SONAR: Export to .wav vs. Outboard Mastering Gear

  • Thread starter Thread starter mark4man
  • Start date Start date
mark4man

mark4man

MoonMix Studios
Gang,

I know we've gone around on this one before, but I'm still not sure (in terms of quality) just how to treat my pre-mastering process concerning SONAR XL 2.2. SONAR's audio engine (at least I think it's the audio engine) exports (bounces) all selected tracks to a stereo .wav file (yes...we all know this, Mark.) But the .wav file never seems to sound as good as the sound of all tracks playing back in the app.

The first major difference (which I'm sure I can rectify in WaveLab), is that the .wav files don't sound as bright . . . almost as if the highs are a bit muted. I can live with this one, since it is very subtle at most (although it does bother me that it is not an exact duplication, which I am assuming it should be.)

The second difference drives me nuts. For those tracks that have passages that are "distortion questionable" (this is my term for those tracks that sound fine when soloed, but that sound as though they have a subtle digital distortion when played with the entire mix), those passages are always somewhat noticeably distorted in the bounced .wav file.

Even when I pull the output buss down to where I have virtually no clipping, you can still hear it (& there should be no such thing as cumulative signal distortion if we're below 0dBFS, correct?)

So my observation is: Maybe the file writing function (audio engine?) of the major PC multitracking software apps is not yet up to snuff, comparitive to mastering gear (or is it me?)

I know audio engine quality has been on the increase through various Cakewalk incarnations, but I was wondering:

How much better off would I be with say; an Alesis Masterlink, or similar tool. How much better would the quality be, given that the transfer process would be essentially an analog output capture, directly to hard disk in the Masterlink, during multitrack playback? (meaning that the signal would undergo double conversion.)

The pro's all swear by this machine . . . but it seems to me that SONAR's export function should be all that is technically necessary to create high-quality pre-masters.

If someone could help with what I might be doing wrong in SONAR . . . or offer some insight as to why essentially a digital data rewrite (which I am assuming is what happens when multiple digital audio tracks are written to a single file from DAW software) doesn't yield precise results . . . or if I would just be better off utilizing outboard mastering gear . . . it would be appreciated.

Thanks,

mark4man
 
mark4man said:
Gang,


Even when I pull the output buss down to where I have virtually no clipping, you can still hear it (& there should be no such thing as cumulative signal distortion if we're below 0dBFS, correct?)
Click to expand...
Not sure what you mean by this.

If you are saying that because EACH track is below 0db, there should be no clipping of the mix, that is not correct. When you put two tracks together, the sum will be louder than each of the individual tracks. The more tracks, the louder the total mix. So, yes, there is indeed "cumulative clippping." That occurs when the sum of the tracks reaches or exceeds 0db - regardless of what the indivual tracks are doing. As such you need to insure your VMain meter remains below 0db.
 
dachay,

What I should have said was: Even when I pull the V Mains down to where I have virtually no clipping, you can still hear it (& there should be no such thing as cumulative signal distortion if we're below 0dBFS, correct?)

-----------------------------------
It's still there, even in that scenario.
-----------------------------------

mark4man

& the fact still remains that the bounced .wavs do not sound as good as multitrack playback.
 
the pro's send it digitally to the masterlink so it undergoes no conversion in the unit they use there $10,000 converters for that
 
I have had some dialogue with Sonusman. He, like you, feels there is a difference in Sonar between playback and mixdown. I have not seen/heard it myself.

OTOH, I feel a similar thing happens in Wavelab when you render the file. Rendering in Wavelab is similar to the mixdown in Sonar, in that it is the point where all the realtime effects actually are applied to the file.

I would assume that the algorithms used for playback and mixdown would be identical; but who knows.

I have no experience with the Masterlink, so can't offer any useful comments. All I can say is someone with Sonusman's experience seems to agree with you about the Sonar mixdown function.

BTW, I'm not sure there is such a thing as "subtle" digital distortion. However, why not try and put a brickwall limiter on the VMain when you do your mixdown. That would insure everything is below 0db. Even though I master in Wavelab, I usually put an L2 on the VMain. I turn off the dithering and only set an out ceiling.
 
Mark,

The results you got from your null test (from another board) got me wondering... so I tried a test of my own.

