MB2 Petition

[madmaxmiller]

- i did not expect the files to null, but was surprised the difference was not subtle

Haha I did exactly the same 2 hours ago. I chose a simple multitrack with 13 tracks, only adjusted volumes and pan (as Jouvert suggested), switched the master limiter off (despite i think it doesn't have an influence at -9db peak), turned the master tape sat completely down, meticulously repeated the settings in the other version, imported the exports into Audacity, flipped phase on one -- and was surprised. I was even more surprised about the bass drum and lead vox content... which are dead center.
However, the single tracks themselves didn't sound much different, but I have to double check on my studio monitors - I had only ATH m50 cans here.
However, I first thought I made a mistake in Audacity, so I did A-B, exported, then B-A, exported and then I re-imported (A-B) and (B-A) - and they nulled out as expected, so no mistake here...
Anyway, it seems there are significant spectral differences between the versions. However, I do not prefer a particular version (yet?) so the advantages of v3 are the deciding factor for me. Yes, I confess I liked the new surface ootb and I like the sound.
Will put up the wavs tomorrow and will still wonder if someone safely identifies which is which

MMM

[the C.L.A.]

I just did another test - also with a full mix imported and then simply routed via master with limiters turned off and exported as 24 bit wav.

The exported files null down completely to dither level - no difference in the rendered audio at all!


So here are the possible problems you might been facing:

- you didn't turn off the limiters, wich are known to be different

- some other setting might differ for some reason (maybe you created the session from a template?) Note that microscopic differences will spoil a null test. For example the numerical displays when turning any of the knobs are not accurate enough.

- you didn't align the exported files properly (for the null test) - they have been off by one sample for me

- your test results may have been affected by a bug that x42 just fixed last night (see details below)

- the DSP on your platform might be broken. This would have to be veryfied and fixed then.

Did I forget something?

In any case it would be nice if you could upload the files if they really differ so it would be possible to further isolate the cause of the difference.


Now regarding the bug: When creating a new session parts of the DSP might not have been initialized with the correct sample rate so that some filters were off. According to Robin (x42) this most likely would only affect you if you used a sample rate other than 44.1 kHz. A workaround is (at least for me was) to just quit and reload the newly created session. [Edit] See: https://github.com/Ardour/ardour/commit/...9c99c6d1f6

So next interim should contain that fix.

[Jouvert]

I am glad that you guys are doing some real scientific tests (i am not that inclined). So am i correct to say that even though both versions null each other (down to whatever db level), there is still significant spectral differences in the rendered audio files?

jouvert

[the C.L.A.]

So am i correct to say that even though both versions null each other (down to whatever db level), there is still significant spectral differences in the rendered audio files?

If files null completely then there is no difference at all. In the case of Mixbus they can't null completely due to dither - which is random noise. But if dither noise (wich should be quit a bit below -100 dBFS) is all that is left after a null test it's safe to say that there is no difference in the actual rendered audio. If you still "hear" some difference it's entirely placebo.

Also note that if you add any effects that add some randomness or modulation to the audio (like chorus, phaser, modulated delays and most reverbs), these won't null either. If you want to compare and especially null sessions wich contain such effects, you would have to bounce those effects to an audio track first and disable the actual plugins, so you can use the exact same FX track on every export.

[Jouvert]

But how could one put anything that does null - (even if not fully due to dither noise) as just merely placebo, because i don't know if you all noticed, but there is a change in the sound stage in the newest interim release.
As you all know some sciences are not exact and subject change, so who is to say that because the audio files null that there isn't something still happening, because mixes in v2, for some reason are always favoured over v3 (at least in my tests)

Jouvert

[jamip]

But how could one put anything that does null - (even if not fully due to dither noise) as just merely placebo, because i don't know if you all noticed, but there is a change in the sound stage in the newest interim release.
As you all know some sciences are not exact and subject change, so who is to say that because the audio files null that there isn't something still happening, because mixes in v2, for some reason are always favoured over v3 (at least in my tests)
Jouvert

Null tests are simple math. If you have a number/value and subtract an equal number/value from it, it will equal zero. If the value subtracted doesn't give you zero, then you can conclude that the subtracted value isn't equal to the value it was subtracted from. That's all null tests are.

Digital audio is made up of values too, and if you subtract identical from identical, it will null. This method can prove if two audio files are identical. It's a proven and objective method.

