Push up the snare microphone alone and it sounds good. Push up the overheads alone and the kit sounds good. Bring them together and the snare goes thin and hollow, as if someone had put a hand over a speaker. Nothing is broken. Both microphones heard the same snare hit, one 5 centimetres away and one a metre above, so the same sound arrived at the mixer twice, a few thousandths of a second apart, and adding two copies of a sound that are slightly out of step cancels some pitches and doubles others.
That one fact, plus a handful of habits, is most of drum mixing. This piece is for the desk: tools you can move, then a working checklist, typical settings as ranges, common mistakes with fixes, and a real session to practise on. It skips what Mixing & Signal Processing (36/63) already teaches (what a bell or a ratio does) and goes straight to drums.
- Who takes it
- Part 1 of the Mixing & Mastering specialization, inside the Music Production specialization of Music; also for Sound Engineering students.
- What it is for
- Making a kit sound like one instrument that hits hard, with a kick and snare that cut through, an even picture of the cymbals and room, and no hollow, thin or flabby moments.
- Comes after
- Session Prep, Gain Staging & Mix Workflow (36/84) and Microphones & Recording (36/61), which introduced phase and the 3:1 rule.
- Leads to
- Finishing the Mix: Space, Automation & Translation (36/88). Next to it in Part 1 are Mixing Bass & Guitars (36/86) and Mixing Vocals (36/87).
- What this piece gives you
- Not a record-quality drum sound; that comes from many mixes. You will know what to check first (polarity and timing), when to gate or edit a tom, how parallel compression works, and you get a checklist to tick through. Easier first reads: Music Technology & Production (35/73) and How a Song Gets Made (09/07).
Phase and polarity: the same hit arriving twice
Two ideas get called "phase", and a drum mix needs both apart. Polarity is whether a signal pushes up or down: flipping it turns every positive sample negative, which is the button marked "ø" on a console or plug-in. Delay is the time a copy arrives late. Sound on Sound's "Phase demystified" makes the same distinction: total cancellation from inverting one of two identical waveforms is not the same thing as the partial cancellations from a delay.
Sound travels at about 343 metres per second in air at room temperature, so every metre of extra path is about 2.9 milliseconds. Put the overheads 1 m farther from the snare than the snare microphone and the snare arrives in them about 2.9 ms late. Add the two at equal level and the sum has notches, narrow dips in the spectrum spaced evenly along a linear frequency axis, the first at one over twice the delay: about 171 Hz here, then 514 Hz, 857 Hz and on up. A graph of those dips looks like the teeth of a comb, and the effect is called comb filtering. It is the hollow sound.
Try it below. Set a path difference (the extra distance one microphone's sound travels), lower the level of the second microphone, then flip its polarity. Two things to notice. A quieter second microphone makes shallower notches: that is why close microphones plus a distant overhead can work, because the overhead is much quieter on the snare than the snare microphone is. And a polarity flip is the right fix only when the two microphones were pushing opposite ways, as snare top and bottom do; on a pair that was already in step it creates the problem. The fix at the desk is to listen to both settings (and run the null test below), not to assume.
Sound isn't available in this reader; the curve and numbers still work.
Computed exactly: |1 + g·e−j2πfτ| (1 − … when flipped), τ the delay, g the second microphone's gain, air at 343 m/s. The sound button runs pink-spectrum noise through a real delay line and adds it to itself.
Without JavaScript: with a 100 cm path difference and equal levels the delay is 2.92 ms and the first notch is at 171 Hz (then 514 and 857 Hz). With the polarity flipped the first notch is at 343 Hz and the lows thin out toward 0 Hz.
Finding the fix. Either align by delay: nudge the closer microphone later (or the distant one earlier) by the extra travel time; most DAWs can shift a track by samples or milliseconds. Or run the null test from Sound On Sound's phase article: invert one microphone, adjust until the combined level is lowest (the copies are then as closely aligned as they can be), set the polarity back and keep the position. Then listen: the right answer is the fullest sound, not the neatest waveform. Steve Albini is quoted there as sometimes delaying ambient microphones by a few milliseconds, so deliberate delay is a tool as well as a repair.
Check three pairs every time: snare top and bottom (the bottom microphone faces the opposite way, so it usually needs its polarity flipped); kick inside and outside (flip one and keep what is fuller); and each close microphone against the overheads.
