The secret of maximum loudness (part 1)

The secret of maximum loudness (part 1)

SUBTITLE'S INFO:

Language: English

Type: Human

Number of phrases: 159

Number of words: 2675

Number of symbols: 11383

DOWNLOAD SUBTITLES:

DOWNLOAD AUDIO AND VIDEO:

SUBTITLES:

Subtitles prepared by human
00:08
Hi, and welcome back. Apologies for the slightly click-baity title. But  yes, there really is a secret of maximum loudness,   and it works regardless of the genre, the medium  or the platform. Obviously I'm not going to give   it up yet, you'll have to keep watching, because  we need some background to properly understand it. I remember when I first got my first little  stereo system as a child, thinking that the volume   control set an absolute loudness for playback.  But later on, when I started using my dad's hi fi   to copy vinyl records onto cassette, I discovered  that the record level could make a big difference:   set it too low and my copy would be  really quiet and buried in noise;   but set it too high and I'd  get horrible distortion. With some material I noticed that pushing the  levels a little into the red actually made it   sound better… but that's not the subject of  todays video. The point is, while you had a little   leeway to choose how hot you printed to tape, the  nature of the medium dictated that most material   sat around the sweet spot where you get the best  signal to noise ratio with the least distortion.
01:23
When digital audio was invented, it was originally  assumed that people would continue to work the   same way: the nominal unity level of analogue  recording would be equivalent to -18 or -20 dBFS,   with the extra dynamic  range reserved for headroom. But of course with digital audio, that headroom  is just as clean and linear as the rest of its   dynamic range. There's no distortion penalty  for pushing the levels right up to full scale,   you just get a louder master.  So that's what people did. And then people discovered that, actually you  could push the levels a little bit further   and deliberately clip the converters, and  that this clipping would be inaudible if   it were brief enough, or might  even sound subjectively good. And then the first digital brick wall limiters  came along, and the loudness war really got going   in earnest. Now you could push those levels  harder… and harder… and it wouldn't clip,   it would just get louder… and suddenly everyone  was trying to be louder than everyone else. Let's pause here to discuss  the elephant in the room:  
02:30
yes, louder is better. At least until you  approach the pain threshold anyway. Below   that, every dB of extra loudness will  translate to perceived extra clarity,   extra solid bass, more detail, more  depth, more width. Whatever emotional   impact the music has at a low level, it  will have more of it when turned up louder. But real loudness is not the  same as artificial loudness.   Real loudness takes the whole waveform and scales  it up. Assuming you're not damaging your hearing,   this is a good thing. Real loudness  is why a live gig can be so exciting:   the natural, uncompressed dynamics of a  drum kit, augmented by a good PA system,   provides a visceral experience that you'll never  recreate with a hi fi system and a studio album. Artificial loudness, as created  by a brick wall limiter,   is not the same. Increasing the average  levels makes the audio seem louder,   but the peaks are now smashed down into the  body of the mix, and this has consequences. Not surprisingly this can reduce  the impact of your transients,  
03:39
making them less punchy or snappy. But there  are also more subtle side effects, such as a   loss of depth and space: while real loudness  can fill a room and immerse you in the music,   artificial loudness forms an oppressive  wall that pushes you back in your seat. Perhaps most egregiously, reducing the peak to  average ratio can suck the life out of a mix,   making it flatter, more boring, less  exciting… if it's an upbeat energetic   rock song this is perhaps the worst thing  you can do: the listener won't know why,   but the music simply won't be as exciting as  a classic rock mix with the dynamics intact. Of course, at this point I'm obliged to point  out that we've come a long way since the first   brick wall limiters were released, and limiters  have become much more sophisticated and powerful.   Pro-L2 with its Modern limiting  style can smash a mix really hard,   yet still retain a remarkable  sense of punch and definition. But no amount of clever program dependency  can change the fact that you're reducing  
04:52
the relative size of the peaks, and filling  in the gaps between them. A clever person   once said that music is the space between  the notes… removing or reducing the space   between your transients risks sucking  some of the musicality out of your mix. However: when used carefully, a limiter can make  the audio significantly louder with no negative   impacts at all. And usually, significantly  louder still with only very minor side effects. So this leaves us with a dilemma: if we don't use  any limiting at all, our mixes will play back much   quieter than the majority of other releases.  Of course your listener could just turn it up…   hold that thought because we'll come back  to it… but you'll worry, quite reasonably,   that they won't, and your mix will just  end up sounding weaker than its peers. But on the other hand, competing with the  loudness levels of many modern releases   is also going to make your mix sound  weaker than it potentially could. So, what's the optimum loudness to aim for when  mastering? Or to put it another way, at what point  
06:02
is the benefit of the extra loudness outweighed by  the disadvantage of a low peak to average ratio? Well, we need some way to quantify  loudness in order to answer that question,   so I'm going to have to talk about metering;  how we measure the loudness of a signal,   which is more complicated than  you might initially expect. Let's start with the channel meters in a  DAW. These are quite simple: they flip the   negative half of the cycle to positive, then  display the sample values on a decibel scale. Of course the signal is continually oscillating  through zero, and the meter would be unreadable if   it showed that, so the values shown decay slowly  to smooth out the dips, and display this sine wave   as a constant level. Hence, if the sine wave stops  suddenly… the meter nevertheless decays gradually. This is perfect when you're tracking vocals and  you need to know how close you are to clipping.   But it doesn't cut the mustard for  mastering, for a couple of reasons. The first problem is that this meter just reads  maximum sample values. But when this signal   is converted to analogue the original curvy  analogue signal will be perfectly recreated,  