You can't subtract 20 from 20 and argue that just because the result was zero, that it doesn't mean the numbers weren't identical. You don't get zero or a null result unless the values, or in this case the audio, were the same.

[Jouvert]

Thanks jamip, that makes sense.

jouvert

[bobos]

I think Jouvert started the Petition thread to convince Harrison to finish MB2 that is a great console emulation. 99% ready.
Surely we do not want in this topic to discuss how to pay anyone to develop it for us and whether it is hacking or legal.

Harrison I think is sufficiently skilled to do the task if he thinks so. MB2 works for me on Win7/64 great with one core.
If it stays one core who cares, it works but has something to be corrected and finished.
I this petition I would rather call Harrison attention to a few small things like this zoom issue in 2.5

https://www.youtube.com/watch?v=gskMGFhu...e=youtu.be

to fix. (I found it fixed in V3 so you already know how to)

There are some more, that do not need any third party expert to solve, even Ben does not suggest it to do according to my understanding.

But this rising the possibility of hiring someone to do multycore and then chewing on it as real "save our life" I really cannot understand

Harrison has a lot hardware consoles and not only one and the one's "upgrades". All for different purposes.

It should be understood that MB2 has a certain purpose, grat in it, and V3 has another purpose, much wider. They are different!

So why not have two great and finished Harrison console emulations? MB2,5 would need some months V3 some years to finish. Why I am sure of it? Because I have been using MB2 for 3 years, much simpler than V3 and still not finished!

I spent mixing only in MB2 more than 4000 hours, (+2500 in others) yes you see it well! So those striving for the latest tooshpaste and the newest detergent... please do not tell me that MB 2 is over and "we" just want the newest V3 version.
And during those long hours there were lots of bugs but was no blog, only email, and Ben was the greatest support I ever met.

It is a shame that only some of us (2 here) stand beside MB2 and ask Harrison not to throw it in the trash.
MB2.5 is the FIRST great sounding console emulation ever created! Be proud of it and finish it, please.

Tassy

(3 here) ...

[the C.L.A.]

But how could one put anything that does null - (even if not fully due to dither noise) as just merely placebo, because i don't know if you all noticed, but there is a change in the sound stage in the newest interim release.
As you all know some sciences are not exact and subject change, so who is to say that because the audio files null that there isn't something still happening, because mixes in v2, for some reason are always favoured over v3 (at least in my tests)

As Jamip already pointed out the null test is hard math - comparing each single sample of the two audio files. That's also why the files have to be aligned sample accurately.

In addition to that I just wanted to add that...

As written in an earlier post there are some differences in the signal flow you might want to take into account, as they are likely to cause some difference in the outcome of a mix. This also might affect opening an MB2.5 session in MB3, depending on where you placed some plugins.

Then as noted, in my first test with a stereo mix I relied entirely on the numbers displayed below for adjusting the tape saturation - I did set both to "+0.0". But as the knobs have higher resolution than those numbers it still resulted in a meaningful difference between those audio files as they nulled only down to about -30 dBFS, and I could clearly tell both files apart in a blind test, repeated multiple times - percieving one as having more depth and width. So a change in value lower than 0.1 on the tape saturation can result in an audible difference.

You're making lots of small decisions like this when mixing, and there is also the possiblility that the GUI, the default settings and parameter ranges lead you to making different decisions in one direction or another. It may also be true that you just need a little time to adjust to some of the changes to get the same or even better mixes out of MB3. Also take a look at the signal flow diagrams in the manuals, spend some thoughts on how they differ, what effect they might have and how you could use those to your advantage (or have done so inadvertently in MB2.5) in either version.

I tend to think my mixes have gotten better - partly due to adjusting and learning and partly due to the new possiblities in signalflow and workflow in MB3. So in my opinion it is worth taking some time to really get to know and adjust to MB3 and maybe then go back after a month or two to check if you still think MB 2.5 was superior.

[Tassy]

"So in my opinion it is worth taking some time to really get to know and adjust to MB3 and maybe then go back after a month or two to check if you still think MB 2.5 was superior."
Checked.
I would do it with pleasure for more than a month but after one minute I get this when quitting:
http://screencast.com/t/nU0YEXNyg

So for me MB2 as something to work with is superior, for the time being.
Tassy