Cubase mixdown sounds different

[luigi71cx3]

Good morning everyone.

I'm not sure that this is the right place to post this question.

I am experiencing a quite unusual situation with the mixdown in Cubase (hopefully due to my limited skills).

My project consists of a few tracks VST (Kontakt player) and a physical Virus TI Desktop (external then connected via USB). The Virus TI is seen from cubase as a normal VST, with the difference that is the Virus TI to generate sounds instead the PC. The sounds thus generated come back into the PC via the USB connection.

While listening to Cubase (playing back the song) everything sounds perfectly, but when doing the mixdown and then listen to it (whether mp3 or wav), some tracks played by the Virus TI sounds without all the voices , for example, if a track consist of an arrangement for three voices , only one voice is played instead.

To make it even more difficult to understand the problem is that if I do the mixdown only for the portion of the song where the problem appears (including few bars before and after), everything sounds perfectly , like listening in Cubase (playing back the song).

W7 x64, 8GB RAM, intel i5, Cubase 6.

Any suggestion is greatly appreciated , thank-you very much to all.

[TweakHead]

An easy fix would be:

record each Virus part individually! this way you have all the DSP power (that translates to the number of voices) available and therefore no note stealing.

Your problem seems to be just that: the Virus has a limited number of voices and DSP power. For complex patches, the number of voices can get high even with just one oscillator going (if using unison, for example); when it overloads, some voices won't be played at all 'cause there's no system resources left to reproduce them. Check the manual, you can set the priority of each channel, setting a channel to high priority means the system will preferably play this channel over others in the event of an overload.

Think of the multiple channels as something you use for composing, arranging and for previewing stuff only. Afterwards, record each part like you would with a Moog or a guitar. Slight change in workflow, but seems like you have no problem at all ;) cheers

[luigi71cx3]

Think of the multiple channels as something you use for composing, arranging and for previewing stuff only. Afterwards, record each part like you would with a Moog or a guitar. Slight change in workflow, but seems like you have no problem at all ;) cheers

So I am obliged to make audio files each Virus track, correct?
The behaviour of the Virus DSP is then different between playback and mixdown?

I thought I could avoid this approach and use the Virus as a VSTi thus saving a lot of disk space.:sad:

I can not explain, however, why doing the mixdown of that portion the problem disappears.

Addendum: at this time I'm making some tests.
The project has three tracks played by the Virus (one is a single voice and the others are chords) and (during the mixdown) seems one of these is played at a different pitch...

[MBTC]

But playing the song in Cubase with your MIDI tracks going straight to the Virus sounds okay in real time, right?

If I'm understanding properly, it sounds related to TI in the sense that the export process of the DAW would need some pretty tight cooperation from the Virus DSP in terms of rendering the audio the way the DAW wants to be "fed" so to speak. So it's possible that the DSP in the Virus is doing some extra work during the export, requiring it to take away some of the polyphony that you hear during regular playback, which would correspond to what Tweak is saying.

And his proposed solution is one possible solution -- and it's what I have to do anyway with non-integrated hardware that's just going straight into audio ports on the interface (Leipzig for example), but yes that's quite limiting in terms of workflow flexibility. Have you tried freezing the MIDI tracks in Cubase first? If it's a DSP resource issue on the Virus, that should reveal it or solve it and is a little more forgiving than treating everything as audio clips (at least you can unfreeze and modify while editing and freeze just prior to export).

I'm not sure any hardware instrument helps save disk space, because they are almost always going to leave you holding more audio clips which of course are huge compared to just MIDI data + patch settings for soft synths. Where you get the benefit from hardware is taking a load off your DAW CPU which there never seems to be enough of, especially working with the really good soft synths which are CPU hogs.

