Car Audio Explained, Part 1: How Stereo Imaging Works in a Car, Reflections and Dispersion (Andy Wehmeyer's Lessons)

4 days ago

34 min read

This is Part 1 of 19 in the series Andy Wehmeyer on Car Audio, Explained for Beginners. Start here · Download the full guide as a PDF

Unofficial guide. Andy Wehmeyer's points are paraphrased from his public forum posts, blog, tuning guides, videos and podcasts, with a link to each source; the plain-English explanations were written for this guide. Not written, reviewed or endorsed by Andy Wehmeyer or Audiofrog.

1. How sound works in a car, and where to put the speakers

Andy Wehmeyer's placement advice starts from how stereo and hearing work, then follows the physics of a small, reflective cabin: which reflections matter, how each driver's size sets its dispersion, and where that puts each speaker. It ends with the special cases he cares most about: two good seats, center channels, rear speakers, upmixing, and "believability" over accuracy.

What this section is about: This section covers the basics behind every placement decision. It explains how two speakers create the illusion of sound coming from places where there is no speaker, and how our ears and brain work out where a sound comes from. It then covers what the hard, close surfaces of a car do to the sound and how a speaker's size decides how widely it spreads sound, and therefore where it should go. Later parts of the section deal with special cases such as two good seats, center channels and rear speakers.

1.1 How stereo imaging works, and why a car images for one seat

What this part is about: How two speakers can make a singer seem to stand in the empty space between them, and why that trick works properly only for a listener sitting exactly in the middle, which nobody in a car does. It also covers what delay (time alignment) can and cannot fix.

1.1 #1 · A stereo image needs left and right to reach you matched in level, frequency response and phase.

Stereo was built to place sound between two speakers for a centered listener; in a car nobody sits centered, so the near speaker arrives first and louder and pulls the image. Match the three conditions by measurement first; listen last. Sources: Audiofrog blog Dec 2020 (Time Alignment Part 1) · CMA Connected Oct 2021 @39:51 · DIYMA 2017-06-28

In plain English: A "stereo image" is the illusion that sounds come from places between and around two speakers, such as a singer who seems to stand in the middle of the dashboard where no speaker is. Three terms matter here. Level means loudness. Frequency response means how loud each pitch is, from deep bass to high treble. Phase means whether the ups and downs of two sound waves arrive in step. His point is that the illusion works only when the left and right sounds reach you equally loud, with the same tonal balance, and in step. Stereo was designed for a listener sitting centered between two speakers. In a car you sit off to one side, so the nearer speaker's sound arrives sooner and louder and pulls the image toward it. So in practice: match level, response and phase by measuring first, and use your ears last.

1.1 #2 · Level places sounds between the speakers: about 6 dB moves an image halfway to one side.

Left-only content stays in the left speaker and content common to both forms the phantom center. Time alignment fixes only the phase part, so a 20 dB left/right response mismatch (one 2017 case, at 80 Hz) still won't image. Sources: EDUCAR · DIYMA 2017-09-12

In plain English: When a recording is mixed, each instrument is placed left or right by making it louder in one channel than the other. A dB (decibel) is the unit for a difference in loudness. He says a difference of about 6 dB moves the image about halfway from the center toward one speaker. Sound that is only in the left channel stays at the left speaker. Sound that is equally in both channels forms a "phantom center", an image in the middle where no speaker exists. Time alignment means delaying the nearer speakers so both sides arrive together, and he says it fixes only the timing (phase) part. It cannot fix a big loudness mismatch: in one 2017 case, left and right differed by 20 dB at 80 Hz (a low bass note), and that would not image no matter how the delays were set.

1.1 #3 · Door speakers put left and right 180° out of phase near 250 Hz, right in the vocal range.

A ~28" path difference to door mids puts the first null at ~250 Hz and its odd multiples (750 Hz, 1.25 kHz…), where you hear speaker locations, not a center. A 1 ms error does the same from 500 Hz; those out-of-phase bands are what people mistake for "width". [RARE] Sources: Audiofrog blog Dec 2020 (Time Alignment Part 2) · DIYMA 2015-09-10 · DIYMA 2017-10-01

In plain English: "180° out of phase" means that when one speaker's wave pushes, the other's pulls, so at your ears they work against each other. In his example, the far door speaker is about 28 inches (about 70 cm) farther from you than the near one. At about 250 Hz, a pitch in the lower part of the human voice, that extra distance makes the two sides arrive exactly opposite. It happens again at 750 Hz, 1.25 kHz and so on. In those bands you hear the two speakers in their own places rather than one image in the center. He adds that a timing error of 1 ms (one thousandth of a second) does the same thing starting at 500 Hz. Many people like a "wide" sound and think it is good. He says those out-of-phase bands are exactly what people mistake for "width".

