16-09-2016, 06:16 AM
(15-09-2016, 04:55 PM)The_Metallurgist Wrote: when i set up the project, rough balanced, no processing, i got -12LUFS PL. that's a pathetic 12dB of dynamic left before we even touch it. this sort of value would normally be seen AFTER MASTERING....
Well, that's how it should be. Recording engineer sets the individual channel gain levels to the nominal operating level of the recording equipment per channel. Emphasis on the words 'individual' and 'per channel'. Usually (almost always) this operating level in high quality equipment is +4 dBu = 0 VU = 1.228 RMS.
When a snare hit, for an example, produces this nominal RMS voltage, it will also momentarily produce a lot higher peak level, let's say 18 dBs higher. So, the equipment needs to be able to handle a lot higher levels momentarily without distortion, than the nominal level. This excess above the nominal level is called headroom. Most high quality equipment can handle at least +24 dBu peak levels, or 'headroom'.
Now, let's enter the digital domain. High quality converters are also calibrated (or can be calibrated) at a certain nominal, or reference, operating level. Usually this is -24, -20 or -18 dBFS RMS for 24-bit converters depending on the standard in use. Let's assume we use -24 dBFS RMS. The snare hit comes in at +4 dBu (RMS) and translates into -24 dBFS RMS at the converter. It peaks +18 dBs higher, so it translates into -6 dBFS peak. Perfect. No clipping and the payload level of the signal was nominal. Cork the champagne. This process is then applied for every individual channel, of course.
Now, if roughly summed, these tracks put together would propably produce clipping and horrible sound. Just as they should, because they are individual tracks recorded at nominal level.
Mixing engineer wants to mix the tracks in the box (on her home computer based DAW). She want's to produce a 24-bit EBUR128 compliant -23 LUFS stereo mix, so she has to calibrate her monitor system accordingly. EBU recommends her, that -18 dBFS RMS pink noise test signal should produce 82 dBA SPL per speaker, as this is their recommendation based on rigorous empirical testing. Although, there are other recommendations as well.
After the calibration she tries to test listen some records. Her head explodes, as they sound way too loud. They are 16-bit records mastered for 16-bit CDs and MP3s, at somewhere between -12 and -9 dBFS RMS, as in 16-bit world the nominal is -10 dBFS RMS for -10 dBV = 0.316 RMS volts. At her new calibration level they sound out at way over 90 dB SPL. She realizes her 24-bit production environment calibration levels are just incompatible with the 16-bit consumer listening environment by definition. Just the way it's supposed to be, and she has find a way to level match them. The quickest and easiest way is to take the level down roughly 10 dBs from her favourite media player. She could also buy a cheap separate USB sound card for the computer operating system audio, and use the 24-bit production system (DAW, audio interface) solely for production, and then switch between these with a separate monitoring controller with level matched +4 and -10 inputs or whatever. There are millions of ways to achieve this, and she figures one which suits her needs best, and now she can A/B compare her 24-bit production system with 16-bit consumer records level matched.
She loads the tracks into her DAW and hits the space bar. Once again, her head explodes as they burst out way too loud. But how can this be? Were they recorded too loud, or hot? Because each individual track is already at nominal level, so their overall level has to go down, so she can produce a mixDOWN, a sum of these tracks, which is again at the nominal level by itself.
Then, why not record individual tracks at even lower gain levels than -24 dBFS RMS, so that the sum of individual tracks produces 'automatically' a nominal level sum, or mix? Because the floor noise per channel is (practically) constant. In subtractive mixing, when you have to only take faders down from their original level, the only thing subtracting is the floor noise, which of course is the goal. And of course each individual recording channel preamp, possible eq and compressor and so on, and converter are designed to provide the best frequency and phase (or transient) response at their nominal operating level. This is the 'one and only job', from technical point of view, the recording engineer has to do: To set the gain levels and equipment to match the predefined reference level, or to go as hot as possible without clipping the tape or converter input if no reference level is predefined. The audible 'sound' is of course a subjective matter due to microphone choices, placement and possible eq and compression, room and so on.
After thinking things through she sets the monitoring level back to EBU recommendation and once again loads the tracks into DAW, and pulls all the faders down and then hits the space bar. First time actually taking advantage of her high headroom 24-bit audio gear and DAW's 64-bit floating point precision mixing engine she starts gradually to open the faders and set panning until the rough mix starts to sound right, and it goes fairly easy as she doesnt constantly follow the level meters. She has faith in the smart engineers who designed the gear, as she knows she's working 'by the book'. The mix starts to sound good, and she checks the levels. -26 LUFS. Should be -23 LUFS. She takes master up by 3 dBs, and now it's the magical (in Europe) -23 LUFS. Hurrah! But her ears start to hurt, as the audible level is now too loud. Once again she thinks over before going into the Internet ranting something about loudness, and just takes her monitoring system down the same 3 dBs. There was nothing wrong with the system, she just had to fine tune her own personal system levels to fit her own personal preference and the current material, room acoustics and so on.
Now the rough mix is somewhat ready and and she compares it (level matched) to the reference track for the first time, which happens to be an international commercial contemporary pop song. It's an 'return to earth' call. The pop hit just sounds better in every aspect, and it certainly feels it has more everything and yet very clean and pleasing. Even the high frequency transients sound better, although in metering it has only half of the microdynamics. But why? Because it was done by the best in the field, who propably applied every established mixing technique developed since the 1950's to make it sound as good as possible for the human ear. Her mix just sounds honky and weak. Off to the Internet to rant about loudness and bash the recording engineer! No.
