Audio Mastering Notes taken by Robert Chen

These notes are written while taking the “Audio Mastering: the complete guide” course by Marco
Galvan. They aren’t a replacement for the knowledge you can learn from the course but could help your
master process. These notes are “general guidelines”, so you don’t have to follow them to the letter.
The X.0 section numbers are also written in the order you should do them in a mastering session.

1.0 Equalization
➢ Don’t fix a specific narrow frequency band. If you need this, then you need to re-mix.
➢ Use high pass filter at 30Hz to remove sub-harmonics (which you don’t want).
➢ Every curve should be soft with low Q values.
➢ Make subtle changes of less than +/- 2 dB.
➢ More drastic changes to the EQ should be done in the mixing process.
➢ Shelf and Baxandall bands are the most common for mastering.

2.0 Compression
2.1 “To Tempo” Release Times

➢ Use multiples of 60000ms/BPM to find a release time and choose the best sounding one.
• 60000ms/115BPM = 522ms
• Release times can be 261ms, 522ms, 1044ms, 1566ms, etc.

2.2 Serial “Downward” Compression

➢ Attack: 50-150ms
➢ Ratio: 1.5-2.0
➢ Release: To Tempo
➢ Threshold: Aim for a threshold that reduces by 2-3dB
➢ Gain: Make up for the reduction

➢ Controls and smoothens out the peaks of the song

2.3 Parallel “Upward” Compression

➢ Attack: 5-30ms
➢ Ratio: 4.0-5.0
➢ Release: To Tempo, but use slower release if pumping occurs
➢ Threshold: Aim for a reduction of 6-10dB
➢ Gain: Make up for the reduction

➢ Makes the quieter parts of the song louder without affecting the peaks by adding a
compressed signal to the original signal.
➢ An aggressive compression like this can cause pumping which sounds like an aggressive
fluctuation in volume. To mitigate this, give a slower release.
 2.4 Multi-band Compression

➢ Attack: 5-100ms, longer on low frequencies, shorter on high frequencies

➢ Ratio: 1.5-2.0
➢ Release: To Tempo, longer on low frequencies, shorter on high frequencies
➢ Threshold: Right under where the song’s peaks are

➢ Compression to control the dynamics of individual frequency bands.

➢ Can make very drastic changes in overall sound so use carefully.
➢ Compressed in serial.

2.5 Invisible Compression

➢ Attack: 50-150ms
➢ Ratio: 1.1-1.3
➢ Release: To Tempo, at least above 250ms
➢ Threshold: Very low threshold, aim to reduce 2-4dB (even with super low ratio)

➢ Lets you reduce the overall dynamic range to achieve higher overall loudness without
affecting noticeably the sound.
➢ The instructor uses a Studio VCA compressor (ie. Emulation of Focusrite Red).
➢ Compressed in serial.

3.0 Mid/Side Processing

3.1 What Is M/S Processing?

➢ M/S Mid/Side refers to the breakdown of a stereo sound. Mid (mono audio only) and Side
(stereo minus mono audio only) can be edited in M/S processing.
➢ Originally, audio comes in the format of L/R Left/Right. The information in Mid and Side is
encoded with Mid audio in the Left channel and Side audio in the Right channel. Plugins are
applied to change the sound in M/S format and then decoded back into L/R.
➢ L/R processing = Stereo and panoramic control
➢ M/S processing = Ambience and wideness control

3.2 Codification

➢ Left channel = Mono = Mid

➢ Right channel = Stereo - Mono = Side

3.2 Processing Chain

1. Encode from L/R to M/S

2. EQ, compression, etc.
3. Decode from M/S to L/R
➢ Ozone already has Mid/Side processing so the encoding/decoding process is taken care of.
4.0 Reverb
➢ Decay: To Tempo, around 1-2s
➢ Diffusion: 100%
➢ Pre-delay: 20-40ms
➢ Stereo spread: 100%
➢ Depth/Density: 50%

➢ Reverb is usually used in mixing. It’s purpose in mastering is to add an illusion of depth, and should
be used very subtly.
➢ Used as an insert with wet/dry (mix%) control as opposed to fader volume control.

5.0 Image and Phase

5.1 Phase Correlation

➢ A measurement from -1 to +1 of how similar the left and right channels are:
• +1 = mono = 100% correlation
• 0 = stereo = 0% correlation
• -1 = opposite polarity = phase cancellation when summed to mono

➢ On Ozone’s Imager, “invert left” and “invert right” inverts the phases of each channel.
Inverting the left side and then summing to mono will kill the track due to phase
cancellation of the left side being an opposite polarity to the right.

