Somewhat Solved: Superior Drummer 3db louder after freezing

[EDIT: See posts #11, 36 and 37 for explanation]

This has been an occasional annoyance for a long while, and I've just procrastinated looking into it. But it's a quiet rainy day, so today's the day to explore it further. I'd be curious to know of others' experiences with this and if you are able to duplicate it. And of course I'd love to hear any hypotheses!

Here is a kick drum track (Superior Drummer 2). Note the peak marker at -5.2db.


Here is the same hit on the same track after freezing. Note the peak marker now reads -2.6db. On playback, the kick drum really is audibly louder.


Here's the snare track from same project before freezing. Note the peak marker reads -16.9db.


After freezing, the same hit on the same track reads -13.9db.


Same project, same soft synth, but a stereo output into a stereo audio track. Initial peak value is -15.5db:


After freezing, the peak value is -15.3db, a mere 0.2db increase.

However, two other stereo tracks (toms and overheads) exhibited different behavior, both increasing the peak level by 2.8db.

This inconsistency is more troubling than the level changes themselves. If there were a consistent difference, I could at least compensate for it on the drum bus. As it is, I cannot mix with any reliability or repeatability if any instruments are frozen. And not freezing tracks is sometimes not an option with this underpowered computer.

---

Here are some observations:

> Freezing this synth consistently raises the peak level by up to 3db. Stereo channels tended to have smaller increases than mono channels.

> This phenomenon is not specific to Superior Drummer, but also does not apply to all soft synths, nor is the level increase consistent for all synths.

> Jamstix and TTS-1 both exhibited the same behavior after freezing: 3db increase on mono channels, a smaller increase with stereo outputs.

> Two different stereo Dimension Pro tracks were tested. One exhibited no peak change, but the other went up by 0.3db. These were two instances of the same synth in the same project, but with different (stereo) patches on each.

> A stereo Kontakt track went up by only 0.4db

> Results are the same regardless of whether or not the Fast Bounce option is enabled in the Freeze Options dialog.

> Straight audio tracks do not exhibit this behavior. Their peaks are exactly the same after freezing, whether mono or stereo.

> SONAR 7 behaves exactly the same way as 8.3.5 in this regard.

I have a bunch of half-baked theories bouncing around in my head (most having to do with stereo-to-mono conversion and pan laws) but so far none that explains all of the above observations.

Comments?

That's actually the next thing on my list to try: different pan laws. If pan laws are at all involved, switching to a different one should result in different levels.

Incidentally, just tried it with Kontakt 4 x64 as well.  No deviation at all on peak levels after freeze (no humanize function on this). 

Dean

Thanks for trying, guys. Yeh, it's inconsistent. I first noticed the phenomenon back at SONAR 5, before SD2 even existed. Over the years it's been sporadic. That's why it's a mystery.

I love mysteries.

I hate mysteries.

Bit, do you get the same inconsistenices when bouncing (to track), as opposed to freezing?

I'm sure I don't have to tell you this, but make sure you don't have any volume automation screwing up a bounce.

Mystery solved. Sort of. I consider this a bug and am not happy with the resolution, but I'll describe the reason for the problem and the workaround, which may be of interest to other Superior Drummer users.

When I switched to the -3db center pan law, the volume increase stopped happening. That clue made me realize that the root of the problem is that Superior Drummer's outputs are stereo.

I have the kick drum mics routed to bus 1/2 in SD2's internal mixer. That's a stereo bus, so at that point the kick drum is obviously stereo. (Or, more accurately, duplicated on a mono pair.)

However, I had selected "S1/2 left (mono)" as the source for my audio track, assuming that this meant a mono output (the left side of the bus). There was nothing visually to indicate otherwise, since it played back mono and the frozen waveform was visibly mono. (Plus the word "mono" in the input selection seemed to strongly suggest mono.)

But although playback was indeed properly mono, for some reason during freezing the audio was processed as stereo (both sides of the SD2 bus even though only the left was selected) and folded down to mono, resulting in a 3db increase due to the 0db center pan law.