1.1 #4 · A phantom image is vector addition, and arrival time "rotates" it toward the near speaker.

That's why a balance control helps only at frequencies already in phase. Align both sides to one point between the ears: even two speakers both on your right image between them, because long wavelengths wrap around the head. [RARE] Sources: DIYMA 2015-10-19 · DIYMA 2015-10-19 (one side) · DIYMA 2015-10-19 (wrap)

In plain English: A vector is a quantity with both a size and a direction, like an arrow. His point is that the image you hear between two speakers comes from adding the left and right sounds together like arrows, so both loudness and timing count. A difference in arrival time "rotates" the result toward the nearer speaker, so the image swings that way. That is why the balance control, which only changes loudness, helps only at frequencies where the two sides already arrive in step. His fix is to set delays so both sides line up at one point between your ears. He adds that this works even if both speakers are on your right: you still hear an image between them, because low sounds have long waves that bend around your head and reach both ears.

1.1 #5 · Phase decides whether two speakers sound like one source or two.

Out of phase you hear both locations; in phase you hear one: midway for a left/right pair, at the higher-frequency driver for a crossover pair. That is how bass gets pulled forward to the tweeters, and why group delay is "less a curse than people think". [RARE] Sources: DIYMA 2021-12-02 · DIYMA 2021-09-07

In plain English: When two speakers play the same sound and it reaches you in phase (in step), your brain hears one source. Out of phase, you hear two sources in two places. For a left/right pair, the single source appears midway between them. A crossover is the filter that splits the music by pitch, sending low notes to a bigger speaker and high notes to a smaller one such as a tweeter. For a crossover pair that is in phase, he says the single source appears at the higher-frequency (smaller) driver. That is how bass from a low-mounted speaker can be "pulled forward" to where the tweeters are. Group delay is the small extra delay that crossovers and speakers give some frequencies, and many hobbyists worry about it. Because a well-designed crossover pair is heard at the upper driver anyway, he calls it "less a curse than people think".

1.1 #6 · Delay fixes one seat: delay the near side (about 2 ms for 28"), and the center lands midway between the speakers, not straight ahead of you.

One sample at 48 kHz is ~0.29", so quarter-inch steps are ample. Set delays to the midrange's distance, not the tweeter's or an average (phase matters most at 200–300 Hz; averaging "makes them BOTH wrong"). [RARE] Sources: Time Alignment Part 1 · DIYMA 2017-07-07 · DIYMA 2021-09-26

In plain English: A DSP (digital sound processor) can hold back the signal to a speaker by a tiny amount of time. You delay the nearer speaker so its sound arrives together with the far one. He says about 2 ms (two thousandths of a second) matches about 28 inches (about 70 cm) of extra distance. This only fixes one seat, and the center image then lands midway between the two speakers, not straight ahead of your face. A DSP works in tiny steps called samples. At 48 kHz (48,000 samples per second), one sample equals about 0.29 inch (about 7 mm) of distance, so quarter-inch steps are fine enough. He says to set the delays using the midrange speaker's distance, not the tweeter's and not an average of the two. His reason is that phase matters most around 200–300 Hz, and averaging "makes them BOTH wrong".

1.1 #7 · Delay does not move speakers.

The DSP-delay diagram used for ~30 years is "seriously misleading": delay changes only what happens between the speakers, never where a speaker or the stage edge appears. Test: play only the left speaker, change its delay with eyes closed, and nothing moves. [RARE] Sources: Time Alignment Part 1 · Time Alignment Part 2

In plain English: For about 30 years, a common diagram has explained time alignment by showing the delay "moving" the speakers so they all sit the same distance from you. He calls that diagram "seriously misleading". He says delay changes only what happens between the speakers, meaning where the phantom images land. It never changes where a speaker itself seems to be, or where the edge of the stage is. His test: play only the left speaker, close your eyes, and change its delay. Nothing moves. (With one speaker playing, there is no second sound for the delay to be early or late against.)

1.1 #8 · Image off-center after tape-measure delays? Fix it with level, never by re-tweaking delay.