[TweakHead]

You can't just bounce down to audio (offline) with the Virus, you actually do have to record it as with any other external instrument. The DSP doesn't in fact have anything to do with any audio rendering of any sort, it's just busy being what it is: a synth engine. The TI side of things is more about being able to edit parameters and record automation within the DSP, total recall of presets within a project, to act as a librarian software, and for sample accurate timing - which I believe includes some kind of latency compensation using it's own audio interface, aligning the phase with the note on information (maybe?). The other advantage is that it dispenses with audio routings, involving at least 3 pairs of inputs so that the channels available are all put to good use; streaming this channels with the usb connection.

Think the important part is really the DSP, though. Having this kind of convenience in a hardware unit that's got such a wide sonic scope is already good enough reason to have it, but any editor, even using Midi can give you most things attributed to TI, with the exception of the sample accurate timing - not exactly a problem if you know your latency, you can deal it on a sample delay (negative value) and it's done.

On topic now. On the mixer you can activate and deactivate the parts. Presumably, if only one part is playing at a time, the engine will have all its power available for it - not divided with the other parts. The more voices you have going at once, the sooner you'll hit the ceiling. That's the sole reason you should record each part alone.

[luigi71cx3]

Yes playing back the song everything sounds fine.

About saving disk space, in the case of the Virus, this is due to the hybrid purpose of the Virus: it is an hardware component, but the tracks are played as a normal VST, therefore, in addition to a less DAW CPU duty, no further disk space is required for the Virus tracks also. In othe words the Virus runs in Cubase like any other VST, with the only difference that the sounds are generated by the Virus DSP in realtime, so no audio clips are required.

After some test it seems that the issue appear while using some high DSP consuming patches.

So I think to solve the issue, the best approach is to test the mixdown from time to time and transferring in audio those tracks based on high DSP consuming patches.

Still remains the fact that rendering only the portion of the song that contains the malfunction, everything runs smoothly.:confused:

[MBTC]

Ok this may or may not help, because there are a couple of big differences in my setup:

(a) I don't have a Virus currently so I'm using my Ultranova which it's own VST and USB integration similar to the TI series
(b) I'm using Cubase 7.5 and I'm not sure if parts of what I'm about to write are the same in 6.0 but hopefully the principles still apply

My Cubase template project for the Ultranova has two tracks that apply to that hardware synth -- one for the MIDI and one for the audio output. It's been a while since I set it up that way so I don't remember all the reasons why, but I seem to recall doing that because if not the Nova output would go straight to stereo out, removing a lot of mixer flexibility. If your setup is not like this, you might consider going this route. If I wanted to I can just capture the output from the Nova on that track, working with it independently from the MIDI data and the Nova, which comes in handy anyway since it isn't a multi-timbral synth (the Virus of course is but there are times when the DSP power might nag at you to work with only one patch at a time anyway). So it's bounced to audio at that point but it's not necessarily a final commitment to that audio take as long as you can get back to the same patch on the Nova if you need to modify and record again.

But if I am sure I only want one sound out of the Nova for a track, I can leave audio capturing out of the picture as long as I do a real-time mix down, and that's the key setting in the export dialog that I think is your best hope. If it's not captured in a real-time mix down the synth DSP probably can't keep up and produce the audio efficiently enough to meet the demand of the DAW. So if I exported batch channels without ticking the realtime box, I've got nothing from the Nova on any wav files that were created. In a realtime export I've got audio from the Nova on both the files produced from Stereo Out and from the track handling the Nova audio output (even though it was never captured as part of the project), and of course in either case the MIDI track that the Nova VST is on produces no sound at all.

Something slightly different is going on with your setup and I'm not sure if that's your version of Cubase or the way VC does things, but the fact that you are able to mix down a smaller clip of the track and not the whole thing leads me to believe doing a "safer" type of export (namely real-time) would address any DSP bottlenecking. Just small things like reverb tails, envelope release values and of course total keys played could make a big difference in whether a non-realtime export would work in one portion of the track versus the entire track.

Sorry if it's old news but I couldn't see that you had tried this or thought of it from your post.

[TweakHead]

I think the problem is really just note stealing from DSP overload. If you care about the quality of your music, then rendering down once the track is arranged is paramount for many reasons. recall is nice and neat, for when you're still in between, once you're sure about some sound and how it fits into the context, plus you have your song structure in place, then it's time to render.

