LESSON 24 · RECORDING & PRODUCTION · PART 4

Mastering, remastering and dynamic range

You put on two editions of the same album. In one, the snare jumps out of the band. In the other, the band sounds fuller and the snare sits closer to the guitars. The players can be exactly the same. What happened between the studio mix and the copy you’re hearing?

Start with the simple idea: mixing balances individual instruments; mastering shapes the finished recording. Heavy compression can make a song seem louder and denser but reduce the difference between forceful hits and quieter moments. The measurements are optional detail.

What happens after the band is mixed?

  1. Recorded tracksSeparate takes of the voice, instruments or programmed parts.
  2. Finished mixThe parts sit together at chosen levels and positions.
  3. MasterThe approved mix is checked and prepared for release.
  4. Release and playbackThe edition reaches your system as a file, disc or record.

Mastering works on the finished picture

The mastering engineer usually starts with the approved stereo mix. They may even out the tone across songs, preserve a big drum hit, tame a sharp sound or leave the mix alone and check it. They also set running order, gaps and fades, check for clicks or overload, and prepare the audio and delivery information for each format. Abbey Road’s mastering overview.

One song, different decisions. A mix engineer might keep a bass guitar steady, control the snare hits on a separate drum microphone, and leave more of the cymbals' rise and fade intact. Once those parts are mixed together, a typical stereo mastering pass works on their combined sound rather than turning down only the snare microphone. See the four-track waveform example.

If the voice and guitar both have a sharp edge in the same part of the treble, an EQ move on the whole mix can soften both. The mix engineer still has the separate guitar and voice tracks. Some mastering jobs use stems, such as vocals and instruments delivered as separate groups, which give the mastering engineer more control.

What the mastering engineer can adjust
  • Tone and cleanup. Broad EQ can rebalance the low end or top end. A de-esser turns down a sharp “s” only when it pokes out. Restoration tools can reduce hum, hiss or clicks on older material, sometimes at the cost of some texture.
  • Level and movement. Compression can act on the whole signal or on a frequency band, such as an unruly low end. Limiting controls the tallest peaks. Some engineers use controlled clipping for level or texture. Peak checks, fades and album sequencing help prepare the release.
  • Stereo. Processing the shared center information and the differences between left and right can alter the sense of width. That move affects any parts in those signals. The separate tracks are where an engineer can move or reshape one instrument on its own.

What are you getting when you buy a remaster?

Original recording
The performances and sounds captured in the session. The original takes can be used again decades later.
Original mix
The balance prepared for the first release, including the parts’ levels, positions and effects.
Original mastered version
The version prepared for the first release. The first LP and CD can have different preparation.
Remaster
A later preparation of an existing mix or source. It can keep the balances of the original mix while changing overall tone, peaks, level or restoration.
Remix
A new balance made from multitracks or other separated sources. A separate 5.1, Atmos or binaural mix can change where parts are placed and how they interact. See how spatial mixes work.
Reissue
A new release of an existing title. It may contain an earlier master, a remaster, a remix or several of these together.

“From the original master tape” still leaves a question. Ask which tape. It could be the finished stereo mixdown or a production tape assembled and prepared for a particular release. The transfer, processing and cutting decisions also affect the sound of the copy you buy.

Yes · Close to the Edge

1972 · Original stereo mix
The US Atlantic LP SD 7244 carried the original Eddy Offord mix. Yes’s credits. Atlantic’s catalog.
2003 · Same mix, remastered
Dan Hersch and Bill Inglot remastered the existing mix for the US Elektra/Rhino CD R2 73790. Yes identifies the 1972 stereo master tape as the source. The specific EQ and compression settings are undocumented in the credits. Yes’s remaster credits. CD edition.
2013 · Multitrack remix
Steven Wilson went back to the multitracks to rebuild the balance and effects. The Panegyric CD/DVD-A set GYRSP50012 includes both his new stereo mix and a direct transfer of the original stereo master at 24-bit/96 kHz on the DVD-A. Yes’s disc contents. Wilson’s release notes.

Try this. On the 2013 DVD-A, select the two stereo tracks and match their playback level. In the quiet “I Get Up, I Get Down” section, follow the voice as the echo trails off, then listen to how the organ enters. Wilson describes rebuilding those small level and reverb moves from the separated parts. The two mixes give you a way to hear which details you notice at the same playback level. Wilson’s interview.

The 2025 Rhino Super Deluxe box makes the distinction explicit in one reissue. CD 1 and the LP contain a newly remastered original stereo mix. CD 2 contains a separately credited 2025 Wilson remix. Rhino’s edition and credits.