A kick is three jobs
A kick drum is asked to do three things at different frequencies: a deep sub you feel more than hear (roughly 40 to 70 Hz), a thump that tells you it is a drum (about 60 to 120 Hz), and a click, the beater on the head, around 2 to 5 kHz, which lets the kick be heard on a small speaker over a bass guitar. Engineers capture them with two or three microphones: one inside near the beater, one outside the resonant head, sometimes a large loudspeaker cone wired as a microphone for the very bottom. Each has its own timing and polarity, so the checks above apply here too.
Augmentation blends a sampled click or body under a recorded kick; replacement triggers a sample from each hit. People disagree about both. One camp says replacement erases the performance and the room, and a good drum should sound right to begin with. The other says a record is a made object, the sample is often a better recording of a drum, and fast metal kicks or a poorly tuned kit are practical reasons. Either way, align the sample to the original's transient and check polarity, or you have built the comb filter of the previous section yourself.
Sound isn't available in this reader; the picture and numbers still work.
All three layers are generated in the page (sine sweeps, filtered noise). Shares are fractions of the signal's energy below 100 Hz and above 1 kHz, measured with 2nd-order filters on the rendered samples.
Without JavaScript: at the starting positions (sub 60%, thump 100%, click 35%) about 86% of the energy is below 100 Hz and about 0.5% above 1 kHz. Pushing the click to 100% raises the share above 1 kHz to about 4%, and still the kick is mostly low end.
Notice what moves the numbers: the click is a tiny fraction of the energy yet it changes what you hear on a laptop speaker, which cannot reproduce 48 Hz at all. That is the argument for layering, and for bringing out the click with a few dB around 3 to 4 kHz, rather than for turning the whole kick up.
Toms, gates and bleed
Every microphone on a kit also hears the rest of the kit. That bleed is not always an enemy, but a tom that sits silent and rings along with every snare hit adds mud.
Three tools, gentlest to most surgical. A gate closes the channel below a threshold, so it opens only for the hit; it needs a threshold above the bleed and below the hits, with enough hold and release not to chop the ring. Editing (strip silence, in some DAWs) deletes the quiet stretches and lets you fade each clip, which is more precise. Clip gain or automation turns the tom down by a few dB when it is not playing. In a modern DAW, editing is usually the first thing to try, because you can see what you are doing.
Below, a synthetic tom mic hears three loud hits, three bleed bursts and one soft hit that is quieter than the bleed. A gate sees only level, so no threshold can keep that soft hit and drop the bleed.
Sound isn't available in this reader; the picture and counts still work.
A peak follower (30 ms release) feeds a gate that opens in 1 ms, holds, then closes over the release time, down 60 dB. An event "passes" if the gate is more than half open within 30 ms. Invented levels: loud toms −4.4 dBFS, soft tom −18.4, bleed −14.9 alone.
Without JavaScript: with the threshold at −12 dB the gate passes all three loud toms and none of the bleed, but also loses the soft tom. At −20 dB it passes all four tom hits and all three bleed bursts.
A soft hit is real drumming (a ghosted fill, a tentative roll), so a gate that removes it is a problem; edit around it. Many engineers leave some bleed on purpose, or use a gentle expander so the channel drops 6 to 10 dB instead of vanishing.
Overheads and rooms: the kit's picture
The close microphones are the drums seen through a magnifying glass; the overheads are the photograph of the whole kit. Many engineers start with the overheads (and rooms) and add the close microphones for weight; others start with kick and snare. Have a reason, and check the pair against each other once the close mics are in.
Look for three things. The snare should sit in the centre of the overheads; if it leans, the microphones are not equally far from it. The cymbals should not be harsh: overheads are where a high boost is easiest to overdo, and a high-pass around 100 to 250 Hz often cleans rumble the close mics already cover. The room microphones arrive late (5 m is about 15 ms), which is part of why they sound big; compressed hard and mixed quietly under the kit, they add size without blurring the attack.
Compression on the bus, and in parallel
With the sounds in place, send the drums to a drum bus (a group channel) and treat the kit as one sound. A gentle compressor there, as in Mixing & Signal Processing (36/63), turns the loudest hits down a little so the kit seems to belong together: "glue". Typical settings are low ratios (2:1 to 4:1), 1 to 3 dB of reduction, and a medium attack (10 to 30 ms) that lets the click of kick and snare through.
Parallel compression, often called "New York" compression, keeps the original drums untouched and sends a copy to a compressor set far harder (high ratio, fast attack, 8 to 15 dB of reduction), blended in underneath. The peaks barely change; the quiet parts (hats between beats, ghost notes, room decay) come up. The Match loudness box is on for a reason: a processed drum that is simply louder always sounds better at first, so turn the louder one down until they match, then listen again.
Sound isn't available in this reader; the meters and traces still work.