Saving disk is an easy fix: get yourself an external one and back things up! You will regret it one day if you don't, as most people who've done this long enough (including myself) know, the hard way. The Virus is just like your computer, once your happy, rendering will free its power for other stuff, plus it will fix the material for good - 'cause there's always the chance you might introduce some random factor to some patches.

@MBTC you can only record the virus in real time, no offline rendering is possible; just like if you're using a Moog.

[TweakHead]

Actually it's quite simple: the number of voices used or complexity of the patches, or playing many of them simultaneously is something that changes with time in a normal track - so when the system gets busier, there's note steal and that's it.

Recording just one part at a time is the secure method here for making sure it's recorded with maximum quality and all the notes going.

[grs]

The op has to know you can't do a mixdown with the virus in any configuration. You need to record in "realtime" the whole mix thus not needing to record individual parts. There is no non realtime way to capture the virus sounds like other vstis.

[MBTC]

@MBTC you can only record the virus in real time, no offline rendering is possible; just like if you're using a Moog.

Makes sense to me, the Ultranova is the same way. The only thing I'm unsure about (and maybe it was addressed and I missed it) is that apparently something he's doing is behaving differently in just listening to the track play in Cubase versus voices being stolen in the export stage only. This indicates some output from the Virus is getting into the final file, so given what you've said I have to assume he's exporting real-time or there would be no audio at all if its like the Nova.

I also mentioned the significance of batch exporting multiple channels in Cubase, because the TI being multi-timbral, if someone had multiple patches sending to multiple audio tracks in Cubase, its important to be sure everything is making it into the final set of output files (multiple files rather than one in the case of hardcore Cubendoheads as they would probably work with them independently in Nuendo next, I'm guessing, for a truly pro result).

[grs]

..... I have to assume he's exporting real-time....
I dont believe the is such a thing as real-time exporting. (edit, ok cubase has it but still I wouldn't count on it.)

The Virus TI is not designed for any Render/Export kind of thing. You might be able to press the button on your DAW and get a file but the TI will just give you all sorts of drama. In a fresh install of your Virus TI a dialog appears when you try to off-line Render/Export/Mixdown etc. saying something like "hey buddy! cant do that. sorry".
In cubase the op needs a vst plugin that captures the stream to a file. or set an audio track to record the master buss and capture/record in real-time..

[MBTC]

I dont believe the is such a thing as real-time exporting. (edit, ok cubase has it but still I wouldn't count on it.)

The Virus TI is not designed for any Render/Export kind of thing. You might be able to press the button on your DAW and get a file but the TI will just give you all sorts of drama. In a fresh install of your Virus TI a dialog appears when you try to off-line Render/Export/Mixdown etc. saying something like "hey buddy! cant do that. sorry".
In cubase the op needs a vst plugin that captures the stream to a file. or set an audio track to record the master buss and capture/record in real-time..

Real-time exporting would put the Virus in the same situation that he is having positive results with now (well, the closest possible situation without the extra safety net of capturing the audio output before hand). In theory, it should not put any greater demand on the DSP of any single instrument than simply allowing the track to play in its entirety. There could still be some hiccups, but at least it might identify the culprit of the voice stealing. Between what Tweak said and what I think you're saying here, I'm getting the impression the Virus doesn't support even this? That would be too bad, I can do it with pretty much any other synth (the only two HW synths I have to date are the Ultranova and Leipzig but it would work with either or both at same time). It should be doable with the Virus by just unplugging the USB and using regular audio outs if not any other way.

[TweakHead]

Again, this is just pretty standard stuff. Virus, as far as recording goes behaves pretty much as any external instrument: you need to record it in real time. It has to play through, while recording.

The only problem I think the Op was having was that at some instances, there was to many things going on and that's originated the "note stealing", due to DSP overload - also a common thing in all DSP based hardware.

Virus allows you to change the priority per part, but ultimately when the time to mix down comes, it's best do record every patch on a multi on its own - specially when there's demanding patches involved or to many of them; so as to assure best performance/quality for each. It's really as simple as that.

