Beeps, let's be clear exactly what you are asking me to do.     And please never again suggest that all this is an attempt to hurt sonar or cakewalk.   Yes I get annoyed at bugs in software I use, but I'm completely disinterested in saying or doing anything for or against some corporate entity that I buy something from.  
 
So, I just did the Freeze (kontakt has four outputs, so that gave me four freeze tracks); then I bounced those freeze tracks into one track.  Posted above; it looks very similar to the 64-bit engine fast bounce, but I haven't tried to phase shift them.  
 
And the OP showed the difference between fast and slow bounce, long ago.    Same material, same set up. 
 
Step by step:  what procedure are you asking for? 
 

williamcopper
Beeps, let's be clear exactly what you are asking me to do.     And please never again suggest that all this is an attempt to hurt sonar or cakewalk.   Yes I get annoyed at bugs in software I use, but I'm completely disinterested in saying or doing anything for or against some corporate entity that I buy something from.  
 
So, I just did the Freeze (kontakt has four outputs, so that gave me four freeze tracks); then I bounced those freeze tracks into one track.  Posted above; it looks very similar to the 64-bit engine fast bounce, but I haven't tried to phase shift them.  
 
And the OP showed the difference between fast and slow bounce, long ago.    Same material, same set up. 
 
Step by step:  what procedure are you asking for? 
 

I don't know if you are for real or not but if you notice I am currently giving you the exact steps to test this without any opinion on your motives.
 
I posted the step by step.
 
Freeze you track so your single sample is turned into a wave.
 
Export 1) Real Time - 64 Bit engine = OFF
 
Export 2) Fast Bounce - 64 Bit engine = OFF
 
Now import these exports into your project into two separate but identical audio tracks.
 
Press the Phase button on one of the tracks.
 
They should null (no sound). This means the exports are identical.
 
Repeat the whole process except this time turn the 64 bit engine ON both times.
 
Again they should null.
 
 
Okay?

sigh.   and can't anyone see that if one process results in a output that is shifted by 30 ms from another process ... that's a problem?  Regardless of which one is "right"?    And if one process results in an output that has a different audio signal than another process, that's a problem?   Regardless of which one is "right"?

williamcopper
sigh.   and can't anyone see that if one process results in a output that is shifted by 30 ms from another process ... that's a problem?  Regardless of which one is "right"?    And if one process results in an output that has a different audio signal than another process, that's a problem?   Regardless of which one is "right"?

If there is a time shift there is another problem with syncing or how you are importing/exporting. That is another issue. Stick to what we are doing so we can get the facts.

williamcopper
last image from my free cw consulting project.   if this is not clear, then I don't know what else to do.   At my old billing rates, I've given CW around $50,000 of free advice over the last year.    
 
Freeze followed by 64-bit-engine fast bounce vs bounce, Real time, 64 bit engine.   CW nneds to get a decent testing staff, and the cw fanboys need to be a little less agressive toward forum posts that are critical. 
 

You know what, I just wasted a bunch of time trying to help you troubleshoot despite the fact I seriously am not fond of you.
 
You are NOT consulting or helping solve "problems". You are misunderstanding a whole pile of stuff, not following instructions for troubleshooting and being belligerent. I do not think CW (or any half decent comapany) would pay for that.
 
These images mean NOTHING. There is nothing really that can be deciphered because the resolution is too low to see the samples/transients.
 
Did you follow my instructions to the letter?
 
Is that what produced those waves?
 
What I CAN see is the time shift. That is some kind of sync/import issue as I've been saying ALL ALONG.
 
Now take one of those waves/clips and use the Nudge feature to move one wave into sync with another (the ones that used the SAME settings). As you nudge them they will eventually sync up and null IF you did this correctly AND are comparing the exports I told you to compare.
 
So even though you are being abusive towards me, Cakewalk and other forum users you are STILL getting help. You have also wasted a good chunk of my morning.
 
Sorry if I'm not all daisies and roses but you are not a pleasant person to deal with and again the only reason I'm doing this is so OTHERS can check this stuff out for themselves and come to their OWN conclusions.
 
