Record Shift Null Test: VSTi Track vs Mixdown vs Frozen

96k/32 bit project
256 buffer
MIDI is quantized to 1/4 note
Matched PB levels

+96 Audio Mixdown
+/-0 Audio Mixdown
-96 Audio Mixdown
VST Intrument Track: Frozen
VST Instrument Track: unFrozen

All exported Audio Mixdown tracks null to almost silent to each other. As expected the mixdowns that are shifted the least from each other null the most but the differences are negligible and I could attribute to round robin of drum VST.

Neither VST Instrument Track, frozen or not, nulled to any of the Audio Mixdown tracks to any degree that was noticeable for practical purposes (like acceptable sync).

Any comments regarding the results and what my next step would be to get the VST Instrument Track nulled with the Audio Mixdown Track?

Thanks in advance.

What do you want to achieve?
You do know a little latency jitter due to buffer changes on a MIDI track connected to a VSTi is inevitable?

Niles thanks for the comment I changed my post to be more clear.

Are you talking about rt mixdonw? With non-rt, I have never seen that here. I can do 20 mixdowns of a MIDI track connected to a VSTi, and as long as there are no RRs, and there is no random-algo anywhere, they will all null down to -inf. So there really is no latency jitter in regards to VIs and non-rt exports. Rt…perhaps; never tested that.


As expected the mixdowns that are shifted the least from each other …

Shifting? What did you shift?

Neither of the VSTi tracks, frozen or not, nulled to any of the mixdown tracks to any degree that was noticeable …

Noticeable? Either the tracks null, or they don’t. If you swap the phase, hit play, and see -inf on the master bus, the tracks have nulled. If you see -117db (for ex.) on the master bus, they have not nulled. Of course you would never hear the difference, but that’s not a null.

I have no experience with freeze; I think I used it once, and that was only because I was curious to see if it actually worked. Is it possible that when the tracks are frozen, there is something else going on during a mixdown (slight, if anything) that would cause slight differences? What type of mixdwon ar eyou doing? Realtime, or non?

No, only talking about live playback of the MIDI channel connected to a VSTi and a recording of the playback.

I’ve not involved frozen tracks…but, I’ve done extensive testing on the midi/VI timing.

Real time mixes do not null with offline…BUT…they consistently null against repeated attempts, even with crazy amounts of other latency inducing plugs inserted all over.

Midi does not capture what you play (accurately)…but, with recent versions of Cubase, once the midi is captured, it maintains the timing consistently–save offline rendering. My guess is freezing basically does an offline render…so, what you’re experiencing is the same thing I found.

Comes down to…if I can play it with my ten fingers, I record audio straight away. VI or hardware. If I have to sequence, I render in real time as soon as I get it where I want it…them freeze, unload, mute the midi/VI.

I’'m not familiar with the term, “midi/IV”.

He didn’t say midi/IV

midi = Musical Instrument Digital Interface. It is an electronic musical instrument industry spec which enables a wide variety of digital musical instruments, computers, and other related devices to connect and communicate with one another.

VI = virtual instrument.

I’d like to try and help, but you didn’t answer any of my Qs.


What exactly is your problem? You start describing your trouble shooting, but what was initially wrong?

Why do you think they should null?

There’s rather a lot of situations where null doesn’t happen or doesn’t matter. It is fairly well documented that it is difficult to stack drums which I think may be at the bottom of this.
Other midi’d instruments are easier to stack because the inaccuracies can make it sound pleasant sometimes but maybe not always. Drums, being short and staccato in character are very prone to any inaccuracies present.
It would not normally be anything new if stacked / overlaid drum tracks didn’t null.

Any comments regarding the results and what my next step would be to get the VST Instrument Track nulled with the Audio Mixdown Track?

Not always possible. Why?

