Core idea: arranging is a space budget, not an instrument list.
Musical space has boundaries, like a canvas; adding beyond the limit will cover up what is already there. Each time you add a part, the other parts have to work harder to compete for the listener's attention. (p0286)
So "adding a layer" is never free. What this skill governs ishow to allocate this limited space。
Which sections to read by task
| Symptom/Task | Read |
|---|---|
| Muddy, murky, unclear | §4 Spectral budget |
| Flat, two-dimensional, no depth | §2 Three depth layers |
| The lead doesn't stand out / someone else is stealing the show | §3 Depth cues + §3.2 Pitfalls |
| Written but inaudible | Read §3 and §4 together |
| The later the section, the messier it gets | §5 Rate-of-change discipline |
| Want to make a very dense but not messy section | §6 Maximum density method |
| Fill in ARR-SPEC's spectrum_budget | §7 ＋ §8 |
Boundaries: what this skill does not govern
| Not here | Where it belongs |
|---|---|
| Which instrument to use, how to blend timbres | mc-orchestration |
| When to enter and exit, the energy curve | mc-arrangement-arch |
| Specific EQ frequencies, compression ratios, mastering | Mixing matters; this package only gives mix_intent this intent |
| Sections making way forintent(which section yields to which) | mc-mix-intent(this skill governsstructurallayering; it governssectionalyielding) |
| Layering within the drum kit | mc-rhythm-section |
mc-texture-layeringandmc-mix-intentthe boundary is easily blurred; remember this: This skill's solution is "rewrite" (change register, reduce notes, change instrument, stagger timing); mix-intent's solution is "remix" (frequency ducking, sidechain, dynamics). What can be solved at the composition layer should not be left to mixing—that pushes the problem downstream.
1. Must fill in before writing
| Required | Question |
|---|---|
| ① Depth map | Which layer is each element in—foreground/middle-ground/background? (§2) |
| ② Frequency-band assignment | Which frequency band does each element mainly occupy? If you can't say, you haven't thought it through yet |
| ③ Conflict list | Which two are fighting over the same block of space? (§4) |
| ④ Solution for each conflict | Frequency ducking / stagger registers / alternate appearances—choose one of the three, and state which one |
2. Three depth layers: foreground, middle-ground, background
Problem: The music feels "flat" or two-dimensional, lacking a sense of depth (p0111).
Approach: Borrow concepts from the visual arts (film, photography, painting),and plan depth layers at the composition stage, rather than leaving it all to the mix (p0111).
| Layer | visual analogy (p0112) |
|---|---|
| Foreground | Closest to the observer, usually the main subject, in sharp focus, rich in detail |
| Midground | Between the two, not the main focus, but closer and more prominent than the background |
| Background | Farthest away, recognizable but blurred, with less detail |
How to push and pull at the production level (p0112–0113)：
| Pull closer (push toward the foreground) | Push farther (push toward the background) | |
|---|---|---|
| Volume | Louder | Quieter |
| Timbre | Brighter (add high frequencies) | Darker |
| Space | Drier (higher proportion of the original signal, less reverb) | More reverb |
★ But what this skill cares about is depth at the composition level, see §3—that part cannot be saved by mixing.
3. Depth cues: on what basis do listeners decide who the protagonist is
3.1 Four perceptual cues
The human ear tends to hear sounds that are "high," "loud," "fast," and "fast-changing" as the foreground. (p0113)
Note that among these four, only "loud" can be adjusted by mixing; the other threeare fixed at the composition stage。
Among them, Howard & Angus gave auditory explanations for three:Loud—Hearing is most sensitive around 3–4 kHz; at the same sound pressure level, this band sounds much louder than low frequencies (p0091, p0093); High—Timbre with prominent harmonics above the 5th–7th is described as bright, piercing, and highly penetrating (p0223, p0228); Rate of change—Hearing is very sensitive to changing sounds; unchanging things get "adapted" away (p0352, p0342).
A precise example (p0113)：
The drum kit is usually very loud, but because of itsconsistency (consistent and unvarying), it is usually placed in the midground or background; when a **drum fill** occurs,the sudden change in texture density and complexitymakes the drums temporarily jump to the foreground.
This explains "rate of change" thoroughly: the same instrument, without changing volume or timbre,can shift in depth by rate of change alone. —So a fill is not decoration, it isdepth scheduling。
3.2 ★ The most common trap
If the synth melody lineis an octave above the vocals, orhas a more complex rhythm and more drastic note leaps, listeners may subconsciously hear the synth rather than the vocals as thelead voice。 (p0113)
This is the real reason "the vocals are buried"—many times it is not a volume problem, it is that the perceptual cues are wrong. Turning the vocals up 3 dB will not solve it, because the synth wins on both "high" and "fast."
