Cubase Audio Engine vs Logic. Cubase wins!

I’m not sure you can assume that the processors actually work the exact same way. The “audio engine” is not the same thing as plugin algorithms.

Either you set your test up to test summing and what not, and that means not using any processing, or you test processing itself. But that’s an essentially pointless and irrelevant test for the most part since people use third-party plugins all the time.

So, I think your test is pretty meaningless unless you know for a fact exactly how those processors work so that the comparison makes sense.

People have done this stuff like a million times already…

Yes. However I still continue to contribute! Everything sounds better than everything!

But my primary test was a raw export without any plugins, remember? Did I not do what you’re describing?

Mattias’s post poses good questions. Did we look at a summing tests or other tests? Something like this.

  1. Import identical files into each DAW.
  2. No plug-ins, systems lined up to the identical reference levels (as reported by each DAW’s metering, and, if possible, some high-end external reference metering, “god meter.”)
  3. Create New Output files with each DAW – DAW-L.wav, DAW-C.wav
  4. Cross import each file into the DAWs, observe any differences, perform null tests, etc.
  5. Import files into 3rd “reference” system – high-end audio editor of choice. Further tests.

Did it really “sound better,” or did I just turn up the volume? ( :smiling_imp: runs…)

Sorry, I actually missed that.

I’ll just have to say then that if you’re truly measuring the “audio engine” then I just don’t believe you. Two DAWs just exporting the same file with zero processing should sound exactly the same unless something is wrong. The only technical difference would possibly be dithering, and even if that is a difference it’ll be inaudible because it’ll exist below the noise floor of whatever room you’re listening in.

This has been done a million times. You’re either doing something wrong, or there’s something wrong with your setup, or you’re imagining things.

But you really have to eliminate things that aren’t part of the audio engine. You wrote: “the exact same MIDI and VSTs”. So, you didn’t use anything on the master, but you used midi connected to VSTi? How do you know that what you’re hearing is the difference of the “audio engine” and not the “midi engine” or the VSTi themselves?

That’s a possibility for sure. There was a difference in the ‘raw’ tracks, but I’m open to the idea of a lurking factor in settings. I just wish I knew what it could be.
It definitely felt related to the summation of all the audio. In Cubase, the highs and lows felt more upfront. In Logic, they were muddier.

The other, and very often most powerful, variable that needs to be taken into account is Cognitive/Confirmation Bias… Unless a listening test is conducted double blind then the results are essentially meaningless.

If someone could post a set of files exported from both DAWs under exactlythe same conditions with no identifying factors in each file then the listener’s own bias can be eliminated too.

It’s been probably over a decade since there was a big test on exactly summation, comparing DAWs of different makes and models as well as expensive analog and digital mixing consoles. The conclusions were in a nutshell;

  • people can’t tell the difference between DAWs
  • the technical difference that exists is so low it can’t be perceived
  • people still thought they heard a difference

So, if you’re saying that Cubase and Logic sums signals differently then you’re just wrong. They don’t. 2+2 always = 4. Logic doesn’t do that differently. Ever.

And again: The way to figure out if that’s the case is NOT to add things like effects and VSTi, you just import files, pan them hard left or hard right, then sum those signals together in each DAW, and then you compare the resulting files. Unless you’re changing the level at some point you won’t be able to tell the difference. Ever.

I agree. I once tested my real Neve 33609 stereo compressor with Universal Audio’s UAD-2 emulation of it. I recorded a drum kit and sent the drums through a bus with either the real thing or the software. I set both up so that they got similar amount of compression for the same input level.

Then I took the analog and digital versions, “A” and “D” and strung them out on a timeline in Pro Tools. But I did so in a ‘random’ order. So it was something like; ADAADADDDADDA.

I then had several people listen to that 8 or 12 bar loop (I forget how long it was) and write down what was what and how they felt about it. I said nothing about the order, and the name in the region wasn’t obvious. Nobody consistently even separated A from B, let alone identified which was analog and which was digital.

That’s what happens when people truly not only don’t know if they’re listening to A or B, but also don’t know if it’s the same from what they just heard.

As far as bias goes I trust my ears as far as I can throw them (not far). I trust double-blind or ABX testing though.

Actually, I’m glad I found this post because I also suspected Cubase had better overall quality…although I may be blowing smoke up my own ass.

I have sent rendered orchestral tracks (MIDI, no actual recorded audio tracks) to an editor, and there were very minute artifacts in the Logic stems; not in the Cubase stems. I did not apply any dithering, but I also did not change the sampling rates or bit depth (both sequenced and rendered at 24/48).

I have no rational reason to believe Cubase is superior to Logic Pro i this regard, but if further tests can prove this than I know I am not going crazy. And if it’s indeed true, then it’s time to stick to Cubase 100%.

Excellent testing set-up. It reminded me of one of the AES videos from several years ago, there’s discussion of a false comparison where a large, wall type switch with a nice positive click was labeled A-B claiming to switching between some “high-end analogue system” and some “high-end digital system.” Both were playing back from, if I recall, a good, but standard CD player of the time over some normal professional-grade headphones. I’d have to go back to get the exact details, perhaps you know the clip I’m referring to? So, with both sides being fed the same source, some listeners reported hearing much more “warmth and whatever” from the “high-end analogue.” Others, just looked at the demonstrator and said, “nice try.”

