Now it gets interesting!
I just did my test: recorded preview audio and created mixdown of the same project. Alignment of the two tracks was done by Cubase, as both recordings started at 0:00:00. I just snapped both tracks to the same 0 point (Changing this point by hand only increases audible audio)
I am talking about a normal project from my studio with guitar, vocals, vst drums and keys. Most levels around -3dB. I do have A LOT of remaining audio. Actually; the whole song (with a phaser/chorus/fake stereo-like effect on it) is audible! Mostly mids and highs clearly but lows until about 100 Hz. (Output VU meter tells me a result around @ about -12dB!)
When zooming in full, the two waveforms are exactly aligned (even until the bit accurate), and resemble each other completely (to the naked eye).
No settings were changed, nothing was altered… I used audio>process>phase reverse for the mixdown version.
I can hardly believe the whole song could be audible, so maybe I’m doing something wrong here? I didn’t change anything in the settings between the preview record and the mixdown calculation, so basicly they should be 100% the same? I think I might have a very interesting result here!
Specs: Cubase 8.5 on i7 4970k, RME Fireface800, Waves platinum (L2 and LinEq used in this test), no automation used.
I can do this same test aswel @ our studio system, that’s Nuendo, and a more accurate external clock there (both Rosendahl nanosync and Grimm CC-1 clocks are selectable) That could make some difference maybe? In fact the same brand RME (madi) interface is used there, however Millenia HV3D AD via ADAT… Also the PC is a powermac there. Although better equipment, I hardly think that that hardware could make so much difference that the outcome of the same test will be Null there?
I would continue by doing what I mentioned earlier, and start with the most simple scenario possible, and then make things increasingly complex. My hunch is that you would find where the problem exists faster that way.
In other words right now we only have two things to rely on: My set of tests, which imply that volume automation accuracy is different under different circumstances (playback vs faster-than-realtime), your test, which implies that something is different in “a normal project” which surely is much more complicated than my test. And because we only have two scenarios tested, and they vary greatly, neither is “confirmed” by third parties. Know what I mean?
I think step-by-step instructions are needed to get other people to reproduce the discrepancy.
Just a thought here… and please forgive if it’s my ignorance but could this be down to a routing issue on the OP’s part as the preview audio differs from the mixdown?
I think that’s possible, yes. I know a lot of users get confused if using Control Room, because they end up “double-routing” the main mix to the speakers. So it’s possible something else is up. I would certainly suspect routing if the difference that is heard is wide-band, consistent, and pretty much sounds like the ‘originals’.
So again this is why a test should be started not on a full mix but on a subset of a mix. It’s far easier to make sure all the routing (and other parameters) is correct that way.
Just for interests sake, I just updated from Windows 7x to 10 and upon running the syncrosoft licenser got the same errors as you complete with red soft license text. I decided to check on the syncrosoft site for the newest version, DLed it and installed. Everything now back to normal.
Compared 6.5 and 8.5. They are different indeed. Better ? Depends. What’s mixed in 6.5 sounds different in 8.5 and viceversa. I think I take different decissions based on what I’m hearing real time. Just stay with the project where it was mixed.
Exactly. I don’t know why, but the PREVIEW audio quality in C8.5 is sounding boxed, metallic and has big transient problems while the same project in C6.5 has none such problems, but it does. In short, it sounds crappy. I believe (compressor) plugins (waves, NI,AIR) suffer the most, they seem to have trouble with handling the input dynamics as they could handle them easily in 6.5.
As I stated before, my null test of recording PREVIEW audio (of a very common project) vs. mixdown audio in C8.5 is irrefutable evidence of a big mistake made in the algorithms of C8.5., or…:
Note this however: I never destinguished the C8.5 installation from the Win10 installation. They were both installed at the same time. (and uninstalled quite soon after ) So: Win10 might aswel be creating the problem…! I use RME Fireface hardware and UAD-1 DSP. Octacore CPU and 16 GB RAM.
So in short: For me it is WIN7 and C6.5 from now on. All updates disabled and big smile enabled.
No, I mean preview audio when working on a project (real time). This includes the use of VST plugins of course.
In the mean time I cannot reproduce the problem anymore and believe it has more to do with the OS than Cubase! I stepped back to WIN7 64 bit and killed all update possibilities of the whole system. I believe the sound problem had more to do with WIN10 than Cubase (strangely enough!) VST plugins perform strangely in W10 and have trouble handling transients (clipping and overall harshness of the preview sound). The mixdown audio would then sound way better but still the clipping and dither-like sound could occur when crancking up the overall gain. With WIN7 C6.5 and 8.5 (without the use of the built in EQ’s) now sound the same to me (I still prefer 6.5 but that has something to do with the changes in interface which I dislike.)
I gave up on this in the end as I could not get it to null. I have since seen it on another guy’s system, and we tested with no plug-ins on. But oddly it nulled on another system we tried it on. No idea why.
) So: Win10 might aswel be creating the problem…! I use RME Fireface hardware and UAD-1 DSP. Octacore CPU and 16 GB RAM.