Vocals also have a cue of their own: opera singers are heard above the entire orchestra thanks to the 2.5–4 kHz "singer's formant" (Howard & Angus p0209; measured 2.4–3.5 kHz for tenors, p0221). Pop vocals do not have such a strong formant, but the principle is the same— this band is the window for vocals to stand out, the harmonic peaks of the synth, distorted guitar, and cymbals pile up here, and the vocal recedes to the back. Self-check: in the chorus, who is occupying 2.5–4 kHz. Failure condition: not applicable when the protagonist is not the vocal.
Self-check action: for each section, list who is currently the highest, the loudest, the fastest, and the most drastically changing. If the protagonist does not occupy at least two of these four answers, the protagonist is not the protagonist.
3.3 Solution
It is not about turning the protagonist up, but aboutpressing the competitors down on the four threads：
| Where the competitor wins | How to press it down |
|---|---|
| High | Move it down an octave, or let it appear only in the gaps between the protagonist's phrases |
| Loud | (leave this one to the mix) |
| Fast | Reduce note density, lengthen durations |
| Rate of change | Turn it into a repeating pattern—consistency will push it toward the background |
3.4 ★ Six means of separating layer from layer (Yang Liqing)
§3.1 discussed what makes listeners hear a layer as the protagonist; this section discusseswhich knobs you have on hand to pull two layers apart(Yang Liqing p0244–0250):
| Means | What to do | Key points |
|---|---|---|
| Timbre | Switch to a different family of instruments | The dominant factor in layer stripping (p0244) |
| Register | Widen the intervallic distance, or put the main line in its most characteristic register | The greater the distance, the less the overlap;when the main line is in its most characteristic mid-high register, it can float out even if it is tangled with the accompaniment in the same register（p0246） |
| Articulation | Change the playing method | Timbres with different temporal transition characteristics can only produce heterogeneous sound, which is a favorable condition for stripping layers (p0247) |
| Density / concentration | Comparison of the number of players in solo, soli, and tutti | ★ Even if two layers use exactly the same instruments, they can be separated by the difference in number alone (p0248) |
| Amplitude | Single-note group versus multi-note group (unison versus octave, multi-layer octaves) | Auxiliary role (p0250) |
| Dynamics | Mark one level higher or one level lower | ★ Always only an auxiliary means. If register, timbre, density, and amplitude are already out of balance, dynamics cannot save it (p0250) |
Usage: when two layers clash, try from top to bottom according to the table,leave dynamics for last. The mixing layer can only move the last row; the first five rows are all at the arranging layer, this is the basis for "what can be solved in the writing should not be left to the mix" (reference §7.4).
4. Spectral budget
4.1 Every instrument must be able to state which frequency band it occupies
This is the use of the two fields roster[].band and spectrum_budget in ARR-SPEC.
If you cannot state an instrument's frequency band, you also do not know whom it is clashing with.
4.2 Three solutions to conflict
When two instruments occupy the same block of frequency band,there are only three paths, and you must choose one and state it clearly:
| Solution | What to do | When to use it |
|---|---|---|
| ① Frequency yielding | One side yields while the other sounds (rewrite: reduce notes, shorten sustains, change chord voicing) | Both must be present, but with different importance |
| ② Stagger the register | Move one side as a whole to a different register | Both are important and both must be present throughout |
| ③ Alternate | The two do not sound at the same time; they take turns by section or by phrase | The two have similar character; appearing together would only cancel each other out |
Item ③ is often overlooked, but is often the best solution——it simultaneously solves the frequency-band conflict and mc-orchestration the neutralization problem in §3.4.
Each of the three solutions has a boundary drawn by Howard & Angus:
- Stagger the register: masking extends upward, and low frequencies mask high frequencies far more strongly than the reverse (p0239)——moving the secondary voice below the protagonistbelowis more dangerous than moving it above; the critical bandwidth increases with frequency (p0083–0084), so the same interval is more likely to sound muddy in the low register——four-part writing requires no less than a fourth between the bass and the tenor, unless they are already above the bass staff (p0146).
- Frequency yielding: within the same critical band, the masking threshold is only about 20 dB below the masker (p0240)——if the secondary voice in the protagonist's frequency band is not at least 20 dB quieter, it is "written but inaudible," and it is better to delete it or change its register.
- Alternate: within 100–200 ms after a sound stops, a quiet onset is still masked (p0241)——an "alternation" whose gap at the end of a phrase is shorter than 200 ms does not count as alternation.
How one measured sample is written (City Pop):
1–3 kHzsection: clean guitar conflicts with the synth lead → Solution: alternate, the guitar only in the verse/lift section, the synth only in the phrase-end gaps of the intro/chorus/outro.
