KVR MIX CHALLENGE - MC03 August 2014 - Voting period has ended, Winners announced (pg 17)

[Compyfox]

I think I have to clarify something here, since this question is popping up for quite a while now:
Dynamic Range is not (in short, mathematical term: !=) Loudness Range

The loudness range is (simply speaking) a measured "range" between the average loudness and the maximum loudness ((ITU-R BS.1770-x specs, SLk ballistics, gated).

The Dynamic Range in the statistics sheet is the range (mathematical value) between unweighted RMS max (300ms) and digital true peak max. Which is actually closer to a realtime readout instead of offline measurements.

The original DR offline meter is actually a bit more complicated - but in short, it takes the "average" signal strength, or better said takes an "average" of 20% of the over 10k measurement RMS avg points (max average values btw!) and subtracts the digital max peak to declare the DR value. I had to refresh my memory on that with the official tech paper, and now I know why it never succeeded - it was just too complicated to understand, it was too program dependent, and you could never reproduce your measurements (since the 20% were declared randomly).

Anyway - I just revisited the Statistics sheet and went for "RMS avg - dBTP = DR value", and suddenly the readouts look way different.
Here is the original Statistics Sheet (RMS max to dBTP max = DR value)
And this is the overhauled Statistics Sheet (RMS avg to dBTP max = DR value)

Night and day - but still revealing (I'll get to that in a minute).



So what does that all mean?

Most of you participants have DR values between 8 and 11 (if we go by RMS max to Digital Peak Max from the original statistics sheet). DR values are very dependent on how dense a track was mixed - or in other words: program dependent. Chances are you can reach a very high DR if you have strong transients, while your "average" signal is just a perceived sound sausage. Therefore it's not an objective measurement (that meter never was, as too many parameters were needed to be put into consideration, and each genre needed own "acceptable parameters"). The value I am offering (RMS max - dBTP max = DR value), can be seen as using the old (realtime) TT DR Meter from either the Pleasurize Music Foundation or the re-issued plugin version by Brainworx themselves, not the offline meter (which is more like RMS avg - dBTP max = DR value).

It's a bonus piece of information for those still transitioning - or those that want to compare metering values. By today's standards, DR meters are non-precise, very random and pretty much outdated.



Now - what is really important?

In our example, the mix challenge (emphasis: mixing - not mastering!), actually a good and healthy peak headroom without(!) using a limiter on the master bus. Therefore you're automatically also using a lower average signal strength - or even a suitable reference level. Both the "Loudness Range" (LRA) and the "Dynamic Range" (DR) show you how dense your mix is (again, simple speaking). There are no ideal values, but give you indications when things are just too much.

For example: a DR of 5 tells me, that the mix was either already pre-mastered, or heavy compression/limiting was used/applied (remember: range between average signal strength max in this case, to maximum digital peak). On the other hand - a DR value beyond 14 can mean "wow - this track must be uber-'dynamic' - absolutely not squashed at all", but this can be terribly misleading (see "sound saussage, while having strong transients" example above).

A LRA of 3LU or 5LU might show me, that the track doesn't really "breathe", while a LRA of +15LRA shows me, that it might fluctuate too much. Keep in mind, we're constantly talking about measuring music, not a broadcast stream with voice overs, FX and music. So these values are also program dependent, but... they still might tell you "wait a minute - something is wrong".


While these values are definitely interesting for a mastering engineer, musicians or plain "mix engineers" usually don't need to bother with these readouts on the meter. Their focus should be: is there a healthy headroom? Is my average signal strength within boundaries?"

These statistics are for a continuing (process of) learning.



Now on to the question by

I just sent Uncle E my votes. There were some really good mixes in there. I was also curious why you felt my track had been premastered as it had one of higher DR numbers?

I just listened to the track again, and took a dive at your statistics.
I wrote in the statistics, that your track is either pre-mastered or used a high reference level. And you wonder why.


Taking a closer look at the values of the original statistics sheet alone (-0,19dBTP, -11,12dB RMS max, +5,9LU LRA, 10,9 DR), this gives me an indication that you worked at really hot levels, have a dense mix, while you might have used heavily limiting. A short listen to the track confirms it. Various instruments are pumping, so heavy compression was involved (the vocals are the most prominent example).

Why does it still have a DR of 10,9? Because the RMS max was -11,12dB and the max digital peak was -0,19dBTP. This gives the mathematical readout. If we go by the statistics revision (again: RMS avg - dBTP max peak = DR value), it's even 14,2 DR. Does that mean this track is "uber-dynamic"? Well yeah - at least according to this value readout and looking at the waveform. But if we look at the Loudness Range (LRA) and having a listen to the mixdown - maybe a bit too much compression was used.