I took a 6 track project with various plugins (gates, compressors, reverb, a separate compressed bus for the drums, etc.) and exported it through the master bus to a stereo track at 24/96khz. I then imported the stereo track back into a new audio track in the same project and inverted the phase. Silence.

I tried the same thing with a mastered version of the same project. LA2A, Pultec eq, L2 chain over the stereo mix-->Export-->Import to new track-->Invert phase = Silence.

I tried this at 16/44.1khz to simulate your circumstances, and got the same result.

SteveD
www.DawPro.com
 
Hey Mark4man, you have some interesting in your observations.
There sure is plenty of info/discussion/opinion out there that says bouncing is inferior in most any of the progs. Slips right into the 'digital summing', and of course, the big Kahuna, whether we should be mixing in the box at all! :D First off I admit I do not get why, if a system can play all this stuff live, it wouldn't be actually easier to spit it out nice and slow on a bounce.

On 'soloing the marginal track' (or any track), right off I guess I would tend to not question it sounding different in and out of context. But I can think of tracks I'd been 'struggling with', that never seem to set well to begin with, let alone days where nothing sounds the same way twice.:rolleyes:
Do you feel other tracks are doing ok on the bounce? I'm wondering what if you took the trouble ones out? (for a test)
Just to check some other things, see if we're on the same page and all...
You're bouncing to a new track, then A/B'ing the mix track (soloed) against the 'real-time' playback.
Both go out the same playback chain. (D/A, etc.)
If there's =anything= going on in the primary V-main (i.e.; it's not set 'zero' or there's inserts in there) your 'mix' track gets it's own clean V-main (at zero) and on the same output path? (Sorry, may as well them all out of the way. :D
On the 'almost not clipping the main' thing, I have to pass on this one.
On one hand I see peaks way past zero that don't seem to hurt anything, and the mix buss supposedly has lots of headroom above 'zero'. (But NOT the digi-out right! ...check!) On the other, I tend to be a chicken shit, try to get my mixes to come in below zero back at the track fader level, then come up if needed at the main. (Here's another spot to play minimalist- if you have main inserts, make one make-up gain change, not at each plug, blah blah blah... :)
There's also this thing I'll mention again in rev 2xl where GAIN envelopes like to play little games with the mix on bounces, but play live just fine. (Makes me wonder if there's other stuff going 'oops! in there from time to time)
Might as well throw in that little bit that progs in general and maybe plugs even more so, might run a little happier at nice conservative levels. (But would that even be a 'bounce' issue? Don't know.)
Basically, I'm looking forward to getting rev 3. The true subs (Yes!), some are saying it sounds even better (hee hee!), (Do I get to use gain automation without second guessing every damn time!?)
But when I shift play here between the mix track and live, I'm not getting anything near the gross errors and stuff like that. They're coming back pretty damn close. Hmm...

SteveD. Are you exporting the 'mix' then bringing it back in to compare (nul) it to a 'bounced-to-track' version? I'd assume (there's that 'ass, u n' me' word) that an 'export would be exactly the same as a 'bounce. Is that supposed to relate to what happens on a 'live play' vs bounce?

Can someone point me to the other thread Mark4 has going?
Thanks all.
Wayne

ps. Sorry for going on so long. Can't get enough...:rolleyes:
 
mixsit said:
SteveD. Are you exporting the 'mix' then bringing it back in to compare (nul) it to a 'bounced-to-track' version? I'd assume (there's that 'ass, u n' me' word) that an 'export would be exactly the same as a 'bounce. Is that supposed to relate to what happens on a 'live play' vs bounce?

Can someone point me to the other thread Mark4 has going?
Thanks all.
Wayne
Click to expand...
I've tried it both ways. The exported/re-imported track nulls against the multi-track and against the bounced stereo mix.

I am not of the opinion that bounced or mixed tracks sound different. I work at 24/96khz, but I don't think that makes a difference for this issue.

The other board Mark posted on is the Mastering WebBoard... hosted by Glenn Meadows of Masterfonics fame. Many of you have probably heard of it. The guys hanging out on this board are award winning mastering engineers... Glenn Meadows, Bob Katz, Alan Silverman, Dave Collins, Ronny Morris, Ed Littman. I don't jump in there often, 'cause I'm not in their league, but I have learned tons monitoring this board.

Anyway, they're kind of ignoring Mark's post... as I thought they would, since we're not exactly discussing Sadie, Sonic Solutions, or Sequia here. When we talk about AD/DA converters here, rarely are we referring to Apogee, Mytek, or CraneSong.