4.3 "Thinning": intermediate means other than complete removal
You do not have to delete an entire instrument; you can just thin it (p0287):
- Reduce the number of notes a given instrument plays
- Shorten the decay time of sustained notes, letting it quickly give way to the rest of the mix
These two are solutions at the composition layer, not the mixing layer: what changes is how many notes are written on the score and how long the note durations occupy, not carving out a certain band of overtones with EQ or compression, so this belongs to this skill, not to mix-intent.
How to choose between thinning and complete removal: if an instrument carries a midground or foreground role from §2, deleting the whole instrument will leave that layer empty, so try thinning first; if it is only filling density in a certain passage and removing it does not affect who is speaking, complete removal is cleaner.
After piling a section up very densely with the §6 maximum-density method, it is easy not to cut enough when going back to do the two rounds of reduction in §4.2/4.3: an over-layered section needsa great dealof erasing to restore musicality, and the slightest softness will leave a pile of notes competing for attention (p0288). The way to judge "enough" is not to count how many instruments were deleted, but to listen again after deleting: if any part still cannot say which frequency band and beat it occupies, it has not been cut enough.
4.4 A premise that must be made clear first
The subtraction strategy is effective only when your existing parts are already as good as they can be. (p0286)
Delete a few from a pile of mediocre material, and what remains is still mediocre.Layering cannot save bad material.
When to think along this line: when a section sounds "small," "weak," or "powerless," the first reaction is often to add things, but the principle is the reverse: every additional part makes the other parts work harder to compete for the listener's attention. Musical space is a canvas with boundaries; overdoing it only covers up what has already been written (p0286). At this point the question should be whether something can be taken away, not what else can be crammed in.
This is not the exclusive province of minimalism, nor a technique useful only to a certain genre: music with a slow rate of change can sound minimal even if the number of elements is not small (see §5 Rate-of-Change Discipline); conversely, high-density commercial EDM can still sound clear and open as long as the number of simultaneously sounding parts is deliberately limited (p0287). The core principle is to add only when necessary, but only when necessary, and at the same time ruthlessly delete anything that does not contribute to the music (p0287).
4.5 Stacking Is Not the Same as Being Louder
Two instruments at the same loudness stacked together only +3 dB, ten only +10 dB, and +10 dB is what doubles loudness (Howard & Angus p0040, p0094–0095). To make a section "bigger," spread energy across more critical bands—total sound intensity unchanged, loudness still increases (p0096)— that is, add toempty frequency bands, rather than stacking unison onto already occupied frequency bands. Failure condition: coherent summation of copies of the same signal is calculated separately (in phase it can reach +6 dB, p0035, p0037). Self-check: which previously empty frequency band does the newly added layer occupy? If you cannot answer, it is a wasted addition.
★ Beyond the frequency axis there is also the time axis(Takeuchi p0028–0031): the impression of sound pressure depends onwhether there are silent gaps. Popping a balloon every three seconds does not give high sound pressure; popping continuously does, with exactly the same maximum volume (p0029). When the kick drum plays quarter notes, the spaces between beats are empty, so the kick drum alone cannot push sound pressure high; the bass must fill the gaps; but no matter how much low frequency fills in, according to the equal-loudness contours it does not bring a sense of high sound pressure,what truly pushes sound pressure high is covering 1–5 kHz with sustained tones(p0029–0030). That is exactly the frequency range where guitar, vocals, and piano fall, so singing while playing is actually easier to make big than a continuous kick drum (p0030).
Arrangement implication: if the chorus is to be "big," first look at two things: whether there are sustained tones filling the gaps between beats, and whether 1–5 kHz has a sustained layer (pad, distorted guitar wall, harmony vocals). If neither is done, adding low frequency will only make it muddy.
4.6 ★ Pedal Tone: When to Add a Sustained-Tone Layer, and When Not To
A band pedal tone is simply a sustained note or sustained chord that continues sounding; pads and sustained strings in pop arrangement are all this (Yang Liqing p0309). Four functions (p0309):
- Mediation: communicates between different timbres, letting the layers merge into a whole
- Stabilization: gives a clear harmonic outline within a flowing figurated texture
- Fullness: makes the sound rounder and thicker
- Contrast: forms a contrast in density and intensity with sections that have no pedal tone
★ When not to add it(p0313): when the melody, bass, or inner voice in the texturealready contains long-duration noteswhen they are already functioning as pedals, adding another layer of sustained tone is gilding the lily. For styles that pursue a harsh, dry sound, also add less; the frequency with which pedal tones appear is itself a style marker.
How to choose notes(p0313): For passages with frequent harmonic changes, choose thecommon tonesbetween the chords as the pedal, letting it become a unifying factor; if no common tone can be found, sift out a few long notes at different pitches and try to make them connect into a pleasant line. The pedal layer does not have to be long notes; tremolos or rhythmically simple figures also work.
