New Mastering Challenge - May 2013

[ObsoleteAcc99]

[DELETED]

[t3toooo]

Could it be that some of you use the individual tracks (STEMs) rather than the "unprocessed" mix?


Oh and... could we stop saying "steps"? Unless we mean "steps in the chain".

Editing steps would mean for example: 1) upsampling, 2) adding FX, 3) downsampling, 4) encoding into WAV/FLAC, 5) uploading.

i would let anyone choose if he want to use the tracks (which is better anyway) or the unprocessed mix.

[Compyfox]

Uhm, sorry t3toooo. But in order to get either "similar" or "similar fair" results for that matter, we should decide on one source. If one uses the stems and another one just the stereo mixdown, it's no wonder that mixes are bound to sound different. To some extend, even drastically.

And this was not how this particular challenge was in mind. At least IMO.



So... my proposal would be:

Stick to the UNPROCESSED version (in the SoundCloud downloader) and then to the 3-5 plugin limit. If you want to, do the STEM version as "bonus" (with as less plugins as possible as well) - but the important thing is the remastered stereo mixdown of the unprocessed source.

As loudness target, we could go for K-12 AZ+2 as absolute maximum for fortissimo passages (chorus), and K-12 on average. That is -10dB RMS maximum.





Technical Setup (for metering):

To those that use configurable VU's, the reference level would be -12dBFS and rise/fall 600ms, no weighting filter. Klanghelm VUMT, PSP Vintage Meter and zplane PPMulatorXL can do this. Or use a K-System meter of your choice (RME DigiCheck has one built in).

If you're using EBU Loudness with the K-System v2 presets, then try to use K-14v2, and try to have 0LU SLk on average and +3LU SLk on fortissimo passages as absolute maximum. Depending on the density of the source, a K-System v1 meter in parallel should show K-12 AZ+0, with certain rouge signal peaks towards AZ+4.

ORBAN's Free Loudness Meter can also utilize the K-System v2 specs. Simply setup the reference level to -14LKFS, then have an eye on the Short Term & Integrated Meter, shoot for 0LKFS on average.

[ObsoleteAcc99]

[DELETED]

[APZX]

Compyfox,

Just for the sake of knowing I set Klanghelm VUMT to a Rise/Fall of 600ms and -12dBFS just to see how it stacked up in regards to the K-System (not technically a K-meter, but close enough). Turns out I'm not too bad at guessing this stuff by ear in the first place (one overshoot when the chorus kicks, only just goes above +1). Personally, I did this master with VUMT at normal settings just raised the reference level to -12dBFS.

I don't know if I'm weird or doing things against what most people do. But I personally mix to VU (-18dBFS) rather than peak meters. I only use peak metering to make sure I don't go above 0dBFS. Other than that I just focus on trying to make it sound as good as I can at my current skill. I do agree with Bob Katz though that metering ought to be based more on average or density rather than peak.

[t3toooo]

Uhm, sorry t3toooo. But in order to get either "similar" or "similar fair" results for that matter, we should decide on one source. If one uses the stems and another one just the stereo mixdown, it's no wonder that mixes are bound to sound different. To some extend, even drastically.

And this was not how this particular challenge was in mind. At least IMO.

didn't listen to the files nor i don't know if i want to participate,single tracks are much better for mastering.

anyway,you can't control it
somebody will choose what he want,that's why i would say let it open and let them choose,everybody can describe what he did.

otherwise i wouldn't bury the challenge in rules,a good sound sounds good.
loud can sound good and bad and a lower sound can sound good or bad as well.

i for myself let not fool myself for a loud sound beeing "better" and i guess most people on the kvr forum either.
same for rms.
it's a matter of taste.

honest opinion is the best for each entry.
listen and decide.

[Aiynzahev]

Yes I think we should listen to Compyfox on this. I will try to summarize his guides and stick them on the first post.

We don't have to limit the steps really that comes down to you. So knock yourselves out. Whatever just as long as everyone has fun and we have some uniform way to judge them.

So download the unprocessed mix

[Compyfox]

I still say 3-5 plugins or "processes" are more than enough for a mastering at this rate anyway.

I just listened to the track, and there are barely any edits needed. At least not in my book.

[s_t]

I think the most optimal metering to use in this kind contest would be EBU R127. With time gated function you could define an exact value that the loudest passage must not exceed.

There is even one free for everybody to use:
http://www.kvraudio.com/news/audiocatio ... eter_16313

With RMS/K-12 it is not practical to determine absolute max, as it would require offline validation of the file. That complicates things somewhat. There could be also slight differences between meters.