One bar at 100 bpm of invented synthetic drums (kick, snare, eighth-note hats, a little room tail) runs through a feed-forward compressor: peak detector, hard-knee gain computer, attack and release smoothing. "Crest" is peak minus RMS in dB.
Without JavaScript: with the gentle bus settings (threshold −12 dB, 2.5:1, 30 ms attack, 150 ms release, fully compressed) the loudest hits are turned down by up to 2.4 dB. With a hard parallel setting (threshold −30 dB, 8:1, 3 ms attack, 80 ms release, 35% blended, +10 dB make-up on the copy) the copy is turned down by up to 22 dB, and the short-term range between the loudest stretch and the lower-quartile stretch of the bar narrows from 30.7 to 27.7 dB at matched loudness.
Two relatives finish the bus. Saturation (tape, tube or transformer emulation, or a soft clipper) adds harmonics that make hits seem denser at the same meter level; judge it at matched level. Transient shaping raises or lowers the attack and sustain of a hit independently, by comparing a fast and a slow level follower.
Starting points, not rules
These are ranges from mixing books and magazines: places to start looking, not settings to copy. Drums, rooms and songs differ, and the 3 dB cut that rescues one snare ruins another. Sweep around them, let your ears overrule every number, and judge each change at the same loudness.
| What | Typical range | Why people look here |
|---|---|---|
| Kick: low emphasis | 50–100 Hz | The sub and thump area; boost a little, or cut if it fights the bass (see 36/86). |
| Kick: boxiness | 250–500 Hz, cut 2–6 dB | The "cardboard" resonance of a drum, often reduced. |
| Kick: click | 2–5 kHz | The beater; lets the kick be heard on small speakers. |
| Snare: body | 150–250 Hz | Weight of the shell. |
| Snare: crack | 2–5 kHz | Stick and wires; where it cuts through guitars. |
| Toms: body, attack | 80–200 Hz; 3–5 kHz | Pitch and weight; stick attack. |
| Overheads: high-pass | 100–250 Hz | Removes rumble the close mics cover; do it by ear. |
| Drum-bus compressor | 2:1–4:1, 1–3 dB of reduction, attack 10–30 ms | Glue without flattening the hits. |
| Parallel compressor | 4:1–10:1, 8–15 dB of reduction, blended low | Raises small details under the peaks. |
| Tom gate | threshold between bleed and hit; hold 20–80 ms; release 50–200 ms | Open for the hit, closed for the bleed, without chopping the ring. |
| Time alignment | delay (ms) = path difference (m) ÷ 0.343 | Arithmetic, not taste. |
Where these come from: the conventional teaching ranges in Bobby Owsinski's The Mixing Engineer's Handbook, Mike Senior's Mixing Secrets for the Small Studio, Roey Izhaki's Mixing Audio and Sound On Sound's mixing articles. The alignment line is physics (343 m/s).
Common mistakes and their fixes
1. EQ before polarity and timing. A thin snare is often two mics out of step, and EQ cannot fill a notch. Fix: flip, align, null-test, then EQ.
2. Judging the louder version. Parallel compression and saturation make the kit louder, which sounds "better". Fix: match loudness, A/B against bypass.
3. Gating everything by habit. Hard gates chop decays and soft hits. Fix: edit and clip-gain first; a gentle expander if you must.
4. Squashing the bus until the hits stop hitting. Fast attack and a heavy ratio flatten the stick. Fix: slower attack, under 3 dB, and put heavy work in the parallel copy.
5. A sample laid in unaligned. A kick sample can cancel the recorded kick's low end. Fix: line up the first peak, flip polarity, keep the fuller one.
6. Soloing the drums too long. A perfect solo drum sound may be wrong in the song. Fix: check in context every few moves, at a moderate level.
Genre and world differences
- Rock and indie. Overheads and rooms carry much of the sound; moderate bus compression; variation between takes is part of the sound.
- Metal. Very fast kick patterns need short, clear hits, so tight editing, gating and sample reinforcement are more common, and rooms are kept low so notes do not smear.
- Hip-hop and trap. Programmed or sampled, so mic phase does not arise; the work is balance, crisp hats and the 808 (a long, tuned, sine-like note whose length and pitch must fit the key; see 36/69 and 36/86).
- Pop. Layered with samples and processed hard, with the room added back artificially, shaped around the vocal.
- Reggaetón and dembow. The dembow pattern, a kick on each beat with a syncopated snare or clap figure (36/72), is usually programmed: keep kick and snare punchy and the pattern clear, and leave space for the vocal.