I did respond to Mark's post there because most of the projects they get in their mastering houses come from high-end Pro Tools studios, and I posted my null test results as a reply to prevent them from forming negative opinions about Sonar.

Although they're VERY friendly and helpful, they are pretty careful to keep the discussion on topic and related to Mastering in the true sense of the word.

Here's the link, but like I said... I try not to post unless I've got something really intelligent to say:
~mastering


SteveD
www.DawPro.com
 
dachay, Teacher, DawPro, mixsit . . .

I leaned so damn much today, my head is bustin' ! ! !

I actually had a few posts going on this one, mixsit...I like to bounce back & forth between forums such as this (where the members really know their stuff regarding SONAR); & forums such as the Mastering WebBoard (where you run into recording engineers with 30 years experience & the like.)

But the piece of knowledge that just blew me away was this gem, in response to a comment I made suggesting that clipping was clipping, regardless of whether or not it happened inside or outside of the box:
Not within a DAW application which uses floating point calcs - as long as you remain in float, clipping may be shown by the metering, but will not actually be occurring. Of course, once you leave the float environment, such as passing through a non-host filter or to an external device (audio card), all bets are off... (Courtesy of: Sean Diggins)
Click to expand...
That just about summed it all up, for me (pardon the pun.) For years, I've wondered why DAW multitrack playback sounds different than a bounced .wav (& mixsit, you were also expressing similar observations in your post, as was dachay.)

And I'm not sure the theory is 100% on the mark here, as related to a straight mixdown environment where no plug-ins are present) . . . as dachay said, playback & mixdown should theoretically utilize the same algorithms . . .

. . . but it certainly has caused me to take a finer look at what is happening in the pre-mastering chain; & for that I'm grateful.

As a matter of fact, Sean also made mention that, within a DAW which operates via a 32-Bit Floating Point calculation format (e.g., SONAR), certain effects filters & plug-ins will not (instead operating via fixed point algorithms.) This will cause the plug-in calcs to overload during the data rewrite (bounce.)

This all makes perfect sense to me, now; & I think I have to rethink my approach as a result. As I told Steve (DawPro) earlier, in the past I sought to make my individual tracks hot; & then compensated by bringing the V Mains down.

I'm now going to grab one of my problem-child Projects; & go back in & pull down all the levels on the individual tracks (so as to eliminate all overs); & then bring the V Mains back up to a normal level (& then bounce & observe the results.)

----------------------------------------------------------------------------------------------------------------------------------------
Steve,

Yeah...the null test results have me worried (& thanks for trying to simulate the problem.) I'm going to perform another test after I rework the Project. (& they don't really ignore me at the WebBoard...I've had compelling chats w/ most of the greats over there touching on more than a few project studio sized issues.)
----------------------------------------------------------------------------------------------------------------------------------------

And...Thanks again to everybody here, as well (I'll take this as my leauge, anyday.)

mark4man
 
Last edited:
Cool. Great stuff.
Every time I get into this I get closer to actually seeing what's going on in there.:D

"Not within a DAW application which uses floating point calcs - as long as you remain in float, clipping may be shown by the metering, but will not actually be occurring. Of course, once you leave the float environment, such as passing through a non-host filter or to an external device (audio card), all bets are off... (Courtesy of: Sean Diggins)"

"As a matter of fact, Sean also made mention that, within a DAW which operates via a 32-Bit Floating Point calculation format (e.g., SONAR), certain effects filters & plug-ins will not (instead operating via fixed point algorithms.) This will cause the plug-in calcs to overload during the data rewrite (bounce.)"

Once again, looks like there's every reason to keep levels in line. There might be all this headroom available, but there's these 'gotcha's in there.
Things I get from this- Yes, there is clean performance above zero, but if the meter don't show it, how does a dumb working stiff like me know when he's exceeded it?
You definitely need to deal with it before it hits the very fixed 24 bit output device. And you could get into trouble even way before that.
 
Last edited:
I have the same beef with Sonar....I've ended up fixing the wav.'s in Wavelab with varying degrees of success...
 
Summing of any perticular program is the LEAST of your problems....worry about the source the sound is coming from then your chain the source takes to get to tape then maybe summing...
 
jerome said:
I have the same beef with Sonar....I've ended up fixing the wav.'s in Wavelab with varying degrees of success...
Click to expand...