Before I rendered the drum kit instrument tracks they were “in the pocket” with all the unrendered percussion instrument tracks. After rendering, the various rhythms no longer worked together. If there is no way, if it’s impossible to null the unrendered drum tracks to the rendered drum tracks then it seems an indication that I will never be able to get the drums and percussion to be tight. In anyone’s experience, does real time rendering solve this issue? It would be time consuming but if it allows me to continue working in Cubase it might be worth it.

As always,


In my experience, if you need critical timing for your dance or hip hop beats you have to manually place audio files. The VIs are always sloppy. The only DAW I’ve ever used that had rock solid MIDI timing was DP, but at the time it crashed so much that I left the program.

I’m not so concerned with MIDI timing per se. The unrendered VSTi drum and percussion tracks are working well together. It’s only when I render that there is a problem between rendered and unrendered rhythm tracks.

What version of DP were you using at the time? I understand it’s cross platform now. I used to use MOTU hardware back in the early 90’s and had problems on my PC. I got them sorted but I remember it being very time consuming. Never used their software, though.

FWIW, as much as I like intelligent hip hop (not a big fan of rap) such as DHoH, DFH, KRS-One, Aesop Rock, etc., etc., I would never presume to insinuate myself into that genre. I’m currently working on a post-tonal piece that relies heavily on polyrhythmic percussion. I hope this provide a bit more insight regarding my concerns.

OK, I’m going to take a stab at this. The unrendered MIDI has to go to the VSTi before being routed out the interface to come out as audio, but the rendered audio skips that step. Wouldn’t that be enough to cause them to be out of sync with each other?

It sounds like a reasonable theory. If Cubase can’t compensate for this then I’m left wondering, “Is there a DAW that is better suited to creating complex rhythms efficiently ?” It’s a can of worms that tends to attract a number of nasty insects if you catch my drift but I need to know. I guess I’m dead in the water for awhile. If there isn’t a solution in Cubase I’ll start posting my question elsewhere.

Well you could check out DP. Back when I was using it was version 4 so probably a lot has changed but I"m sure it still has tight MIDI timing. If you have a MOTU MIDI interface (I have the MIDI Express XT) it offers sub millisecond MIDI timing even on your external MIDI devices with time stamping.

The other thing it has which I really miss in Cubase for doing crazy rhythms is the ability to step input a “n-tuplet”. Useful for doing those flute glisses etc where you want the highest note in a gliss to line up on the downbeat.

One more thing, I’ve noticed that there seems to be some undocumented maximum amount of latency compensation in Cubase. If you try using a buffer of 2048 you might find the timing starts to be off. So try setting your buffer lower and then doing the offline bounce.

How is the timing when you export the entire project from MIDI to Audio as one stereo file?

Yes, rendering in real time fixes timing…want to absolutely sure? RECORD THE VI audio output in real time while you’re listening to it. Hell, record the percussion at the same time to a different audio track.

You should always render to audio once a track is where you want it. I have found that, yes, real time is what has to happen…particularly for drums and piano (sharp percussive attacks).

I used DP3 year ago, but all external midi devices (including a Gigstudio PC)…OS9…MTPav…solid as a rock. But, honestly Cubase 4+ with a MoTU interface has been tight, too. When you’re talking about rendering VIs in faster than real time, that’s a completely different scenario than midi timing being tight or not.

I would say that Steinberg should just remove the freeze and offline renders for VIs if they can’t get it working 100%. it effectively IS removed as far as I’m concerned after my own testing. I’m used to real time bounces…when you work as much as I do with external audio gear, there is no other way. Render…get coffee…come back…just a natural part of my workflow.

All good stuff, thanks. I’ll try raising the buffer, exporting the whole project (should be very telling), and rendering in real time. Of course, at first, only one at a time. Also, what Alexis brought up got me wondering if using a “VST Instrument Track” instead of a normal “MIDI Track” assigned to a VSTi would make any difference.

Robotpriest, back in the day I think I remember MOTU touting their MIDI timing as a major selling point; I wonder if the timing of my MOTU Express PC spoiled me very early on. I actually was doing a bit of 4x4 dance stuff at that time. Hehe, good times.