As far as outputs go, you can go with either analogue outs connected to a sound card; or the audio stream using the usb connection - in which case he'll have the advantage of sample accurate timing; this last point is where the TI is similar to software instruments.

I'm able to record my Virus (C desktop) using the "freeze" function on Ableton, what it does is it records the parts and then the Virus stops receiving and sending audio, it's just an audio clip; so if that's what you mean with this Cubase function, it should - presumably - work properly.

Offline rendering, as you'd do with software is not supported. Not only with the Virus, but with any hardware synth that I know off. Unlike many people say, I feel the Virus DSP is strong enough for most uses, provided the user is familiar with what it can do and can provide a very smooth workflow - even with the C. The TI range has a major advantage though: you get delay and reverb per part on a multi, whilst before that you'd have those set as global fx - for the whole thing; even when using different outs for the parts on a multi. Thus, knowing this, rendering the parts on their own is a no brainer and I feel the same must still be true for the TI, except it makes life easier for arrangement and previewing the sounds while composing - plus some other tasty features that have been added meanwhile.

Multi implementation is actually very strong with the Virus, compared to many other instruments like the Blofeld for example - which is sort of incomplete btw. Number of voices available is also flying above most anything VA on the market, except maybe the Nord Modular (but that's discontinued anyway).

[MBTC]

Cubase has freeze also but it's a little different than audio export - maybe its the same way in Ableton? Freeze comes in handy more when I have a plugins running in the DAW that are CPU heavy (instruments or fx). The advantage is that if I have something like Dune2 playing a thick strings pad with a lot of notes and fx, that single instrument might take up 15-20% or more of CPU by itself. Freeze gives me a way to bring the CPU down to almost 0% usage with one click (actually a single button press the CH Cubase channel controller since it has its own button). It is essentially just creating invisible audio clip of the track behind the scenes, locking the midi data so it can't be edited, and managing it all invisibly, then giving an easy way to unfreeze individual tracks as needed for editing. It's one of the reasons I prefer soft synths from a workflow perspective, treating of midi data and audio as one, whereas I have to treat them as separate entities with my hardware. Freezing of course doesn't work with external instruments which would need to be recorded in real-time anyway. What I wasn't sure about originally is if the TI had the ability to take midi and render it digitally internally to feed back to the DAW - it makes sense to me that it would not, because while the technology is there to do it (the Virus is could be thought of as just a soft synth running on dedicated hardware), it would complicate the TI aspect and it seems they already have their hands full with that.

But I have to say I don't understand why if the OP is talking of simple note stealing from too much DSP demand, why he is getting such different results with simple playback. It should manifest there too unless there is something creating additional DSP demand during export.

[TweakHead]

The extra struggle while recording (or mixing down) should be the computer's cpu and not the Virus. The extra stress of the internal cpu could mean some added latency in the communication with the Virus, maybe - but all I'm doing here is guessing, really.

[MBTC]

Yeah that's what I mean, it seems like the computers' CPU would take the hit in a mix down, which doesn't explain voices from the Virus missing.

[TweakHead]

It could explain it. In theory there's a lot going on at once: during a real time exporting of a full track there's a huge cpu load; certainly hope this isn't what the OP is doing - 'cause that's just pushing your luck imo. The thing is: if using the usb for audio streaming and the usb connection is also taking care of the TI thing, where Midi in and out is an open route (if you will), then this gets busy. At the same time, if the cpu is struggling to keep the pace - and that's a common thing considering there might be a lot of latency inducing plug-ins (heavy duty dsp processing can be a cpu hog, as I'm sure you know better then I do); to my mind all of these things together can induce cpu spikes, in which case the added latency could originate this kind of thing.