If you HAD followed the procedures I laid out for you you would also have a thread to send to Cakewalk support so they could help you sort out any system/hardware issues that may be causing issues.
 
I have other stuff to do now so you are on your own.

48Gb of Ram used?
 
Sorry, there must be something wrong here surely?
 
How many tracks are you talking about? The most any of my projects consumes is about 9Gb (135 tracks, running EWQLSO Platinum plus via 6 instances of Play), multiple Fx, busses, drum maps (for artics)

williamcopper
Ok, thanks beepster.    I too have other things to do ... been about 6 hours on this today.     One final observation: all of this is in a large project, about 48 GB memory used.   After doing the assorted bounces and freezes I was somewhat shocked to see suddenly nearly 64 GB memory used, and the little Performance indicator showing Red alert for memory.    Nothing added, all the various bounces deleted from the project after they were created.  
 
 

Take a break and do all these tests, as I typed them up, to the letter in a clean/new project. This is pure science and needs to be approached as such which means strict controls, consistency and attention to detail.
 
If your RAM or any other hardware stuff is going wonky that of course could indeed be tainting the results which was something I was going to mention after the initial tests were completed.
 
The time shift could simply be because your system is being maxed out so use a new/clean project. I THOUGHT that was what you were doing (or at least a reasonably clean project).
 
Like I said... it is science and needs to be strictly controlled for accurate results. If you truly want to find out the accuracy of Sonar's export outputs you need to do it in a "sterile" environment first. Then you can expand/extrapolate the experiment to larger/more complex projects which will reveal where the problems may be occurring.
 
Right now we are at "petri dish" level. You were starting from "terraforming" level.
 
It's all moot anyway if the end product sounds good BUT if there are syncing issues that's a problem that needs to be fixed. If there are seriously AUDIBLE varitions that are destroying your work that again is a problem that needs to be dealt with.
 
If we are only getting imperceptible variations then you are just twisting your own nuts in a knot over the theoretical. That is NOT conducive to the supposed end goal of creating/producing pleasing music.
 
Knowaddimean?
 
Cheers.

williamcopper
Bristol_Jonesy,  the project in question is a large one, but it is far from the largest I've used.     The memory size is mostly from fully loaded Kontakt instances for a full orchestra with effectively 3 string sections made up of VSL solo strings, VSL chamber strings, and OT Berlin Strings plus a bunch of woodwinds, brass, and percussion.    It's about 14 minutes long at present, maybe 40,000 midi events.    But this is not a memory issue ... the observation about memory was from surprise after several hours of bouncing, freezing, exporting, and importing, and including running Samplitude and Sonar at the same time.   
 
I don't really know how long I've spent 'consulting for cakewalk', so maybe the estimate above is high.   But it is many many hours over a year devoted simply to finding out how and why Sonar fails at different things.   And, sure, sometimes operator error -- that happens in any craft or profession, but as a consultant you bill for it anyway! 

Do you're tests in a blank project!!! Cripes... you have been doing this in THAT project the whole time? You could have a TON of crap interfering with your exports. Sends, synths, effects, hardware, etc etc etc...
 
If you want to blame Sonar then Sonar can be the ONLY variable (and even then there are still hardware/system/settings issues that could be going on which would need to be ruled out).
 
As a consultant... who is supposed to be educated... and get paid... and you don't grasp this?
 
Sorry to say but... you're fired!

williamcopper
last image from my free cw consulting project.

 
And that free consulting is worth every penny.
 

if this is not clear, then I don't know what else to do.

 
I recommend you study up on "the scientific method." Control groups, variables, test conditions, test environments, methodology, etc.
 

CW nneds to get a decent testing staff

 
See below.
 

and the cw fanboys need to be a little less agressive toward forum posts that are critical.

 
Critical posts are accepted in the spirit in which they are given. The reception to your posts confirms that premise.
 
Please observe the following screenshot. Here is the methology.
 
The top track is the original audio track from SONAR recorded at 44.1 kHz. The second track is the result of a real-time bounce in SONAR with 64 bits unchecked. The third track is the result of a fast bounce in SONAR with 64 bits unchecked. The fourth track is the result of a real-time bounce in SONAR with 64 bits checked. The fifth track is the result of a fast bounce in SONAR with 64 bits checked.
 
