thats exactly what im suggesting you do. and you will hear a difference. Keep kicks, bass, claps, big snares all to themselves on their own tracks. Basicially group pads and strings and things that are not "transient"...and leave things that are percussive and punchy to their own track. If Cubase is only outputting one sound such as a kick for a render, it will not damage the kick sound. My belief is that things only go pear-shaped when lots of sounds are thrown at the miserable summing-buss algorithm all at once...thats when i can hear it "choke".

cepia...yes that was my thoughts too...that Ableton has the timestretching algorithm so deeply embedded into its entire being that its still running signals thru some of its algorithm still even tho the algorithm might be set to Zero for its settings when turned "off". Might explain it. And it sounds like even Ableton admit it. hehe

Ok ive decided to post the wav files on speedyshare to put an end to this so you can all hear the kick and clap transient difference for yourself. You need to click on the link, which will take you to the speedyshare site, then right-click on the blue file name and then select "save as" and save to your hd first.

They are excerpts from 4min29secs for about 30 seconds each, 24 bit wav files. These are the raw renders...what Nuendo and PT spat out ....no normalising, dithering, anything.

The track is from an official remix i have done for a record label and not released yet so pls put the wavs in a non-shared folder and delete when finished....just in case the label stumbles over them somehow ... i dont think they will have a baby over 30 seconds worth but id appreciate your co-operation in this regard. Thanks.

Here is the Nuendo 3.2 Render:

Nuendo 16 trk Render.wav

And here is the Ptools 7.3 Render.:

PTools 16 Trk Render.wav


As i stated in the original post...listen for the more open bass with more overall "strength", punchier kick and sharper more detailed clap in the Pro-tools render. First time pass-thru you might not hear it...then after a few mins you will most definately "feel" the difference and be able to recognise it.

All you Null-enthusiasts...here are the wavs....knock yourself out.

Cheers,
darren.
Post Reply

Welcome to the Gearslutz Pro Audio Community!

Registration benefits include:
  • The ability to reply to and create new discussions
  • Access to members-only giveaways & competitions
  • Interact with VIP industry experts in our guest Q&As
  • Access to members-only sub forum discussions
  • Access to members-only Chat Room
  • Get INSTANT ACCESS to the world's best private pro audio Classifieds for only USD $20/year
  • Promote your eBay auctions and Reverb.com listings for free
  • Remove this message!
You need an account to post a reply. Create a username and password below and an account will be created and your post entered.


 
 
Slide to join now Processing…
Thread Tools
Search this Thread
Search this Thread:

Advanced Search
Similar Threads
Thread
Thread Starter / Forum
Replies
TheArchitect / Music Computers
0
Nino / Music Computers
0
Earwitness / So much gear, so little time
11
Your Add Here! / So much gear, so little time
1
Dave Martin / Q+A with Dave Martin
33

Forum Jump
Forum Jump