She returns to her mix, and starts to apply all those dirty tricks from those YouTube tutorials. Smiley eq, transient enhancing compression, serial, parallel, levelling compression, limiting, saturation, side chaining, emphasis processing, delays, reverbs, stereo imaging, creative effects, subgrouping, automation etc etc, she does all the hard work that requires actions and subjective decisions. Now she thinks she has the best mix ever, and still working at the -23 LUFS level. And she compares it to the pop hit, and it endures the test. At least it's not 'that' bad. In the metering she sees, that the new mix has actually less microdynamics, but it still, to be honest, sounds better. That's because the material just doesn't need headroom above 10 dBs. Our human brain will 'invent' the missing parts, when the mix is right, giving emphasis on the transients etc.
Now she wants to 'publish,' or in another words, put the track to Internet. The 24-bit mixdown will sound quiet, and the audio quality will suffer greatly from encoding, unless she does something. Off to the internet to rant about loudness and that all the people in the world should calibrate their $200 laptops, Logitech desktop speakers, phones and earbuds, car stereos etc etc to -23 LUFS, even if they are only 16-bit systems. No. She finds about a process called mastering, or technically, premastering in which the program material is conformed for the publishing medium. She already knows, that in 16-bit domain a widely used level is -10 dBFS RMS for -10 dBV, and in Europe there are regulations that demand portable and handheld electronic devices to operate at so low voltages, that much lower levels than this wont produce decent listening levels on such devices. So, she puts a limiter on the master bus, and brings up the average level to around -10 dBFS RMS hoping, it wont affect the sound too much. And it does not, as her mix is already solid enough.
She sees in some video on the Internet, that she should leave some headroom for something called intersampled peaks. She uses intersampling peak meter, and discovers that hear 'true' peak levels go a little over 0 dBFS. She tests the MP3 or AAC encoded file, and the peaks go higher on the meter. This is impossible! Nothing can go over 0!!1 Well, these peak readings are theoretical predictions of what will happen to the audio in the D/A conversion during playback, and they can occur, but almost all consumer, even the cheapest, D/A chips are designed and built to handle built to handle ISP levels cleanly up to +2 dB dBFS, and her master is just a hint over zero, so she doesn't have to worry about that, although they look a bit scary on the level metering. Some ridiculously loud metal and EDM masters can easily produce ISP peaks higher than this.
Now her master track is on YouTube, Spotify and iTunes. And it plays at audibly different levels on each service. That's because all of these services use their internal loudness normalization, based on ITU 1770. The goal of course is to match the audible level of the tracks in the service, regardless of the RMS or peak levels. iTunes uses basically ReplayGain +/- 0 dB (-15 LUFS Short Term), YouTube RG +4.5 (-10.5 LUFS-S) and Spotify RG +6 (-9 LUFS-S), Spotify being the loudest. Going over the Spotify level won't of course pay any extra 'loudness impact' on Spotify, and Spotify optimized master will play 6 dBs quieter in iTunes etc etc. Main point here is, that the production shouldn't pay too much attention to the LUFS metering at mastering stage, as long as it sounds good and suits the publishing service, or is close enough. The idea of the scale is to provide broadcasters and streaming services a way to more accurately match the differences in delivered audio material. One can of course optimize masters for these platforms for that 'extra' what a 'loud as possible without going over' might achieve. The end user ultimately decides the listening level.
Only production situatation, where the use of LUFS scale is 'mandatory' is live production, where the broadcaster expects the incoming stream to be -23 LUFS, for an example, as they then process it further to, for an example, -9 dBFS RMS for FM radio, usually with hardware equipment preset for -23 LUFS program input. And LUFS meters aren't of any use during recording either. -23 or -24 LUFS standards can be useful during mixing of music, but only if the mixing engineer has to explicitly provide material at such level.
It can be a good practice to mix at -23 LUFS, for an example, but by no means does this mean that the mix has to have excessive amounts, or too much, microdynamics. Or use every available decibel of the headroom. By definition, the idea of the excessive headroom in 24-bit productions system is to provide a fail safe against clipping, not to be the main register for top snare mic.
About the raw tracks. I hear very little, or any, processing in the raw tracks. The snare, for example, has a good amount of peak over RMS constantly, around 15-20 dBs constantly, and is no where clipping. On drum tracks, there are some artifacts that sound like sloppy editing with Logic Pros Flex Time Slicing -algorithm, but this is just a hunch.
About that overly loud supposed CLA mix. It think it's actually quite impressive, that it still sounds like music, and, after the initial shock and turning the volume down, was quite good. I listened the latest Muse album from consumer gear, and only after listening it in the studio, realized that this loud as hell, around -6 LUFS, and still, to my ears, very good.
Listened your mix a couple of times. Generally it's good. Just about the dynamics. I set the initial listening level at the start, and gradually had to take down the volume as your mix builds energy. Your first verse is around -20 LUFS Short term and in the final chorus hits a little over -12 LUFS. I think difference is just too great, and this is no more in the category of micro or macro dynamics, and is just a volume change.
Also, during the first verse, you're averaging around -20 dBFS and your snare is peaking somewhere between -5 and -2 dBFS, which would indicate that there is very little if any compression in the source track or your mix during that part, unless you purposely compressed the original snare down and then used transient enhancer to bring back the transient, which I don't think you've done. So I wouldn't say that the source tracks are overly processed. They are mostly very clean and have no eq and compression, in my opinion.