Long story short:
As you can see - analyzing a track at either scenario (post mixing/pre mastering or post mastering i.e. with a regular release) is definitely not simple. Especially not with one value alone.

Meters can be stupid sometimes (for example: unweighted ones respond stronger to bass, the original DR meter even has in the manual, that the measurements can't be reproduced!). You have to put the readouts into perspective and work with that information.

Does that automatically make you a better mixer if you improve on that?
It's really hard to tell - but at least you're thinking about overusing compression and limiting. Read: your awareness is definitely brightened if you know what's going on, and what your metering tool shows you on the GUI.




EDIT:
Updated the "voting" post on page 10 with the re-visioned statistics.

[psycho43142]

Actually no hot reference levels. After I bounced the track I normalized and burned it to CD to listen to on different systems to see what it sounded like on them. If you take your track and normalize it to the same as mine it is very similar in loudness. by all aspects I looked at. This whole thing on what was or wasn't done is pointless if you are doing the leveling like you did so all tracks are close to the same level. Instead of worrying about things like this we should be looking at the more important things like is the mix muddy? Does it have to much reverb? does the bass have to much midrange? Do the vocals sound a little nasally? How can we fix these things? As far as what the GUI and meters show, don't trust them, use your ears.

[Compyfox]

Sorry - I disagree here.

"Peak normalization" never works out for an objective comparison of tracks.

Maybe you like a loud mix - but this is why we do the Loudness Normalization process - to offer a balanced listening experience and objective judgement. And here, the provided statistics does help to see where you could improve on. On top of the general questions you should ask yourself during/after mixing.

So by understanding your comment, you didn't even need to peak normalize your track - but you did it anyway due to reasons I can't follow. You have a volume knob, so why "normalize"?



BTW:
I also never trust my ears alone. I always have some sort of analysis tool running in parallel. And using Digital Peak and VU meters are a second nature for me.

[psycho43142]

So the volume in my car isn't at max to have a decent level. It has no other reason than for that purpose. Lets see if I get this right, to is bad to mix a dance track that has bold dynamics and pumps. You're right nobody does that. This track was to be aimed at being a club track, so in my mind it should pump. Again maybe instead of talking about how loud a track seems maybe we should talk about why some tracks are very muddy in the mid bass area and how to fix that which is something I felt most of the submitted tracks suffered from. How to get the kick and the bass to sit together in the mix. I know the mastering engineer can fix these things so why bother.

[blind]

Honestly mixing is an objective thing, mastering is more of a hard-rule type thing, so I don't think getting into the politics of it is gonna help at this point. Weren't we doing this for fun?

I agree about this being a dance track and it needing to pump though, hence why I added some sidechain pumping to the track. I think compy even suggested doing that and "being creative" with the mix. Just cuz the pumping continues from the kick/key track without the kick being there doesn't mean I am breaking a rule necessarily. Sound design is a hazy line in mixing as well. A lot of dance music is mixed as it's made for this very reason since mixing is so integral in it's sounds and styles.

also, the loudness thing, isn't TOO big of a deal, though we should stick to the rule of no louder than -3 peak db. RMS db is a different story because we are all using saturation tools and maximizing busses in the track, so I don't think that should be hard-fast rules as much. We can all use our ears and volume knobs

[Doc Jon]

Thanks for the explanation Compyfox!

[Compyfox]

So the volume in my car isn't at max to have a decent level. It has no other reason than for that purpose.

Actually - the DR and LRA "adapts" (read: shifts) if you peak normalize a track. So if the original RMS avg. value was originally around -18dB and not -14dB as it was with your entry, the LRA and DR values would still be the same (see math formula) - only the dBTP and RMS avg values would be lower (i.e. -18dB RMS avg - -3,6dBTP).

So as you can see - the Peak Normalization didn't gain you anything other than not reaching for the volume knob in your car. And if you would have "Pre-mastered" the track, the LRA and DR values would have been lower as well.

Lets see if I get this right, to is bad to mix a dance track that has bold dynamics and pumps. You're right nobody does that. This track was to be aimed at being a club track, so in my mind it should pump.

If is a difference if a vocal track sounds like being "sucked down" by a vacuum cleaner every time it is setting in, or a pad having a grooving element once a kick drum is setting in. This is what I mean with "pumping" sound. This is usually an indicator that a compressor was overused - or wrongly used.

Again maybe instead of talking about how loud a track seems maybe we should talk about why some tracks are very muddy in the mid bass area and how to fix that which is something I felt most of the submitted tracks suffered from. How to get the kick and the bass to sit together in the mix.