Same drummer, three mastering choices

One musical excerpt. Three ways to master it.

These are visualizations of the same invented musical phrase, using the kind of waveform you might see in Audacity or a recording program. Listen with your eyes first: notice how much the biggest hits stand out from the music around them.

Very dynamic

Big musical contrasts

The snare and other accents can jump well above the quieter instruments.

Moderate

Some control, some contrast

The backing sounds steadier, while stronger hits can still stand out.

Heavily limited

Less space between peaks

More sound fills the gaps, but individual accents may feel less dramatic.

Same stereo volume setting: the heavily limited version generally sounds louder overall because more of the music stays near the peaks.

What these pictures do and don't show: One channel of an original synthetic, music-like phrase; the same timing and display scale for every version. The processing is a simplified teaching example, not a measurement or waveform from a commercial song. Waveform density alone cannot tell you whether a master sounds good. The adjusted mode uses matching average electrical signal strength as an approximation, not a measured perceived-loudness (LUFS) match.

Why would the dynamic version seem quieter? With the stereo volume left alone, the highly compressed version may have more sound filling the space between hits. It can grab your attention because the music stays loud more of the time. A dynamic recording often leaves more room for those sudden accents, so its quieter passages may make you reach for the volume knob.

Now turn the quieter version up to roughly the same overall loudness. The snare's crack, the shift from verse to chorus and the pause before a big entrance may stand out more. Turning up didn't invent that contrast. It was already there in the music.

What this asks of your stereo: Those bigger, brief hits need enough extra amplifier power and speaker capacity to come through cleanly. Engineers call that reserve headroom. A heavily limited master can have smaller peaks but keep the amplifier and speakers working harder over time because the sound is more consistently strong. Neither approach is automatically better. See how individual guitar, bass, snare and cymbal tracks can be processed differently before mastering.

Want to see the numbers behind this?

Think of a drum hit jumping above the music around it. Its loudest instant is the peak; the average describes the signal across the whole passage, including the quieter moments. The gap between the two gives one clue about punch, though music has other kinds of loud-and-quiet contrast too. The example below uses volts RMS, a measure of effective signal strength, to show how the peak can ask more of an amplifier.

What a compressor does

It turns down the signal when it gets past a chosen level, called the threshold. The ratio sets how much smaller the rise will be. A slower attack can let the first edge of a snare through. A longer release keeps the gain down further into the ring of the hit. Raising the level after compression can bring the sound between hits forward. FabFilter’s compressor guide.

What a limiter and a clipper do

A brickwall limiter aims to keep peaks below a ceiling by turning them down quickly. A look-ahead limiter delays the signal briefly so it can start turning it down before a peak arrives. Clipping reshapes the waveform as it runs into a limit, which creates new frequencies and can sound gritty. That can happen at the edge of a digital file’s range or when an analog circuit runs out of swing. See what happens when a signal clips.

How to read the graph. All three lines come from the same four-second performance made for this lesson. Follow the dashed mark to the big drum hit, then look at how far each line moves from zero between hits. Use its buttons to compare equal peak levels with equal average electrical levels. This is a separate simplified electrical model, not the DAW-style waveform above.

Signal voltage (V) over time

−10 −5 0 5 10 0 1 2 3 4 s

Same peak level (10 V). With heavy limiting, the signal stays stronger between drum hits. That raises the average level, so the amp delivers more power over the passage.

Restrained

Peak level
10.00 V
Average level
1.99 V
Peak above average
14.0 dB
Average power
0.50 W
Power at peak
12.50 W

Moderate

Peak level
10.00 V
Average level
3.41 V
Peak above average
9.3 dB
Average power
1.45 W
Power at peak
12.50 W

Heavy limiting

Peak level
10.00 V
Average level
5.42 V
Peak above average
5.3 dB
Average power
3.67 W
Power at peak
12.50 W
The figures show one amplifier channel with a simple 8 Ω resistor standing in for a speaker. Power at the peak is the power at that brief instant. See what that means for an amplifier.

Restrained

A level adjustment preserves the shape. Listen for the jump at the start of each hit and the drop as it rings out.

Moderate

Compression and a little limiting pull the tallest hits in. The bass and drum body can feel steadier between accents.

Heavy limiting

The backing takes up more of the available level. The beat is still obvious. The snare’s first crack and the space after it may feel less distinct.