[Compyfox]

This is why I said: EBU Loudness or Orban Free Loudness Meter.

LUFS at -14LUFS, focus on the SLk meter, average 0LU, fortissimo passages at +3LU maximum.


Or in short - utilize the K-System v2 (which I pretty much told you in a nutshell). Want to know more, hit up my KVRmarks.

But even then, if meters are that drastically different, then I wouldn't advice using them. I'm using K-System, Crest Factor (Dynamic Range), RMS and EBU R-128 meters for quite a while now, and I know their quirks.

Measuring is always purely technical. Feeling wise it's totally objective. I can agree on that end.




SOME SPECS:
According to Wavelab 7's analysis, the unprocesed full mix has a digital peak of -5dBFS, and an RMS signal level of -11,4dB (50ms measurement frame).

Taken a closer look with the K-System meter (600ms), and the DR-Meter, we get the following values:

DR-Value: (8 to 5,8, average 6)
Digital Peak: -5,7dBFS
RMS (K-12v1): -6,24dB

So in theory, the loudness can be raised by +6dB, past +9dB and the sound degrades too much.

I just tested this with Barricade, with an Ceiling at -1dB TP. At +6dB gain, the signal is reduced by 1,4dB - the readout on the K-System Meter would be K-12 AZ+0 on fortissimo passages. At +9dB it's reduced by 4,3dB - the K-System readout results in K-12 AZ+2,5.

Stronger gain raise and the source material starts to pump. Unless you utilize heavy compression/clipping which reduces the dynamic of the song even more.




The STEMs are of no use.


Reason:
available are only the vocals, the acc guitar and the drums. The latter sounds even significally different than in the mix. The rest of the arrangement is missing for mastering with STEMs.

[Compyfox]

Okay, here is my entry including some background info:

FLAC 44/24 (Encoding Level
KVR MC-02 2013-May - Studio Compyfox Entry, FLAC (ca 16MB)

MP3 44/16 (256kbit VBR)
KVR MC-02 2013-May - Studio Compyfox Entry, MP3 (ca 2,6MB)



Technical Stuff:
I really had to limit myself, since I also wanted to add some sort of saturation / glueing effect with a tape machine (in this case Slate VTM), but then I would have used 6 plugins rather than 5. I also tried a compressor to "gel" things together, only to realise that it didn't add anything to the production. So I simply dropped that idea.



I used the following plugins:
Melda Production MDynamic EQ
T-Racks CS Linear Phase EQ
Brainworx bx_control v2
VoS Thrillseeker XTC Black
ToneBooster Barricade


For metering, I used:
ToneBoosters EBU Loudness
HOFA 4U Goniometer
Steinberg Cubase's built in Control Room Meter (K-12)
Steinberg Wavelab's built in Master Meter (K-12)

Since these plugins only analyse, I do not(!) count them to the processing limitation of 5 plugins maximum. IKM ARC is also an essential tool in my studio, but since it's plugged into the Control Room section, it doesn't count as "processing" plugin.

I worked in Cubase 7 in 96/24, and raised the loudness in Wavelab 7 in 44/24. The MP3 in 44/16 was dithered down with the trusty Apogee UV22HR (16bit, Hi, Auto Black).




Step by Step editing:

STEP 1 - Upsampling:

I like to create HD WAV files these days. So my first step was to upsample the stereo mixdown from 44kHz to 96kHz with a stand alone SRC tool. I decided to go against 32bit float this time, since Cubase and Wavelab work at 32bit float internally, and I did not intend to exceed 0dBFS anyway (maximum peak while editing prior to the loudness raise was -5dBFS).



STEP 2 - Editing:

The editing was handled in Cubase 7.

I first fired up EBU Loudness with the K-System Preset K-20v2. I then aligned the track input gain so that the SLk meter was hitting 0 LUFS max and went from there. -18dBFS would have also worked, though important to me was a suitable headroom either way.


MDynamicEQ is taking care of the two sub-bass drops in the song with a reduction of about 3dB, LowShelf at 120Hz. I also used a 24dB/Oct LC at 25Hz to tame the lowend rumble a bit.

The T-Racks LP EQ is used as M/S EQ to boost some frequencies in the mid channel (85Hz with +3,5dB, this gives the song a more modern Pop/RnB sound). But I was also also cutting around 700Hz (-1dB) to free up the mids a bit) and some of the nasty high hats at around 15kHz (very small bandwitdh). I furthermore rolled off the upper frequencies a bit with a highshelf. The sides were mostly cut for the too prominent guitars and hats. To be precise, at 325Hz, plus a roll-off (highshelf) at 12kHz. Since I like a more bass-intensive sound for guitars, I added +2,5dB at 120Hz.