07:08
and its normal for the peaks of the  analogue signal to be in between samples,   and higher in level than the  sample values either side. Measuring the actual peaks of the waveform  requires a more sophisticated true peak meter,   that calculates values in between  samples. More on this later. Perhaps more significantly however, peak  metering… and also true peak metering…   tells us almost nothing about  how loud the signal will sound. This is partly due to the nature of human hearing.  Its not flat in frequency response, so a sine wave   at 2k… will seem much louder than a sine wave at  200Hz… even though both have the same peak levels. Also, a long burst of 2KHz… will seem much  louder and more obnoxious than a very short blip…   even though both are again at the same peak level. A more real world example would be  drums and distorted electric guitar:   if I normalise these to the same peak level,  the guitar completely obliterates the drums…  
08:16
the drums need to peak at a significantly  higher level than the guitar for the two to   sound balanced, because of the short  spiky transients in the drum track. Finally, peak levels don't really mean much  anyway! The easiest way to demonstrate this   is with a sawtooth wave: here's how that looks  on Pro-Q3's analyser, with a low fundamental   plus a harmonic series above it. Each of these  harmonics is essentially a separate sine wave. Now I'll add a high pass filter: I’ll start with  it set way too low to have any audible effect   on the sawtooth wave… but notice, the peak  level is already reading 2dB higher than it was.   Now lets start to wind up the filter cutoff,  until it starts to shave away the level of that   lowest partial… I’m sure you’ll agree that this  doesn’t sound louder than the unfiltered version…   and yet the peak level of the filtered version  is a full 6dB higher than the unfiltered version. If you've ever wondered why an EQ cut  or subtractive filter can make your   mix apparently get louder, this is what's  actually going on: phase shift has caused  
09:27
all the individual partials to add up  differently, and the peaks of the wave   have got higher as a result. But it's  not actually louder in any real sense. So measuring the peak levels is pretty much  useless when trying to determine how loud   something will sound. That needs a different  approach, and there have been a few over   the years: the classic VU meter relies on the  physical ballistics and inertia of the needle   to smooth out the peaks, and display  more of an average level instead. This is much better than a simple peak meter. But  you need a little practise to interpret what a VU   meter is telling you, especially when it comes  to signals with lots of low frequency content,   which tend to pin the meters even  at perfectly sensible levels. Another approach is to measure  RMS levels, or Root Mean Squared:   without delving too deeply into the maths  involved, this averages out the levels over time;   “mean” being simply a more  mathy way of saying average. This solves the problem of phase shift:  inaudible changes in phase will not cause   changes in RMS levels, and our sawtooth wave  doesn't read higher when I high pass filter it.
10:45
It also potentially solves the problem of  duration, as we’re averaging the signal levels   over time. But the question is, how much time? If  you average levels over a short 50ms window you’ll   get a much faster moving, bouncier reading than  if you average over 500ms… or 2 whole seconds… With longer, larger windows RMS levels correlate  quite well with perceived loudness. But we still   have the same problem as VU meters with bass  heavy signals that tend to read too high. So some meters use weighting to try to emulate  the non flat frequency response of human hearing.   Basically this means a filter  to boost the upper midrange   and cut the low and high extremes, so the  meter responds more or less respectively. This   is a standard A weighting curve as used  for measuring environmental noise levels… I told you measuring loudness was  more complicated than it seems! But we do have a better option these  days, thanks to the broadcast industry   which has developed a new  standard way to measure loudness,  
12:01
with a new unit of measurement known  as Loudness Units or LU for short. This new scale is based on RMS style averaging,  but includes frequency weighting to avoid the   problem of over reading bass heavy signals,  and also defines the averaging window used.   In fact it defines three different ones,  called Momentary, Short Term and Integrated. Momentary averages signals over 400ms,   and gives us this relatively bouncy  display labelled M in Pro-L2. Short Term uses a six second  window, and provides a much   steadier reading that doesn't move  much in response to drum transients.   This is the most useful measure of "how loud  is the signal at this point in the song". The integrated reading is a little  different: this averages out the levels   over the whole song… or the whole  album, or whatever you want to measure.   In Pro-L2 you can click to reset before the  start of the song… and by the end of the song,   or the end of the album, you'll have a single  integrated loudness value for the whole thing.
13:16
Your DAW or audio editor might also  allow you to analyse clips offline   to calculate an integrated  loudness for the whole clip… This might seem like a perfect way to quantify  the loudness of a song, but it does have one flaw   which I can illustrate with Led Zeppelin: here's  Immigrant Song… I won't play it to avoid copyright   strikes, but I'm sure you know how it sounds: it  kicks in hard, and rocks hard all the way through. Now compare this with Stairway to Heaven: this  starts off famously quietly, and builds gradually   all the way through. As a result this mix has a  lower integrated loudness than Immigrant Song,   even though the climax at  the end is about the same. So actually the best way to derive  a single loudness value for a song   might be to isolate the loudest section,   typically the last chorus, and measure the  integrated loudness of just that section. Failing that, watching the Short Term reading  during that section will do just as well. Note that Loudness Units correspond to decibels:   if a mix has a loudness of -16 LUFS,  and then you turn it up by 3dB,  
14:32
it will now have a loudness of -13 LUFS,  which helps to keep things nice and simple. So, now we have a way to measure  loudness, you doubtless want some numbers:   what loudness readings should you be aiming  for? Should you even have a target level?   And does it need to be different  for different platforms or formats? Well, we'll get to that in part two,   along with the secret in the title  of course. Thanks for watching.

DOWNLOAD SUBTITLES: