By Dan Loh
Objective – Achieve a loud enough track (0dBFS peak) with not much increase/decrease in volume between songs on an album, no digital clipping, no overly reducing dynamic range and altering the balance too much with EQ.
References – I would’ve already started using this at mixing to check my overall mix competes with commercial mixes to some extent and to recalibrate my ears, I check that I’m not overdoing something in my mix i.e. reverb, frequency imbalances, weird stereo imaging and to check if my song sounds as good as I thought it was.
At mastering this is the final line of offence, being objective to see if the song is missing something and or something that completely crossed my mind (that’s why I prefer to send my mixes to mastering engineers as they will have be listening to the song for the first time, and will capture something deadly obvious to them). We see frequency masking at its absolute best here, what I did is drag a .wav file into a session (.mp3s are compressed so quality will be lower), some plugins include a ‘mid side’ function, this allows you to only affect the mid or left & right channels, A technique Glen Martin uses is to send these to mono aux tracks and flip the phase relationship on one of the speakers with any plugin (in PT), send these to mono aux’s but on a different bus and flip the inverted channel back again before finally arriving at your mix bus, each fader you bring up will cancel the phase on either the mid or sides. Insert an EQ and create a band pass, PT’s stock EQ includes a frequency solo option by holding down SHIFT SHIFT + WIN, SHIFT + CMD/CTRL.
Doing so allows us to directly listen to what sounds occupy the frequency spectrum so that everything can be heard without clouding the mix, when referencing I check how the instruments are well separated and have their own space, most commercial mixes will have bass frequencies in the mid channel so both monitors share the load of bass, and mid – high end on the sides, this reveals a lot of panning pockets that allows space for other things like effects to fit in.
This is gold for those with a limited monitoring system (what I suffer from), investing in a pair of revealing monitors and room treatment (if not HIGH QUALITY headphones are a better solution), regular changes of each system but still use references as they are like mix/master tutors.
References should be similar to my own mix in some way, or can be used for its specific characteristic i.e. ‘Natalie Imbruglia’s Torn’ is known for its very prominent mid-range, and sibilance. I will use this as a threshold guide, so if my mix has more sibilance or mid than the track, I know I should probably tone it down.
RMS – Route, Mean, Square or ‘Crest Factor’ as it’s known, is the difference between the loudest peaks and its average level. There is a school of thought that suggests all music should be a certain RMS level, while others believe that some styles should have a shorter dynamic range than others. You don’t want the dynamic range so large that people have to constantly reach for the volume (imagine the car incidents, should a sudden loud chorus frighten motorists). You want it loud enough to compete against other music, so there’s not much of a volume change in between songs on radio and on the album.
Loudness Wars – It is most important, that none of these mixes are ‘squashed’ dynamically to achieve maximum loudness for every part of the song, this means that snares, kicks, vocals, cymbals are so similar in volume it bores and fatigues the ear. “Ahem” Death Magnetic “cough”, what happens is the peaks of the song have been pushed beyond 0dBFS and essentially the waveform information above this is ‘chopped’ off creating a ‘flat top’. Kicks and snares are main victims, where the punchy/snappy info is lost and digital clipping occurs.
Ian Shepherd at Production advice produced the graph below as an example of how the less dynamic range music can impact sales and overall success of the album.
The benefit of dynamic range other than not clipping the audio thus riddance of good info above 0dB threshold, also the build-up of the song that naturally occurs, you don’t want the final chorus to be the same level as the first, you want sections jumping out at you (not like Albini) and music to impact all of our human senses not just hearing. Engineers work hard to ride faders, sculpt tones that will impact listeners, don’t let loudness be the main solution to grab audiences by the……ears.