5.2 Sonic Imaging

➢ Rule of thumb: Low frequencies are more narrow and high frequencies are more wide.
➢ Overall sound should have an 0.8 phase correlation, which is the industry standard.
• This will also give your song good mono-compatibility.

5.0 Limiting
5.1 True Peak

➢ On a standard peak meter, the maximum sample level of the signal is not always the
maximum level of the signal. The actual maximum level of the signal is True Peak.

➢ True peak level (dBFS) should be set accordingly to the intended listening format:
• -0.1 dBFS for CD
• -0.3 dBFS for WAV, FLAC
• -1.0 dBFS for MP3, AAC (YouTube, Spotify, iTunes, etc)
5.2 Limiter Release Times

➢ Release: To Tempo
➢ Release time should also take into account the song’s dynamics to compromise between
fast and slow release times:
• Fast release: more loudness (good), more distortion (bad)
• Slow release: less loudness (bad), less distortion (good)

5.3 Direct Current “DC” Filter

➢ Usually added as a fix for DC problems that resulted from analog hardware and some
microphones. The point of reference, where the 0 dB reference axis is drawn in some DAWs
and Audacity for example, is supposed to be where the actual zero noise signal is. Any
displacement is called a DC Offset and must be fixed.
➢ This is not always needed, but when it is, it is crucial. If not removed, the dynamic range and
audio signal can have unintended distortion.
➢ DC Filters should always be enabled. On Ozone, it’s on the bottom right called “Dither”

5.4 Lookahead

➢ Usually enabled in the mastering process as latency is not a big issue. Use your ears to gauge
the effect on perceived loudness to hear if the lookahead is worth it or not.
➢ Computes audio samples a few milliseconds before gain reduction is applied. This results in
a more precise limiting. The cost is a slightly longer plugin latency and sometimes less
loudness.
• Short or no lookahead: Transients remain the same but it may sound more distorted.
Higher risk of going above the True Peak due to calculation error.
• Longer lookahead: Better control of the transients for a smoother sound. Less risk of
going above true peak, but more latency and less perceived loudness.

5.5 Loudness

➢ Loudness is the subjective perception of the sound intensity.

➢ In layman’s terms, it is called Volume to the average person.
➢ Depends on the intensity, frequency content and duration of the sound.

➢ Measured at the end of the mastering chain in the Loudness Unit Full Scale LUFS unit.
• M = Momentary measurement (past 400ms of playback)
• S = Short time measurement (past 3s of playback)
• I = Integrated measurement (average of the whole playback / entire song)

➢ Most streaming services use a normalized loudness of -14LUFS, which result in some louder
songs getting heavy limiting applied, so make sure your master accounts for that.
• LUFS targets can be found online. Use LoudnessPenalty.com.

➢ Rule of thumb: Increase the gain on the limiter so that the integrated LUFS is at the target
loudness, but the actual reduction meter on limiters should only show limiting on the
highest peaks. This way, you aren’t cutting off noticeable dynamic range.
5.6 Hear what a Limiter is adding with a Null Test

➢ A Null Test can be performed on a song to hear exactly what a plugin is adding to your
sound. In the case of limiting, this can be done through these steps:

1. Make sure the limiter on your master track is turned off.

2. Duplicate the track.
3. Invert the polarity “180° Phase Shift” of the duplicated track.
4. Play the two tracks in unison. If done right, the resulting sound should be silence. If
there is sound, that means something other than the phase inversion is different about
the two tracks.
5. Increase gain of duplicated track on a new limiter at the end of the duplicated track’s
plug-in chain by the same amount the limiter on your master track is adding. Do the
same with the release time.
6. Reduce duplicated track’s output level by the amount of gain you’ve added. On Ozone’s
Maximizer, this is done by lowering the Ceiling. IRC I-IV have different sounds.
7. Play the two tracks again. The resulting sound should be pumping, and it is what your
limiter on the master track is adding to the sound.
8. Now you can adjust the release time of the duplicated track to find what release time
flows best with the beat of your song. Be careful: really fast release times causes
distortion – which is not always bad. On Ozone’s Maximizer, the release is called
“Character” which you can change from Fast to Slow.
9. Mute the master track to hear the duplicated track with the limiter applied.
10. Once you find the release time you like, apply it to the original master. You can then
delete or mute your duplicated track.