We can of course do that. Then again, everyone has a different opinion on mixing in that section. As can be clearly heard with the various entries. Some might even struggle if they used headphones or have a less than ideal speaker setup, others might have tired ears at that moment (on my end for example - I know my flaws if I pull everything at "the last minute", especially if work, family and "questions" on KVR pile up - a revision is already pretty much done at this point).

This is the purpose of this challenge - the focus on mixing (and mixing only!), learning from others, improve on it. Not "peak normalize" and maybe even "pre-master".

I know the mastering engineer can fix these things so why bother.

No, the Mastering Engineer can not fix these things. The Mastering Engineer can only build upon what the Mix Engineer provides. If the mix balance of the track is off whack, he can't fix it. If the reverb is too strong, he can't fix it. If the tracks are not aligned correctly - again, he can't fix it. And if the track was "premastered" already, this can't be reversed to a satisfying result.

So you need to bother.

An overall "dull mix" that has no flaws otherwise is easier to handle than something that has more severe issues but otherwise "clean sounding".

also, the loudness thing, isn't TOO big of a deal, though we should stick to the rule of no louder than -3 peak db. RMS db is a different story because we are all using saturation tools and maximizing busses in the track, so I don't think that should be hard-fast rules as much. We can all use our ears and volume knobs

At this point, it isn't hard to use a realtime RMS or VU meter on top of a digital meter. This seems to be the biggest learning factor for this challenge - how to use both utilities to their efficiency.

Else yes, -3dB digital peak without using a compressor/clipper/limiter array on the master bus to keep the levels in check, should more than suffice for those struggling with metering tools (though I provided plenty of posts by now, that easily describes what to do, and how to make the best out of it).

The statistics sheet on the other hand, is for those that want to dive a bit deeper into the topic.

[photonic]

Voted !

BTW: I wonder, if we will ever talk or discuss so much beside levels, dynamic and statistic values? A mix is more than just setting levels. Don't take we wrong, I appreciate all the dicussion here in the forum and I learned a lot. But there are other things like frequency balance, use of FX (more or less reverb...) or mixing good layers of depth and so on.

[blind]

amen to that, although those things are key to a good mix, frequency curve (also appropriate to genre) careful application of fx and the rest are what catch my ears more than anything

[Compyfox]

I think I repeated myself quite a lot now as well.
So I'd also like to talk about something else, please.

[Uncle E]

This whole thing on what was or wasn't done is pointless if you are doing the leveling like you did so all tracks are close to the same level.

I agree with this. We made up the mastering rules before Compyfox had figured out how to batch normalize by rms. Now that that's done, I think we can stop worrying about that. The fact is that voters aren't responding well to the smashed mixes and I think the pre-mastering is probably hurting most people's mixes.

Also, unless you guys find it very handy, I propose we eliminate the analysis file. Compyfox has to do that manually and I personally don't find it to be a good use of time.

[Uncle E]

We can all use our ears and volume knobs

Well, one of the goals was for people to get familiar with mixing to standardized level, which does go beyond mixing with your ears. You don't have to do this, obviously, since we'll be normalizing it for you, but it's a good thing to learn if choose to take it on.

[Compyfox]

Also, unless you guys find it very handy, I propose we eliminate the analysis file. Compyfox has to do that manually and I personally don't find it to be a good use of time.

Actually, the main manual process is copying the data over from Wavelab's Analysis window into an Open Office Calc sheet.

MC02's statistics was purely done by hand. With 12 tracks this took me about half an hour. I expected more work than that with MC03. But now I have a usable template (created from scratch, which all on all took me about 2-3 hours to create and sort the data) - so things can be done within a short time span (30-45 minutes - which can be handled during uploading the RAW bundle). Unless we get over 50entries, then this will take a while.

But I think the information is worth it - it's adding to the learning process. And actually, from the last two challenges alone, people start to rethink their concept of metering, mixing and whether or not using limiters/clippers on their master bus.


I see that as a good thing.
The focus goes back to plain mixing again.

[psycho43142]

Listen to this song pump!

https://www.youtube.com/watch?v=FGBhQbmPwH8

Much in the same way mine does so is this one wrong also? Maybe it is all just a matter of taste.

[Uncle E]

Much in the same way mine does so is this one wrong also? Maybe it is all just a matter of taste.

Maybe I've missed something in this discussion but there's nothing wrong with what you submitted. The analysis is just for your own knowledge, to empower you to understand how your mixes appears when technically analyzed. There really shouldn't be any finger pointing or wrong making attached to it.

FWIW, the mix I submitted (which we didn't end up adding to the playlist because we didn't want to confuse things) pumps MUCH more than yours does.