Signal and processing settings

A generated 48 kHz stereo phrase combines a 55 Hz bass part, pitched backing, upper partials and six drum accents. The two channels share the compressor and limiter gain. The chart shows the range of left-channel sample voltages in intervals of about 17 ms. The compressor follows the louder of the two channels. Thresholds below are fractions of the source’s maximum sample magnitude before treatment.

  • Moderate. Threshold 0.32, ratio 2.5:1, 7 ms attack and 95 ms release. Raise level by 1.3×, then limit to a sample ceiling of 0.72.
  • Heavy. Threshold 0.23, ratio 4:1, 4 ms attack and 82 ms release. Raise level by 2.15×, then limit to a sample ceiling of 0.48.
  • Both use a 4 ms look-ahead and 42 ms release time constant in the limiter. The final comparison scales each signal to a 10 V sample peak or 2 V RMS. Both use the same time and voltage scales. The model measures stored samples.

A remaster can change the tone and the punch together

Use the same fictional mix again. The original master leaves the mix’s shape alone. For a later edition, imagine the engineer wants more bass weight and a softer top end. They make a broad EQ change, then add compression and catch the tallest hits with a limiter. Playback gain stays fixed in this comparison.

Signal voltage (V) over time

−10 −5 0 5 10 0 1 2 3 4 s

Find the accented drum hit. The later master’s tallest peak is lower, while more level fills the gaps between the hits. You might hear a steadier body and a smaller jump when the drum hits.

Original master

Peak level
10.00 V
Average level
1.99 V
Peak above average
14.0 dB
Average power
0.50 W
Power at peak
12.50 W

Later remaster

Peak level
8.90 V
Average level
3.13 V
Peak above average
9.1 dB
Average power
1.22 W
Power at peak
9.90 W
Original synthetic example. The equalizer settings and sample measurements describe this one fictional master.
  • Listen to the weight. A little extra bass and less top can sound warmer. The original can sound brighter by comparison. The actual result depends on how much sound was in those bands to begin with.
  • Listen to the edges. A vocal can feel more forward if its strongest frequencies or quieter syllables gain prominence. Less treble can soften some sibilance, the sharp “s” and “sh” sounds. Changing treble can also change how easily you hear cymbals and breath.
  • Listen to the space around parts. EQ and compression can change how easy it is to hear reverb tails or separate a bass note from a kick. The mix still provides the positions that create imaging and the depth cues that can make a soundstage feel holographic.

Before judging how punchy the later master is, turn it down until its body sounds about as loud as the first. Play the same few seconds again. Can you still pick out the drum’s first hit and hear its ring fade?

One approved mix can take two paths to your system

The engineer may start both formats from the same prepared master. They can also go back to the mix and make separate decisions for a file or a record. The last steps change because the two formats carry sound differently.

Digital delivery

  1. Approved stereo mixBalance the tone and peaks for the release.
  2. Digital masterCheck the file format, peak levels, fades and track order.
  3. CD or streaming filePrepare 16-bit/44.1 kHz PCM for CD. Deliver the service’s required file for streaming.
  4. PlaybackDecode the file or read the CD. Streaming may apply playback gain before conversion to analog.

When delivering a 16-bit file, an engineer may deliberately add a tiny amount of noise called dither. That sounds backwards, but it changes how very quiet signals are rounded, replacing certain rounding distortions with low-level noise. They may also check a lossy streaming version after encoding because the process can change the peaks. A 24-bit or high-sample-rate copy cannot restore musical peaks already removed from the original master.

Typical release paths. The cutting engineer chooses adjustments for the actual source, side length and desired level.

Why a phono stage matters. Standard RIAA cutting EQ reduces bass and raises treble. The phono stage boosts the bass and reduces the treble by complementary amounts. This restores the intended balance and lowers high-frequency surface noise. The pressing, stylus, cartridge and phono stage can still affect tone, noise and the shape of a recorded peak. See the RIAA curves. SST’s lacquer-cutting notes.

Eric Clapton · I Still Do

For the 2016 US Bushbranch/Surfdog releases, Bob Ludwig says he prepared Glyn Johns’s original analog mix tapes for Bernie Grundman, who mastered and cut the 45 rpm double LP (51266-1) from them. Ludwig mastered the digital release, including the CD (51279-2). His account identifies separate mastering paths, with the exact EQ and peak treatment unspecified. Ludwig’s account. LP edition. CD edition.

Elbow · The Seldom Seen Kid Live at Abbey Road

Miles Showell says the earlier version used digitally peak-limited files. For the 2019 UK half-speed 2-LP edition (ARHSDLP002), he went back to the 24-bit/96 kHz stereo mix and skipped digital peak limiting. The band identifies the earlier vinyl release as part of its 2012 box. Try following the quieter orchestra between full-band swells if you have both cuts. Showell’s account. Band’s release note.