So that explains the mystery. Now, what to do about it? Changing the pan law is not a viable option for a 99% mixed project.

The workaround is to set the kick and snare tracks' interleaves to stereo. This causes the 3db increase to happen during playback as well as during the freeze. Although it irks me to see a stereo waveform in my kick drum track, this will allow the drums to be mixed without worrying about unexpected level changes after freezing.

I'm not at home to check for myself (I certainly will when I get there) but - does SD's internal mixer ALWAYS have stereo pair for bus assignments or is that only if you choose ALL STEREO OUTPUTS for SD when inserting it as a softsynth?  if you had chosen ALL MONO OUTPUTS for SD would the internal mixer still give you a stereo bus?

even if it did, can you route directly to the MONO outputs and then effectively not have this problem you're describing?

(I understand that you might want a bus in the SD mixer for routing/FX purposes).

I'm just saying that if bit is chosing LEFT 1/2 MONO for his source and the bus is actually a stereo bus that apparently (and I agree that it's errant) sonar interprets that the user wants to sum the bus to mono, but maybe if the output of that bus was MONO already then sonar wouldn't interpret that it needed to sum anything and then the levels would be correct.

ba_midi



Agreed about the summing.  That could do it.   But I still think freezing should be 'as is', so that what you see/hear is what you get.
 
 

Not arguing that at all - I agree it shouldn't sum the bus when choosing LEFT 1/2 MONO as source.  it should choose only the left channel of the 2 channel stereo bus.  apparently it doesn't - and that's not what it should do. 
 
I'm just curious about how it would interpret the output if MONO outputs are chosen instead of stereo for SD.

I'd just get more horsepower so I don't have to freeze.

Shoo! Back to the coffeehouse with you, you vagrant!

I'm just curious about how it would interpret the output if MONO outputs are chosen instead of stereo for SD.

You get the same thing whether you specify mono or stereo outputs in the Insert Synth dialog. There simply are no true mono outputs available from SD2.

I don't use that method (automatic track creation) with SD, since I've never had a need for 32 outputs from any sampler. But I just tried it for grins, and you do get all the "left, mono" and "right, mono" combinations. The first track's assigned input source is "S1/2: left (mono)", the same bus I normally use for the kick output.

There is no question that this is a bug. The amplitude of a track should not change when frozen. Whatever rules and processes were being applied during playback should be exactly the same as what are applied during the bounce. If it's a multi-output soft synth in which some instruments are boosted but others are not, then that pretty much makes freezing of soft synths unusable.

There is no question that this is a bug. The amplitude of a track should not change when frozen. Whatever rules and processes were being applied during playback should be exactly the same as what are applied during the bounce. If it's a multi-output soft synth in which some instruments are boosted but others are not, then that pretty much makes freezing of soft synths unusable.

 
I totally agree.  And I also think it's a bug.   This is one I HOPE CW is looking into.
 

So I'm wondering why it matters if a kick is a stereo signal but is actually 'centered' vs a mono kick? If you put a mono source signal in a stereo track that ends up "stereoized" but is still mono, right?

That is correct. If you import a mono file into a stereo-interleaved track and bounce it, the result will sound the same because both sides of the stereo file contain identical information. You'll just be using twice the disk space and RAM for no benefit.

I cannot tell if the SD2 samples are stereo or not, as they are in a proprietary format. However, my guess would be that they are not stereo recordings. I believe they are all mono. The busses in SD2's mixer are all stereo, but they make each side of each bus available as a separate output, so you can get mono outputs from them.

By selecting the left channel of a bus, SONAR should be seeing effectively a mono output. Which it does, at least during playback. Just not during a freeze.

One question comes to mind,  will the exporting of the audio either with frozen tracks or without exhibit the same results?  In other words are the same problems going to arise with a total project bounce down that are occurring with the individual track freeze?

I have not tested an export yet. If the tracks are frozen prior to the export, then yes, the export's going to be messed up as well. But whether the export does the same thing as the freeze, I have not determined that.