In most common situations (where one isn't pushing his luck), missing voices is just "note steal" due to DSP overload on the synth. It's got it's own mixer, you can solo the parts, you can record one track at a time in solo, making sure both system resources and the Virus is free from unnecessary struggle and performs as expected. This is, or rather should be, common practice and best way to avoid issues - even timing issues with all types of hardware synths, where latency is always a factor to have in mind. Some ms is ok (sometimes a little bit off the grid doesn't hurt), you can compensate it with sample delay function (many times included on the channels themselves, otherwise as plug-ins, few times the host can compensate the latency and align with the midi).

What matters is to get a routine that works and stick to it and focus on the music, really.

[luigi71cx3]

Hello people, I am very happy that my question has ignited a discussion so exhaustive, and I wold to thank each one.

I understood that to make an audio track for each midi track is the best way, so I will do this way. Following an answer from the Access support:

Dear Luigi,
No, the Virus does not operate differently when creating the audio file via the "audio mixdown" function. I monitor the output of Cubase during a realtime export, so I hear if anything goes wrong already during this process. Are you doing the same?
Also does this already go wrong if you do a test and simply solo the TI tracks and then create an audio file out of this? (by "printing an audio file" I literally mean creating an audio file with the audio mixdown function or whatever process a music software provides for this).

As you can see doing an audio track each midi track is also suggested by them.
Probabily the Virus TI2 does not suffer of this issue (maybe it can play the tracks simultaneously during the mixdown without matters).

[TweakHead]

^TI2 is the same as TI, only with 25% more DSP power. Depending on the complexity of the patches, it will get to a point where there's note stealing.

This is something that has to be done with all hardware synths: you need to record them! The thing is: if you had, say, a polyphonic analogue, you would then be limited to the number of voices it's got (say 5 on a Prophet 5, for example). In this case, if you were to press 6 keys at once, it would only play 5, 'cause it can't possibly reproduce more. What happens with DSP is that every time you add a unison voice, for example, it's replicating everything that's going on with a patch (so do the math). So this has been (still is) the major advantage of virtual analogue technology over analogue - the limits are either set previously with some margin (so as to allow the engine not to overload even when pushed to extremes, which is kind of conservative); or - that I think is the case with the Virus - you can use the entirety of the DSP power available until you hit the ceiling. The later, of course, has to rely on the user's best judgement. If you were to use your Virus as a single instrument, you'd have a hard time hitting this ceiling (even with the Virus C). When you make the entire engine double itself a couple of times, each part is using its own set of calculations for all the modules: oscillators, filters, so forth and so on; same as when you duplicate some software instrument in software, your cpu takes on a bigger hit. It's the exact same thing. I think its meant to facilitate previewing and not having to make decisions right away, not a substitute for recording. If you were to rely on external instruments to save disk space, if by chance one of them got broken, you wouldn't have your track.

Just take a look at the modern branch of instruments with more demanding DSP going on, it takes a lot of CPU. Stuff like Dune 2 or Diva or Lush 101. Having a hardware unit that can touch those complex patches without compromising workflow (making the project less responsive, more stressed, cpu intensive...) is a great thing to have. Even current low priced mono analogues are good stuff to have. Euro rack is even nicer to have (coupled with ES4). =)

[MBTC]

As you can see doing an audio track each midi track is also suggested by them. Probabily the Virus TI2 does not suffer of this issue (maybe it can play the tracks simultaneously during the mixdown without matters).

Let us know how it goes. As I mentioned I ended up needing to do this with the Ultranova editor anyway, because otherwise the output just goes to Stereo Out and you have limited mixer options that way (including not being able to control audio/pan of that track independently aside from whatever features the VST plugin provides). I wasn't sure how much has changed between Cubase 6 & 7.5 because supposedly the mixer was a ground-up rewrite in 7.0. It was very different the last time I worked with the TI plug-in, because at that time, I only tried it in FLStudio which treats instruments and mixer tracks as two very separate things which do not necessarily have a 1:1 mapping (or in the case of external hardware, a 1:2 mapping with an instrument track for midi and mix track for audio). That felt a little cumbersome at first to me in Cubase but it really does give you a lot more flexibility once you get used to it.

[TweakHead]

Yeah, Cubase is top notch.