A higher DR score from an LP capture needs context. Abbey Road cutter Geoff Pesche describes cutting from already-mastered CD files and making adjustments such as de-essing and cut level. In a separate demonstration, Ian Shepherd obtained a higher TT DR reading from a vinyl capture even though the cut came from the same digital master. Treat such scores as clues for an edition comparison, alongside source and replay information. A turntable capture alone leaves the master’s original dynamics uncertain. Pesche’s account. Shepherd’s same-source test.

How getting louder changed the sound

When listeners switched between CDs or tracks at the same volume setting, the one with more level could grab attention first. Once the tallest digital peaks were near the ceiling, engineers could get more level by reducing the peaks and raising everything together. In some releases this left less room for a snare to jump out or for notes to fade between hits. Researchers found changes to peak-to-average relationships in mainstream music during the loudness war. Larger-scale contrast, such as a quiet verse leading into a loud chorus, is a separate part of the picture. AES research on mainstream dynamics.

Some listeners like a dense, immediate wall of guitars. Others hear congestion or a harsh edge when sustained parts compete with the hits. That may get tiring for them. A recording can feel dynamic when the arrangement opens up, or have strong separation because the mix makes space for each instrument. A single peak number leaves both of those experiences unexplained.

1985 · Vertigo CD 824 499-2 · matrix 12

Dire Straits · Brothers in Arms

Listen for the build. In “Money for Nothing,” the keyboard and distant voice give way to the drum build and a big guitar entrance. The producers Mark Knopfler and Neil Dorfsman shaped the performance. Dorfsman recalls committing EQ during recording, then concentrating on balance in the mix. Bob Ludwig describes making a dedicated CD master and later preferring his own digital remaster. An independent analysis documents the high peak-to-average contrast on this early CD matrix. Dorfsman interview. Ludwig interview. Matrix and analysis.

2008 · US Warner Bros. CD 508732-2

Metallica · Death Magnetic

Listen through the full band. When the guitars and drums are going together, follow how much the kick or snare steps out from the sustained guitar sound and whether you hear a rough edge. Producer Rick Rubin described pushing the mix toward breakup for excitement. Andrew Scheps and Greg Fidelman mixed and Ted Jensen mastered. Reporting at release described clipping and Jensen’s account of receiving already-limited mixes. Several stages helped shape the result. The band also sells a download identified as a separate 2015 mastering. US CD credits. Rubin interview. 2008 reporting. 2015 master.

2013 · US Columbia CD 88883716862

Daft Punk · Random Access Memories

Listen for the section changes. In “Giorgio by Moroder,” follow the speaking voice into the synth pulse and on to the much fuller drums, orchestra and guitars. Mix engineer Mick Guzauski describes aiming for the sound of records before the loudness war. Mastering engineers Bob Ludwig and Antoine Chabert worked on the album. Ludwig describes retaining punch and contrast as part of the record’s impact. The changing arrangement gives that decision room to matter. US CD credits. Engineers’ interview.

A credit tells you who worked on the record. Their explanation can tell you what they were trying to preserve or change. Ludwig mastered the Dire Straits and Daft Punk examples, but each began with its own performance, mix, brief and release format. Choose the edition by listening to the choices you care about.

Which kind of “loud” did the meter measure?

dBFS and sample peak
dBFS measures a digital level relative to full scale. A sample-peak meter reports the largest stored sample. This tells you how close the file’s samples are to the digital limit. Your volume setting, amplifier, speakers and room determine the sound level at your chair.
True peak (dBTP)
When a player turns the stored samples back into a continuous signal, its highest point can fall between two of those samples. A true-peak meter estimates that highest point. It helps an engineer check how much room is left for the peaks when preparing a file.
Average level (RMS)
RMS voltage tells you what steady voltage would deliver the same average power into the same load. For example, a musical passage measuring 2 V RMS delivers the same average power as a steady 2 V into the lesson’s 8 Ω load. To compare how loud two songs sound, use your ears or a loudness reading.
Overall loudness (integrated LUFS)
Estimates how loud a whole song sounds. It accounts for how sensitive our ears are to different frequencies and excludes very quiet stretches. Streaming players can use it to adjust the playback volume across tracks. A song with a huge drum hit and one with a steady backing can get similar readings.
Loudness range (LRA)
Compares the typical quiet and loud stretches, leaving out the extremes. For the size of a single drum hit, check its peak or look closely at the waveform.
Peak above average (crest factor)
The gap in dB between the peak and the average level (RMS) over the same passage. A larger gap means the peak sticks out more. Change the passage you measure and the number may change.
Musical dynamics
What you hear when a player accents a note, a drum decays or a chorus opens up. A song can keep its verse-to-chorus contrast while losing some drum attack. PCM bit depth affects how finely a file can represent amplitude and its noise floor.