- Cumbia and percussion-led traditions. The "kit" may be hand percussion (congas, güiro, guacharaca) and a bass drum; timing between microphones still matters, but the cymbal-and-snare picture does not apply. Balance the ensemble by ear against the bass.
- Jazz, acoustic and folk. Few microphones, light or no compression, brushes and soft playing; overheads and a room carry the sound, and the dynamics are the point.
Practise on a real session
Mike Senior, author of Mixing Secrets for the Small Studio, runs a free library of multitrack sessions on his Cambridge Music Technology site (cambridge-mt.com), offered for mixing practice and education, with unprocessed tracks for each instrument. Its search lets you require or exclude instruments: look for a rock, pop or acoustic song with a full set of drum microphones and a bass guitar, or start from the library's suggestions for beginners. Read the terms beside each download before sharing a mix or using it beyond practice.
Solo each pair, run the polarity checks, and build the balance from the overheads. Save a version, note what changed, and compare against a commercial track in the same style at the same loudness.
The drum checklist
In the order you would do it at the desk. Ticks aren't saved.
DAW notes. The drum bus is a folder track or send in REAPER, an aux input in Pro Tools, a bus or summing stack in Logic, a group track in Ableton Live and a mixer insert in FL Studio. Polarity is a button or plug-in on each; for a few milliseconds of delay use a track delay, a sample offset or nudge.
What a graduate just sees
- Two microphones on one hit are two copies a few milliseconds apart; the sum has notches, and the first is at 1 ÷ (2 × delay).
- A thin snare is a phase problem until proven otherwise: flip, align, null-test, then EQ.
- A kick is a sub, a thump and a click, and the click is what survives a small speaker.
- A gate only sees level, so it cannot separate a soft tom from loud bleed; editing can.
- Parallel compression raises the small details without flattening the peaks; matched loudness is the only fair test.
- A drum bus does a little, several times, instead of a lot once: 2:1 to 4:1, 1 to 3 dB.
- A drum sample must be aligned to the original; polarity and timing apply to samples too.
- Every range here is a place to start looking; the song decides.
Where it shows up next
- Mixing Bass & Guitars (36/86): the kick and bass share the bottom of the mix, so the kick weight you chose here sets what the bass can use.
- Mixing Vocals (36/87): snare crack and vocal presence share 2 to 5 kHz and must be balanced against each other.
- Finishing the Mix (36/88): depth, width and mono checks apply to the overheads' stereo picture.
- Beatmaking, Sampling & Arrangement (36/69): programmed drums skip mic phase and bring their own problems, such as 808 length and tuning.
If you want to go deeper
- 📘 Textbook. Roey Izhaki, Mixing Audio: Concepts, Practices, and Tools (Routledge, 3rd edition, 2017); and Mike Senior, Mixing Secrets for the Small Studio (Routledge, 2nd edition 2018), written for people mixing at home.
- 📘 Short and practical. Bobby Owsinski, The Mixing Engineer's Handbook (Cengage), a quick guide with starting settings by instrument.
- 📘 Phase in depth. Sound On Sound, "Phase Demystified", free to read.
- 🌐 Free multitracks. The Cambridge Music Technology "Mixing Secrets" free multitrack library (cambridge-mt.com), for practice.
- 🌐 Free course. MIT OpenCourseWare 21M.380 Music and Technology: Recording Techniques and Audio Production (Fall 2016).
A drum mix is a set of microphones that all heard the same hit at slightly different times. Line them up first, decide which ones carry the weight and which the picture, tidy the bleed without losing the soft hits, then glue the whole thing together with a little compression on the bus and a hidden, hard-squashed copy underneath, and judge every move at the same loudness.
Try this: take any recording with drums, mute everything except the drums, and play the opening bar. Count the instruments you can hear (kick, snare, hat, toms, room) and ask which of them you would turn up first if the whole kit sounded thin. Then play it in mono and notice which of them gets weaker.
Sources. Sound On Sound, "Phase Demystified" (polarity versus delay, the 3:1 rule, aligning direct and miked signals, Albini on delaying ambient microphones); speed of sound at about 343 m/s in air at 20 °C; Robert Bristow-Johnson's Audio EQ Cookbook and Giannoulis, Massberg and Reiss (Journal of the Audio Engineering Society 60(6), 2012) for the compressor design in Mixing & Signal Processing (36/63) that the bus compressor above follows; Owsinski, Senior and Izhaki for the conventional ranges.
- 📘 Izhaki, Mixing Audio
- 📘 Senior, Mixing Secrets for the Small Studio
- 🌐 Sound On Sound: Phase Demystified