Red Book Standard – dither will create noise under the recording; you have the option to EQ this (typically a choice of harmonics i.e. triangle, rectangle etc.) Also a method used when lowering the resolution of bit depth (I.E. 24bit to 16bit). When lowering the value of bit depth you lower the values available for measuring amplitude with audio samples, this leads to quantisation errors (aka. Truncation distortion). We use this when converting higher bit files into 44.1 KHz/16bit which, is the format used for CD.
iTunes – In our era the majority of music will eventually make its way on iTunes, which uses the same format as CD however, the quality standards of sample rate conversion are very specific on iTunes in order for your music to make it on there. Image below shows how a sample rate conversion should look, the one below (with all the colours) is how it shouldn’t.
The latest version of Apple’s AAC encoder works better for sample rate conversion, iTunes recommends that none of the files should have any clipping (good practice for all mastering). The requirements are at least 24bit format and original sampling format, -0.1dB of peak limiting. Avoiding clipping isn’t as simple as it sounds, take Apple’s Roundtrip AAC which is a live encoder to check compatibility before loading it on iTunes. The image below detects inter-sample peaks even though it won’t necessarily show up in your DAW’s own master meter.
Because of digital audio, the DAC may sometimes produce an analogue signal momentarily exceeding the digital peak 0dBFS, this is known as an ‘inter-sample peak (image below: http://www.hometracked.com/2007/11/08/prevent-intersample-peaks/) red = digital, the blue line is the analogue signal which briefly overlapped the 0dBFS threshold.
The side-effects could result in audible digital clipping should it cross the 0dbFS mark a lot further. And that’s why I average my own signals at -18dBFS way before mixing which is plenty of headroom to avoid this without use of special gear.
EQ for loudness – In most cases the bottom end is the first to suffer at the hands of limiting, due to the high energy and headroom it occupies, an EQ carve at this region allows for the overall level to be boosted without clipping, the same will be true for any frequency that is hitting the ceiling before you can get the rest of your mix up to 0dB. A linear phase EQ is a better choice here for bigger cuts and boosts as they won’t alter the phase that a regular EQ induces.
Multiband compression – Are frequency bands that split the signal so that only that particular frequency range can be compressed, it is common for a particular frequency to get excited and resonate during a part of a song, by using just an EQ this might affect all the other sections of the song i.e. if you cut a high frequency to tame one part of the song, the rest may sound dull, a multiband compressor with only compress the assigned frequency when it crosses the threshold allowing for the rest of the track to maintain its balance.
I dragged all songs into a blank session to help me get even levels between each track. A limiter was inserted on my master fader with the ceiling at –0.1dB to prevent any clipping.
Down to Dusk – Again a band pass filter, this time a cut at 1.85KHz as some of Kaia’s vocals were a little too strong in this mix. The song’s most narrow RMS was -14dBFS, the cymbals caused clipping to occur in some sections which is why the max peaks were around -0.3dBFS.
Love me like you do – The room noise is audible due to majority of mics being room mics. -1dB at 400Hz with a narrow Q and -1dB at 750Hz rid of some weird resonance in this area but made track sound less open, so skipped that and cut some 3.1KHz by -0.9dB as this area was to harsh but still being careful not to rid the presence of the vocals in the process. finishinf the song with peaks of -0.1dBFS and an RMS of -13dBFS at busiest sections.
Don’t you pity – the balance here was best left alone here, as it sounded decent across many playback systems, a high and low pass was all it needed to tame the extreme highs and lows. some compression just to bring down peaks from the snare to bring the rest of the track up, leaving the most narrow RMS at -10dBFS
Hole in the Earth – High and low pass filters to cut the extreme lows and highs that most people cant hear anyway, a -0.9dB cut at 750Hz as I felt this frequency was a little over resonant and prominent in the mix. My theory as to why my mid-range seems to be drowning may be due to the hours I spend making critical EQ moves (normally 2-3 hours into a mix), my monitoring levels aren’t loud but my perception does lose track rather quickly in that amount of time.
Mother – finishing weak a max peak of -0.1dBFS and an RMS of -12dBFS at busy sections.