If your ok with the mix but still hearing 'export or 'bounce issues, the OTHER answer is actually simpler; track it live (digi-to-digi or course)into a second recorder. Bam. End of story!:D
Wayne
 
Yo mixsit,

Man, this pre-mastering process thing is really starting to develop into a quest for me (& probably a pain in the ass for everyone else)...but, I made some interesting discoveries today which I wanted to share (in hopes they will help someone else.)

The first was based on a discovery of my own screwup. When I originally performed the null test, my V Mains were at -4dB (to compensate for slammed tracks & cumulative gain.) I had tried the test both ways, which was to first export to .wav; & then import that .wav to another track; & secondly, I bounced internally to another track.

Of course, since the output buss was not set precisely at 0dB, the null test gave me a low volume result.

I went back in & performed the same test, only this time with the V Mains set perfectly to 0.0dB; & the results yielded a true null (for both exported .wav & internal bounce.)

(Granted, I should have picked this up earlier...sometimes it’s the simple stuff.)

But I now know that SONAR is competent in that function (my apologies to Ron Kuper; & the boys at Twelve Tone Systems.) The other observation I made in the process was that; going back to my original problem (exported .wav sounding different than multitrack playback), this time, the two sounded identical!

That set me off down another road. Thinking that the null test may have proven that the issue wasn't one of headroom, but maybe of output balance, I wanted to be sure.

After I brought all of my individual tracks down equally to eliminate overs, I pushed the V Mains back up to 0dbFS. Yeah, I know that slammed the output buss, but I wanted to see how much headroom I could get out of exporting at a higher resolution.

With the clipped output, I exported first to 16-Bit .wav; & then to 24-Bit .wav. I opened them both in WaveLab; & the 24-Bit file sounded noticeably nicer. The file had more presence; & the spikes were cleaner, almost like I'd gotten away with too high of an output level.

I converted the bit depth back to 16; & dithered with UV22HR at the same time...& that 16-Bit file sounds better.

I've heard a lot of different opinions on whether or not up & down conversion sounds better (& the SONAR User Guide even states that exporting 16 to 24 will increase the precision for effects), but I think I'm now a convert to this method.

Thanks for listening.

(man, I hate that phrase.)

OK, try this one:

Take it for what it's worth,

mark4man
 
"Man, this pre-mastering process thing is really starting to develop into a quest for me (& probably a pain in the ass for everyone else)...but, I made some interesting discoveries today which I wanted to share (in hopes they will help someone else.)"

"...going back to my original problem (exported .wav sounding different than multitrack playback), this time, the two sounded identical!"
Man, are you kidding? This is what it's all about as far as I'm concerned! I learned, you learned, the word gets spread, the music gets better. WTF!:D

"...the null test may have proven that the issue wasn't one of headroom, but maybe of output balance..."
"...I know that slammed the output buss, but I wanted to see how much headroom I could get out of exporting at a higher resolution."

Want to get into this?
(Hey, I don't want to come on as some know it all here. But if we get too far out-a-whack, hopefully someone will come in and straighten us out anyway right?...:p

Not 100% sure what you mean about output balance, but 'slammed output' + 'higher resolution'? As in 'the further up the scale = the higher the resolution'? The story goes that anything that fits within the available dynamic range is good to go. Higher or lower (within reason, say, well above the bottom which is easy enough) is not to worry.

"With the clipped output, I exported first to 16-Bit .wav; & then to 24-Bit .wav. I opened them both in WaveLab; & the 24-Bit file sounded noticeably nicer. The file had more presence; & the spikes were cleaner, almost like I'd gotten away with too high of an output level."

You seemed to have found variable in the 16-bit ver Vs the 24. No clue here what that's about. But on some quick spikes, it might not be much worse than some brick-wall limiting, or maybe so quick they're just not audible.

"..I converted the bit depth back to 16; & dithered with UV22HR at the same time...& that 16-Bit file sounds better."
...Than the 24 bit?
Again though, IF you accept that pushing the levels up gets you nothing but louder...:rolleyes:

"...I've heard a lot of different opinions on whether or not up & down conversion sounds better (& the SONAR User Guide even states that exporting 16 to 24 will increase the precision for effects), but I think I'm now a convert to this method."
Maybe all they are saying is if you're stuck with a 16 file and have to do more processing, go ahead and do it in 24 bit (for the sake of the new processing). Are you saying an extra (or unnecessary) trip up-and-down might do something preferable?
Maybe it just gets some more juice?:D
Wayne
 
mixsit said:
Are you saying an extra (or unnecessary) trip up-and-down might do something preferable?
Click to expand...
I hope that's not what he's saying. Stay at the same bit depth and sample rate until the VERY last stage when you downsample and dither to 16/44.1 for CD. Do this once and once only.

The exception to this MIGHT be a mastering house using a Lavry or equivalent SRC into their HIGH-END gear for mastering. Bob Katz does this, but most mastering houses do not... I checked. They will stay at the same bit depth and sample rate as the mixdown recording.

If for some reason you get a project at 44.1khz and are asked to add some tracks with heavy effects or processing, then upsample the whole project to a new project at 96khz or what ever your system can handle and record your tracks and add your effects and processing.

Then, apply effects to the audio, and export the tracks and convert to the bit depth and sample rate of the original project externally.

Then import the new tracks into the original project at the bit depth and sample rate of that project.

I do this when I record remote drum tracks for friends and clients, and they do not use the same bit depth and sample rate I do. Analog drum tracks use lots of mics and lots of effects and they benefit greatly from the higher sampling rate.

This is the same process sample CD and synth manufacturers use to record their audio samples before exporting and converting into the target CD or device.

But notice I did not upsample and then downsample the original tracks. I only upsample to play along and monitor, and then export the new drum tracks to be converted and imported into the original project.

Converting up and down can be very damaging to the audio. Bit depth and sample rate conversion is one of the MOST DISRUPTIVE things you can do to audio.

Don't do it more than once.

SteveD
www.DawPro.com
 
Steve,

Who in the heck is talking about upsampling or downsampling. I'm talking about upward Bit Depth conversion...about adding #'s for greater processing precision. Did you read my post?

I had a 16-Bit Project. I exported it as a 24. I listened very carefully to both. The composition begins with a long percussion lead in, to which I applied one of SONAR's stereo reverb effects. The 24-Bit file has much more clarity & depth (which can be heard best during this percussion intro.)

Bit depth conversion is one of the MOST DISRUPTIVE things you can do to audio.
Click to expand...
Hooey.

You seem to be mired in the technical end.

I used my ears. Upward conversion works.

But, speaking of the technical end:

"With the output of my workstation patched to the bitscope, I can watch a 16 or 20-bit source expand to 24-bits when the gain changes, during crossfades, or if any equalizer is changed from the 0 dB position." Bob Katz.
Click to expand...
And a 16-Bit source also expands with the addition of effects.

All Cakewalk is saying in their User Guide is; that exporting to 24-Bit will capture more of the expanded signal (expanded by the addition of FX.) More #'s equals greater precision. This is especially true of reverb, where more #'s captures more reflections, adding greater depth to the effect.

Granted, my source tracks (the dry percussion) were recorded at 16. I then added verb within the host app (which works via a 32-Bit float.) It seems to me to be simple logic that, the finer nuances of the reverb, which are heard during SONAR playback (because of the 32-Bit operation), are also captured with the 24-Bit export (& they were not with the 16.)

What amazed me was, I retained much of the quality when I converted downward in WaveLab w/ Dither !
----------------------------------------------------------------------------------------------------

mixsit,

"..I converted the bit depth back to 16; & dithered with UV22HR at the same time...& that 16-Bit file sounds better."...Than the 24 bit?
Click to expand...
No . . . better than the original 16-Bit export.

Not 100% sure what you mean about output balance
Click to expand...
My thinking was...maybe with the output set precisely at 0dbFS, the summing was equal when the file was written. Kind of like...for example, the pros will tell you that, for *downsampling* (SRC), if of course the destination medium is CD (44.1)...record at 88.2, which is an exact double. By working in equal divisibles, there are far less rounding errors during the downward sample rate conversion.

I thought maybe having the output right at 0dbFS, the same would hold true for bit depth conversion. We'll have to look into that one.

mark4man
 
Sorry Mark4, I got lost in here. I know you can import 16 bit files into a 24 bit project, but expot at 24 from a 16 project?
(There are two places in Sonar to set bit depth that i don't grock yet. Maybe that's it...)
Wayne
 
Last edited:
You must log in or register to reply here.
Back
Top