Gluing(p0311): Add a layer of sustained chords above a dense figure, making its register, arrangement, and density align with the figure layer on the first beat of each bar, and the two layers become glued together. Conversely,ostinato figures and percussion rhythm patterns are also glue(Yang Liqing, Volume 2, p0514–0515): A clearly contoured, repeatedly occurring rhythmic pattern can integrate an intricate structure, both providing vertical integration and forming horizontal connections between sections.
5. Rate-of-change discipline
The rate of change of midground and background elements should be kept to a minimum, while foreground elements allow more freedom of movement. (p0114)
This one is extremely practical, and it is the antidote to "the later it gets, the messier it becomes":
- Background is changing → the listener's attention is constantly being pulled around, and the protagonist cannot hold its ground
- Background is unchanging → every change in the protagonist is heard
Self-check action: Write out "how often every few bars" each element changes. If background elements change as frequently as the foreground, then there is no background, only a bunch of foregrounds.
Failure condition: The aesthetic of certain styles (IDM, breakcore) is that everyone has a high rate of change— but in that case the "foreground" is established by other cues (usually the extremeness of the timbre), not by the rate of change.
6. Maximum-density method: how to make a very dense but not messy passage
Problem: After writing the first part, adding a second part feels like starting from zero; linear left-to-right advancement is daunting (p0265).
Method(note that it is a method for thecomposition stage, whereas the subtractive arrangement of mc-arrangement-arch §3 is for thearrangement stage;
the two are similar but act at different points) (p0266):
- Start from anymaximum-densitypoint in the middle of the song—all elements that could possibly occur simultaneouslyare in fact all occurring simultaneously (p0265)
- This step will pile upan unmusical mess, which is expected
- Subtract elements one by one, until reaching a level that is both high-density and musically logical → this is the song'sclimax point (p0265)
- Copythis peak passage, and test lower-density versions by muting/deleting
- When you find a combination you like, copy and paste it into a new passage
- Thus arrangement becomes "organizing these passages of different densities to match the song's ebb and flow" (p0266)
The value of this method: It makesall passages naturally share the same origin—low-density passages are subsets of high-density passages, so there will not be a situation where "each passage sounds like a different song."
Cost: Prone to over-layering. DeSantis himself warns that after using maximum density or subtractive arrangement, over-layering may occur unintentionally, requiring a great deal of erasing to restore musicality, and one must be especially careful in judgingwhether enough has been removed (p0288)。
7. Correspondence with ARR-SPEC fields
| Field | Which section of this skill |
|---|---|
arrangement.spectrum_budget |
§4 (each entry contains band / owner / competitors / resolution） |
arrangement.roster[].band |
§4.1 |
arrangement.depth_map |
§2 three depth layers |
arrangement.roster[].band four-part division |
§4.1 (see reference §1) |
Not filled in by this skill：roster[].entry/exit（→ mc-arrangement-arch）、
roster[].instrument（→ mc-orchestration）、mix_intent.ducking_map（→ mc-mix-intent）。
8. Must-pass checklist after writing
- Every element is labeledforeground/middle ground/background
- Every element can statewhich frequency band it occupies
- For each section, "highest, loudest, fastest, most drastically changing" can be listed;the protagonist occupies at least two
- Every frequency-band conflict has its solution written out asfrequency carving / staggered registers / alternating appearance, whichever one it is
- When the protagonist is vocals, the chorus's 2.5–4 kHz belongs to the vocals; the spacing between the two parts in the low range is no less than a fourth (§3.2, §4.2)
- The rate of change of background elements isclearly lower thanthe foreground
- If the maximum density method was used, checkedwhether enough has been removed(the cost in §6)
- Any conflict that can be resolved at the composition layerhas not been pushed to mixing
Appendix: Sources
- Dennis DeSantis《Making Music》(Ableton)——the backbone of §2, §3, §4.3, §4.4, §5, §6 (Foreground/Middle Ground/Background, Maximal Density, The Power of Erasing). Page numbers are from that book.
- The auditory basis for frequency-band budgeting is in the masking chapter of **Izhaki, Mixing Secrets**, organized in reference.md。
- Howard & Angus, Acoustics and Psychoacoustics(Acoustics and Psychoacoustics)—— §3's loudness/harmonics/singer's formant, §4.2's masking asymmetry and temporal masking, §4.5's loudness summation; the ruler for critical bands is in reference §1.1. Page numbers are from the Chinese translation.
- For the corresponding mixing-side techniques, see Izhaki, Mixing Secrets——only the parts on frequency-band budgeting and intent are taken, EQ/compression practice is a mixing matter and is not included in this package.