The bx_control is taking care of the Mono Compatibility (up to 85kHz) and spreads the stereo field a bit (plus 15%). I also used bx_control to solo either the M or the S channel to find issues there.

Thrillseeker XTC Black was then used to add the so called "fairy dust". I raised the Air Band around 10khz by plus 8, added some MOJO, but turned off the Drive. On second though, I could have overdone it on purpose at around 8kHz with the knob fully clockwise (10). This overemphasized sound is often liked, but in this particular example, I would have made the hats too prominent. This might be the very same reason that this production could sound dull compared to other productions.

But heck, it's a squashed mix anyway, so no chance to go more in-depth.



STEP 3 - Loudness adjustments:

TB_Barricade was used in WL7 to raise the loudness by +8dB. The Ceiling was setup to -1dBTP, Look Ahead 0,002, Stereo Link 100%, Auto Saturation 30% (0% would have just been good as well), Dithering/Noise Shaping off, no futher changes.

Barricade reduced the incoming signal by 3,5dB maximum. It literally cut off the peaks only. Would there be a stronger signal reduction, I would have used a clipper in front of it. But the limit was 5 plugins, and it thankfully went well.





Technical readouts:

The song clocks in at about +2,5 LU at -14LUFS (K-System v2, K-14v2 preset), or roughly K-12 AZ+2 (K-System v1, K-12 preset) due to the differences in both ballistics and weighting filter. The Integrated Loudness is at -0,5LU after the second or third loop.


Numeric readouts:

Code: Select all

Digital Peak: -1dB TP
K-System v1 (K-12v1): K-12 AZ+1,8 (or +1,8dB max at ff)
K-System v2 (K-14v2): K-14 AZ+2,6 (or +2,6LU SLk at ff), +4,3LU MLk max

Integrated: -0,5 (after 5 cycles)
Loudness Range: +7,8 (-5,7 at 10% to +2,0 at 95%)
True Peak: -1
PLR: +13,5

Invested Time:
With rendering and writing this summary, about 1,5 to 2 hours.





Summary:
If I had access to the STEM's, I would have tamed the bass drops a bit, and rather edit the percussion section directly than in the full spectrum. The acoustic guitars are a bit too agressive in my opinion, and certain instruments drown in the arrangement. I'd put more emphasis on strings and the electric guitar. If I was in touch with the mixing engineer, I'd get back at him to do these changes for me.

But then again, this is how I'd work.


I only did subtle chances, but changes nonetheless. Others might get (way?) better result. But to me, there is nothing more to change or add. Maybe (as already mentioned), overemphasis on the higher frequency ranges, but I'm definitely not going there.

Another thing that stood out, was the drastically spread stereo field. Mono compatibility is guaranteed for the bass section, but the strings are pretty much nonexistent (either way) and while pressing a mono switch, a lot of the guitars were lost as well. I could have tamed that with bx_control, but then the stereo field would have fallen apart, and the song would have sounded "too narrow".

It does work on headphones, but on a stereo in the kitchen, it might drown out.

[wesleyt]

I'm in. I will download the unprocessed mix this evening and work on it over the weekend.

Thanks Sami for proposing this!

[manducator]

Can someone help out a dummy?

I have Klanghelm VUMT, how do I have to set it up?

VU @ -12dB? And let the needle trespass occasionaly by 3-4 dB?

I have to start learning somewhere...

[Compyfox]

In my KVRmarks is a link towards using a fully configurable VU as plain K-System v1 meter.

In a nutshell:
-12dBFS is the reference level of K-12.
The ballistics need to be 600ms rise/fall in order to be within specs.

The 0dB to +4dB VU zone is considered the "Amber Zone" of this recommendation, the 0dB VU is the reference level and where the signal should hover on average. Fortissimo and forte fortissimo passages (in this case the refrain/chorus) should/must not exceed +4dB VU.

The less amber zone is used, the better.
For example: K-14 AZ+4 = K-12 AZ+2

[s_t]

Here is mine...

S_T_KVRMC.flac

Software used:

Samplitude Pro 8 with the following plugins:
1. PLPAReq3
2. ThrillseekerXTC (black)
3. TesslaPro mkII
4. Samplitude's built-in multiband compressor
5. Limiter6