Standards define how to take these measurements consistently. ITU-R BS.1770 (loudness and true peak). EBU Tech 3342 (loudness range).

What the streaming service changes at playback

Imagine a hot remaster followed by a quieter original in a playlist. With normalization on, the player may turn the remaster down and raise the original if there’s room for its peaks. A gain change moves the whole waveform together. The space between the snare and the backing survives that move, including any change to it already made in the studio.

Spotify

Spotify documents −14 LUFS in Normal mode on supported clients, plus Premium Quiet at −19 and Loud at −11. It normally uses a shared adjustment for an album played in order and track-by-track adjustments for playlists or shuffle. Available peak room can limit how far Normal raises a quiet track. Loud mode can apply a limiter during playback. Spotify says normalization is skipped on its web player and third-party devices such as some TVs and speakers. Spotify’s artist documentation.

Apple Music

Apple’s Sound Check adjusts playback level across songs. You can switch it on or off in Music’s Playback settings on a Mac or in Music settings on an iPhone or iPad. Confirm which mix you’re playing as well, since switching between stereo and Atmos can change the music’s balance and space. Apple’s Sound Check guide.

Spotify’s Normal setting is a playback reference and its artist guidance is a recommendation. A separate broadcast recommendation, EBU R 128, calls for −23 LUFS programme loudness in its own context. Mastering, player gain, playback limiting and lossy encoding can each affect a different part of what arrives at the DAC. EBU R 128.

Try a fair listen. Find the same passage in the two editions. Use the same device, listening position, output format and settings. Match how loud the backing sounds, using a small volume adjustment if needed. Then compare the drum’s attack, the weight of the bass or the end of a reverb tail. Check the release credits if the positions of instruments or effects seem to change.

What changes for the amp and speakers?

One version can ask for more output over time. Another can ask for a bigger burst for a fraction of a second. Both can be true of the same two releases depending on which level you hold equal. Here are two small examples, each across one ideal 8 Ω resistance.

Same peak · The denser signal uses more average power

Both reach 10 V at the peak. Between peaks, the denser signal stays at a higher level.

Average power at equal peak voltage
SignalAverage levelAverage power
More space2 V0.5 W
Denser4 V2.0 W

Four times the average power. If that higher average holds over sustained playback, the amp delivers more energy per second and the speaker’s voice coil can run warmer.

Same average · The bigger accent needs more swing

Both have the same average level, 2 V RMS, and use 0.5 W on average. One has a bigger drum hit.

Power at the loudest instant with the same average level
SignalPeakAt that instant
Taller hit10 V12.5 W
Shorter hit5 V3.125 W

Twice the peak voltage and current. That’s four times the instantaneous power. An amp needs enough headroom to make that short swing cleanly.

Show the calculations

For a fixed resistance, average power = RMS voltage² ÷ resistance. So 2² ÷ 8 = 0.5 W and 4² ÷ 8 = 2 W. At an instant, power = instantaneous voltage² ÷ resistance. So 10² ÷ 8 = 12.5 W and 5² ÷ 8 = 3.125 W. At those two peaks the currents are 10 ÷ 8 = 1.25 A and 5 ÷ 8 = 0.625 A. Crest factor = 20 × log₁₀(peak voltage ÷ RMS voltage).

Read the amp rating under its test conditions. Power at the peak tells you what happens at one instant. A continuous power rating comes from a sustained test, usually with a steady tone. A burst rating comes from a short test. A real speaker’s electrical load changes with frequency, so it can draw different amounts of current on different notes at the same voltage. Strong bass can make the cone move farther. High average power over time can warm the coil that moves it.

More clean headroom can preserve a peak that the amp would otherwise clip. If the release itself has already flattened the peak, a better amp or DAC still receives that version of the waveform. Your system’s response, distortion, DSP and room can change how prominent the effect sounds.

Sources and figure notes

Album references identify specific releases and link out to credits and engineers’ accounts. The four-second signals, EQ and delivery diagrams were created for this lesson. Waveform values use sample peaks, RMS and crest factor. The power figures assume a fixed resistance.