Separating what we want to hear, or even think we should be hearing, from what is actually there is the critical thing, and is why good testing such as you described is important for solid, reliable, repeatable engineering. As discussed before, someone should release a calibration and testing package for all the DAWs. Ideally, the package would also do some DAW computer system analysis and perhaps suggest or implement actual system optimization. Take care for now.

I always find topics like this hilarious… Cubase is a superior program to Logic for sure, but if you want to get technical, Logic now has 64 bit float mixing engine.

regardless, when it was 32 bit, i nulled it to cubase to infinity… at -170 db on span there was nothing. Zero.

You have to understand, a phase null proves there is no difference…

Even the best ears can hear a “difference” at various points of the day. Even a slight change in pan could cause it… just a setting that you think is dead on the same as cubase’s and vice versa but isn’t.

Ultimately what sounds better, is which program facilitates your workflow to allow you to get the best result possible. It’s entirely believable that someone could get a better sounding mix in cubase over logic and vice versa. But there is nothing “faulty” about logic’s audio engine maths.

I tested ableton vs cubase vs logic vs pro tools native and they all nulled completely. I did a very scientific and simple test. I put 4 audio files in each, on 4 audio tracks. Did not touch pans and left volumes at zero. No plugins. Exported. Put the four exported results, one from each daw, in a new project and did phase cancellation tests. All were identical, and i mean identical. This is maths… we are dealing with digital audio, not analog. It is ALL NUMBERS. If a phase cancellation test cancels it out to infinity, there is no difference to hear as it’s simply not there.

The reason i don’t like these topics is because it misleads people especially newcomers… trust me, the workflow is what enables one to get a better result vs one daw over the other, AND of course, the subtle change in our ears even within a few seconds of direct comparison testing. I have been doing this for 26 years now, and the other day i was hearing the differences between the plugin on and off, i was hearing a volume difference every single time i pressed bypass, But what i hadn’t realised was that the plugin was disabled the entire time with it’s internal bypass… it was not activated and doing absolutely zero! This happens to everyone and is normal. Hope this helps explain it somewhat.

I can’t help feeling sometimes that very often some people can get sucked down the rabbit hole rather than just getting on with making music, perhaps it’s some kind of subconscious avoidance mechanism at work… i can think of one or two people i know in RL that definitely seem to do just that.

Welcome to the forums. That test would make for an excellent youtube video or an article. I’m very glad to you did it and got the results you did. Thanks for posting about it.

Maybe the next big plug-in will be named PlaceBO and everyone will be “raving” about it. :mrgreen:

These forums have helped me avoid many pitfalls and I’ve collected tons of useful information from the people posting here. Posts and threads like this thread and your post, do help newer users; and, what’s more, are just interesting to read.

I think you’ll find that a lot of us who read up on this years ago just make the single odd comment and then move on, because it’s already been tested… years ago… with documentation etc. Not to be a downer, I do appreciate people taking the time to test this for themselves, it’s just that it’s been done and discussed a million times already.

Yep, but make the name an anagram so people don’t dismiss it!

“Paleo-BC”, for “clean, natural processing”… your turn…

thanks for the welcome… you really made me laugh with the placebo plugin LOL! Cheers

ps been here a while, at gearslutz much more active under same name, these days i tend to only really post when i have a problem, i am busy making music 99% of the day, every day!

spot on! wish we could up vote posts here…

I had to know once and for all which is why I did my own test… i did a much simpler test in that i just tested that the audio that went in was the same that went out… and it was, in all of them.

I doubt i’d bother doing it again but if people really want it, I will do it and prove it when i have some time :slight_smile:

Even little things like time stretch perhaps being enabled in default in one daw and it wasn’t in the other, can change results… this is why a very simple careful test is the way…
For example, people argued for years that live sounded bad… but live enables stretch on everything by default. Not only that, until recently it’s automation was completely out of time due to a faulty plug in delay compensation engine… this has been rectified somewhat, but all little mixing and timing errors could trick the ear into thinking it sounded bad.

The real kicker for me was what happened the other day when i was bypassing and activating an actually internally bypassed plugin, and i was hearing a volume difference LOL. This is why i say, the slightest change, the way the head is tilted even say one inch differently compared to the time one is testing a different daw… means our ears respond differently.


It’s very disappointing that Steinberg is representing 64-bit processing the way it does. If it has stated clearly and technically accurately what the benefits were I’d have thought it might have been a good idea (assuming they were technically correct), but the video to me looks like a bunch of garbage. Exactly the type of language we could do without.

I am looking forward to being able to at some point A/B compare mixes in 32-bit float with 64-bit float, doing a null test, and see just to what degree there’s a difference attributable to the actual summing engine. I’m guessing it’ll be in the LSB or two, and not audible, and thus a pointless exercise in spending development money (as opposed to fixing Nuendo v7/8 for example).