Bouncing not nulling!

I´m playing a kick drum sample realtime from Kontakt, I have 3 insert plugins on that VSTi track (Waves SSL Channel, Waves API 2500 and EQuick).

I bounced that track without the insert FXs to an audio track and then copied the same FXs to that track.

Then I did a null test. Flipping the phase from the audio track and playing that track same time as the Kontakt track. It won´t null perfectly. There´s some random “clicks” here and there from the kick drum.

Also when I´m ABing these two tracks I hear/feel that the Kontakt playing the sample in realtime has a bit more bass “oomph” in it than the bounced track. I don´t if it´s just confirmation bias, but I could swear I hear extra low end. And because it won´t null it´s definitely not exactly the same.

So does Cubase somehow degrade the signal when bouncing (Audio Mixdown)? Something wrong with the audio engine?

I´m running Cubase 6.0.5 64-bit.

You should not expect nulling in those circumstances. Many plug-ins introduce subtle inconsistencies into the signal as part of their design. Get rid of the plug-ins and the signals will probably null. As for the bass in the kick, I don’t know. What bit rate did you use when you bounced the track?

Ok, I tried again without the plugins and the results were the same. Didn´t null.

All the files (and the project) are 24-bit/44.1kHz, including the bounced kick… hmm except the sample that Kontakt is playing realtime is 16-bit/44.1kHz. Would that make the difference?

Would bouncing a 16-bit sample to 24-bit file degrade the signal? Taking a little bit low end out from the kick?

1 - Those 8 bits will make a difference, subtle but its a difference.

2 - I wouldn’t say degrade, raising the bit rate shouldn’t degrade the signal nor make it better.

To complicate things a bit, Cubase works internally at 32-bit float.

So that null test is pretty much doomed.

Those two points are kind of a contradiction. :slight_smile: If it will make a difference (although very subtle), then the signal is different = better or worse (that´s of course subjective). But I´m hearing less low end so to me that´s degradation.

Now the question is, how can I bounce this so that I´ll get a perfectly nulling audio track? I really want to bounce it, but I want to be sure it´s exactly the same as playing it straight from Kontakt in realtime.

Ok, first off, this has nothing to do with s.rate/bit depth. You might be triggering 16 bit samples while your project is set to 24 bit - but that has no affect on the null vs. not null.

Secondly, KBJRock > any difference that there is (and there is a difference - more on that in a bit), you will not be able to hear. With respect, if you think you hear more lows or less lows in one than the other (mixdown vs. playing back and triggering MIDI data), you are imagining things. This is assuming that the levels are the same; that the master bus was at 0 when the mixdown was done. If it wasn’t - that’s a different story.

Finally, there is this. When you trigger mulitple Kontakt samples in succession, they do not always play back exactly the same. Why, I don’t know. I do not know if this is a Kontakt thing, or a Cubase thing. I have not tried this with PLAY…maybe I should. This is evidenced by loading a drum patch in Kontakt, creating a MIDI file which triggers a bunch of samples from that patch (they should not be RR samples - b/c then we will almost likely not have two nulling files regardless), and doing two real-time exports. Those two audio files will not null. Again, I don’t know why. AFAIC they should, but they don’t. Now, if you open say, a violin patch, and lay down say, three notes that are several seconds apart from one another, two resulting files from rt exports will null. Assuming there are no plugs with any randomness, and that there are no Kontakt inserts.

This is not a concern for me, b/c the difference is inaudible; it’s usually around -75db. Nobody can hear that.

Now the question is, how can I bounce this so that I´ll get a perfectly nulling audio track?

If you do two non-rt exports, those files will null. But a non-rt export will not null against playback, b/c remember - the samples do not always play back exactly the same.


–edit— just checked this in PLAY - same thing. It appears to not be a Kontakt thing, but a Cubase thing. Support ticket time.

I remember doing a null test few years back with PLAY and it didn´t null (No FX, no RoundRobin, same bit depth). I didn´t pay enough attention back then and just hoped this would be fixed sooner or later, I guess not.

I don´t know if it´s in my head, but I can hear everytime some low end missing, even in the whole mix using the bounced kick vs the realtime kick.

Null tests, unless you REALLY know what you are doing are just that.
“Oh, dear! My bass drum’s a millimollimicrometer off!”
If you’re not recording signals off the moon and lining them up with Jupiter at 5 oclock I wouldn’t worry your sweet heads about it.

If they removed it it would save so much grumbling and time-wasting.

Maybe I’ll start requesting that certain features be removed. :open_mouth: We’ve got too many already. :mrgreen:
If one is working from a bedroom you shouldn’t really be worried about nulling.

ps The nulling illustrated here is meaningless as no two signals from different sources will ever quite soung the same over time.
The only time you check for TRUE nulling is when you have copied a track.
ie: Bouncing [Track A] GA1 drum machine bass to a [Teack B]track will likely not null. COPYING that (bounced) track to ANOTHER bass drum track [Track C] should then null with the first bounced track.
Track A will not null to Track B but track B should/will null to Track C.

Now have your cocoas and sleepybyes and tomorrow it will have all been a bad dream. :mrgreen:

Why so hostile?

I´m just wondering how will I get that kick sample bounced to an audio track, so it´s 100% the same. It can´t be that difficult in a modern DAW, or is it?

Same goes for PLAY samples and everything else I´m bouncing.

You can’t get 100% same, if there’s programmed randomness in the virtual instrument. For your situation (hearing less low end on bounce) I would recommend:

  1. Try to use use “realtime export” instead of bounce, if it makes any diference
  2. Try to use output-to-input loopback virtual audio interface and record the signal from there instead of bouching (you may have this feature built into your audio interface …whatever this interface is … you are not telling us what it is).
  3. Just plug a cable from your audio interface’s output to it’s input and record what you get from there. Yes, it passes an extra D/A and A/D stages, but if you have any decent audio interface there should be no significant deterioration in the sound.

It´s just one sample played back by Kontakt, so there shouldn´t be any randomness.

  1. I´m using the “Realtime Export”. I don´t trust the non-realtime export at all, too many bad experiences with it.

  2. I´m using Symphony I/O. I don´t know how this is done? I understand that in Pro Tools you can just record the Master Bus to a new track and many swear by it. Like Dave Pensado for example never bounces his mixes, he records/prints them to a new track. He swears he can hear the difference between printed vs bounced and the printed/recorded one always sounds better according to him.

Can Cubase record the Master Bus to a new track? That could work.

  1. I don´t trust this solution at all. It just has to degrade the signal, ´cause I can even hear differences in cables too.

Even if the programmer of Kontakt has introduced some ramdomness on playing samples?

I don’t know how Dave Pensado does it and I don’t care. I never bounce or export my mixes, but that’s because I use external equipment to do my mixdowns. If I did it ITB, I would use “realtime export”, since I have found some plugins to mis-behave while exporting non-realtime.

But why don’t you just select “Outputs - Stereo Out” as an input of your mixdown track?

Does he have any statistical A/B/X test proof for his claims? If not, then … yes … I can bet he can hear it … but if there’s any real difference … that’s another issue.

OK. Case closed. Please continue discussion at audiophile forums.