The bounced WAVs were dragged to the desktop. They were then brought into Sony Vegas, to remove the variable that an error might be compensated for within SONAR and therefore not appear in that context.
 
In the screen shot below, the "10" calibration refers to 10 samples. That is 202.67 microseconds. Your "30 ms delay" is the result of your inability to conduct a meaningful test. The tracks are aligned with sample accuracy and are identical in level. They all null perfectly if you throw one track out of phase compared to the other.
 
If you find out what you're doing that produces the anomalous results you obtain, let us know.
 

 

williamcopper
 And, sure, sometimes operator error -- that happens in any craft or profession, but as a consultant you bill for it anyway! 

Maybe you do. I don't. If I make a mistake that takes me down the wrong path, unless the company asked me to make that mistake, I will not charge for my errors. Anyone considering doing consulting should realize that this is standard operating procedure. For example if you lose an hours' of work because you didn't back something up, you have to eat that hour, not the company paying you. 

williamcopper
Craig, that is really not helpful, though i suppose you may be trying to fool other readers. 
 

 
Irony... meet head desk.
 
Head desk... meet irony.

Okay... so hopefully everyone watching can see how ridiculous this OP is. Don't let him/it waste your time.
 
cya...

williamcopper
Craig, that is really not helpful, though i suppose you may be trying to fool other readers.

 
Right. And to make sure people would be fooled, I provided all the details about the test and the methodology so that people could reproduce it themselves. Unlike you. 
 

The issue is not bouncing of audio.
 
The issue is bouncing of midi data fed to Kontakt to produce audio --- 'bounce to tracks' applied to midi tracks and their audio output tracks. 

 
 
Well, that helps, because your first sentence was [Italics mine]:
 
"See image.   Same music, slow bounce vs fast bounce.   While they align and sound similar, clearly they are not the same."
 
I took that as you were mixing down an entire composition of audio tracks, which were generated by bouncing Kontakt to create that audio. If you had been that clear at the outset, it would have saved me quite a bit of time. However, it's fortuitous you clarified that because in that case, I have a much larger body of data to draw on than the quick experiment I ran a few moments ago.
 
I just finished all the mixes for my new album "Neo-." These mixes have been done over a period of several months. I don't recall for sure if all the tracks used multiple instances of Kontakt, but many had five instances. In any event they were typically a mix of audio and multiple virtual instruments playing back in real time; some songs had tempo changes, some did not. Some synths and processors were upsampled, some were not; there was even a mix of 16- and 24-bit files, as well as looped and non-looped files. Some of the mixes were done using a real-time bounce because I wanted to hear it "one last time," and some were fast bounces because I was in a hurry. Some used the 64-bit engine and some didn't. In other words, it would be hard to imagine a situation with more variables.
 
I bounce all the tracks down to a final two-track within the project for reasons I've explained in other posts, so I won't go through that again. Typically, there will be four to ten bounces during the course of mixing a song, each with slight changes so I can compare them. Furthermore, I sometimes cut areas from one mix and put them in a different mix. As a result, I often hear separate bounced mixes play back simultaneously, and when deciding which intro to use, am zoomed way in on the beginning the tracks.
 
Throughout the entire creation of the album, no bounced mix exhibited a start time offset compared to either the original song, or the other tracks. 
 
 

I've edited the OP to clarify.   The operation that produces audio from a VST in association with midi tracks is called BY SONAR "Bounce to Tracks" ... Select Midi Tracks and VST Output Tracks, Select a time range, Track View - Tracks - Bounce to Tracks.    
 
Surely you'd not have me call it something else? 
 
Privately, actually, I call it "YO" .. because I've mapped a shortcut to solo all of a group of selected tracks as "Y" and a shortcut to use the Bounce to Tracks Dialog as "O".    
 
 
edit.   I don't generally use 'Freeze', because that involves a rather time-consuming unload of all the samples from Kontakt -- I have enough memory for what I do that this is not necessary or desirable. 

Come on Beepster.   I'll readily agree that I don't write well or clearly.  But help me out ... I think you know what I'm trying to show, why not say it clearly as you believe it might be better understood?