➢ Different limiters can introduce different kinds of sounds to your master.

➢ Too much limiting will add distortion to your signal – which is good or bad depending on
what you’re aiming for. You can check this using the Null Test too.

5.7 Measuring Dynamic Range with Dynameter

➢ The course instructor recommends Meterplugs’ Dynameter plug-in.

• The plugin only measures dynamic range, not loudness.

➢ The plugin gives two values:

• PSR (Peak to Short-term Loudness Ratio): Difference in dBs between a short-term peak
of the signal and the loudness.
▪ Small PSR (red/orange) = signal is more compressed/limited.
▪ Large PSR (blue/green) = signal is less compressed/limited.
▪ PSR values of less than 8 dB are considered squashed and should be avoided.
• PLR (Peak to Loudness Ratio): Similar to PSR, but with a longer time scale, such as the
whole track.
▪ Large PLR = song is more dynamic, meaning song has loud and soft sections.

➢ Comparing the PLR of your master track with other genres is also possible.
➢ Most streaming services use a PLR of around 13 dB.
➢ Overall this section of the course sounds like an ad but the plugin seems really nice.
6.0 Dithering
6.1 Dither

➢ Dither is the signal applied to improve the bit reduction process when the signal changes
from, for example, 24 bit to 16 bit. It does the following:
• Reduces distortion caused by truncating the bits.
▪ 24 bits sample 100101001010111101011001, when truncated, becomes
▪ 16 bits truncation 1001010010101111 (01011001 is lost info).
• Improves perceived dynamic range.
• Adds noise to the signal (bad).

6.2 Noise Shaping

➢ A tool that distributes the dither’s spectral content towards less audible areas.
➢ Reduces perceived dither.
➢ Only use dither once in a master, only at the very end, and only when you need to change
the bit rate of the audio file.
➢ There are different types of dither, these are the most popular:
• POW-r
• Apogee UV22HR

7.0 Master Delivery

7.1 Master Format Checklist

➢ WAV or AIFF
➢ True peak level (dBFS) should be set accordingly to the intended listening format:
• -0.1 dBFS for CD
• -0.3 dBFS for WAV, FLAC
• -1.0 dBFS for MP3, AAC (YouTube, Spotify, iTunes, etc)
➢ Fade in (very fast) and fade out (around 2s)at the start and end of the export range
➢ No clipping nor distortion
➢ Good mono compatibility

➢ Audio CD standard (red book CDDA)

• 44.1 kHz
• 16 bits
• Dither
• Burned to the CD at lowest possible speed (1x)
 ➢ Hi-Res Audio Master
• WAV or AIFF
• 96 kHz
• 24 bits
• This gives a larger dynamic range and the best harmonic resolution
• Great for iTunes, BluRay, DVD
• Con: “Unfortunately people don’t really like this high resolution audio, I don’t know
why, I hope it will change”, says the course instructor

➢ Disk Description Protocol “DDP”

• The format for disk manufacturing
• Allows for error correction
• Allows to set lengths and gaps between songs for the album
• WAV or AIFF
• DDP software required: Sonoris, HOFA, Reaper
• Con: Some places do not accept this kind of file

➢ Soundcloud, Youtube, Spotify, etc.

• The formats for this are specified by Youtube and music distros, but it will not hurt to do
extra research on specific streaming sites.
• The encoding process is going to result in a loss of harmonics and dynamics, so make
sure you compensate for that.
• To make a single master that covers the biggest variety of streaming platforms, use
these settings:
▪ WAV
▪ True peak level: -1.0 dBFS
▪ Integrated LUFS: -14 LUFS
▪ Resolution: 24 bits
▪ Dither: Only needed if original recording is higher than 24 bits

7.2 Audio Bouncing

➢ Bouncing is turning the project file from your DAW into a stereo audio file on your device.
➢ Make sure your fader does not have any audio playing before bouncing your audio to avoid
tailing effects.
➢ File type should be Interleaved, or L/R Channels (they are the same).
➢ Resolution / bit rate is dependent on what you’re mastering for (read section 7.1)
➢ Dithering is usually either POW-r or UV22HR. You can also try others to see if they are
different.
➢ Do not have Normalize enabled if your DAW offers that option.