Delay tweaks only add comb filtering. As the last step, turn one whole side up or down until the singer is centered, ignoring RTA numbers (the mic can't separate direct sound from reflections); for a left-hand-drive driver the right side usually ends up louder. It has worked in "hundreds of cars". [RARE] Sources: DIYMA 2017-07-07 · DIYMA 2018-05-22 · DIYMA 2022-03-19

In plain English: "Tape-measure delays" means measuring each speaker's distance to your head with a tape measure and setting the delays from those numbers. If the singer still isn't centered after that, he says not to keep fiddling with the delays. Each tweak only adds comb filtering, a row of dips in the sound caused when two copies of it arrive slightly apart. Instead, as the very last step, turn one whole side up or down until the singer is centered. An RTA (real-time analyzer) is a microphone and display that shows loudness at each frequency. Ignore its numbers here, he says, because the microphone can't tell the direct sound from the reflections. For someone driving a left-hand-drive car, the right side usually ends up louder. He says this has worked in "hundreds of cars".

1.1 #9 · Delays chosen to make drivers sum at a crossover also shift the left-right image.

Left-right position is set mainly by level and phase at the ears; for one-seat imaging use tape-measure delays, LR4 (Linkwitz-Riley 24 dB/oct) slopes and EQ, or an all-pass filter (a filter that shifts phase without changing level). Exception: with a real center and upmixer, crossover-optimizing delays are fine. [RARE] Sources: DIYMA 2015-10-19 · DIYMA 2015-07-24

In plain English: Some tuners set delays not from distance but to make two drivers on the same side, such as a woofer and a tweeter, blend smoothly ("sum") at their crossover point. He says those delays also shift where sounds land between left and right, because left-right position is set mainly by level and phase at your ears. For imaging in one seat, his recipe is tape-measure delays, LR4 crossover slopes and EQ, or an all-pass filter. LR4 (Linkwitz-Riley, 24 dB per octave) is a standard crossover filter shape. EQ (equalization) means turning particular frequencies up or down. An all-pass filter shifts phase without changing loudness. The exception is a car with a real center speaker and an upmixer, a processor that creates a center signal from a two-channel recording. There, he says, delays chosen to optimize the crossover are fine.

1.1 #10 · A correct center sounds "too loud" or "mono" at first, because we've learned to like fake width.

Small path errors keep fundamentals centered but throw harmonics to the speakers, like a boombox "stereo wide" switch. Of two famous producers he demoed a strong-center car to, one loved it and one called it "mono"; listeners usually come round to the stable center. [RARE] Sources: DIYMA MS-8 FAQ #10 · CMA Connected Oct 2021 @41:19 · DIYMA 2013-01-24

In plain English: "Mono" means a single channel, with everything coming from one place. He says a properly centered image often sounds "too loud" in the middle, or "mono", the first time you hear it, because we have learned to like fake width. A musical note has a fundamental (its main pitch) and harmonics (higher overtones). Small errors in path length keep the low fundamentals centered but throw the higher harmonics out toward the speakers. He compares this to the "stereo wide" switch on a boombox. He demonstrated a car with a strong center to two famous record producers: one loved it, and the other called it "mono". In his experience, listeners usually come round to preferring the stable center.

1.1 #11 · Ultra-precise high-frequency alignment is right at only one point in space.

Move the mic an inch and drivers are out of phase again; 1 mm delay resolution is "nonsense" and laser-aligned acoustic centers are "garbage" where heads move. Fussing over mid-to-tweeter crossover phase is "ridiculous"; sub/midbass and midbass/mid just need to be far from 180°. [RARE] Sources: OFCAP Ep. 19 @2:56:00 · DIYMA 2015-05-29 · DIYMA 2021-07-12

In plain English: High-pitched sounds have very short waves, about an inch or less at the top of the treble range. So timing that is perfect at one exact spot is wrong a little way off. He says that if you move the microphone an inch, the drivers are out of phase again. That is why he calls 1 mm delay resolution "nonsense". He calls aligning speakers' acoustic centers with a laser "garbage" in a car, where heads move. He calls fussing over the phase at the midrange-to-tweeter crossover "ridiculous". For the lower crossovers, subwoofer-to-midbass and midbass-to-midrange, he says the two drivers just need to be far from 180° (fully opposite), not perfectly matched.

1.1 #12 · The goal is one apparent source above the dash, not a "rainbow" of bands in different places, and the cure is a good crossover.

What counts is arrival at your ears, not at the baffle (a mid 6.8" farther than the tweeter is 180° out at 1 kHz). Right phase at sub–midbass puts bass in front; right phase at midbass–mid–tweeter holds vocals at tweeter height. [RARE] Sources: Audiofrog blog Dec 2020 (Time Alignment Part Three) · DIYMA 2021-11-03 · DIYMA 2015-05-29

In plain English: He wants all the music to seem to come from one source, up above the dashboard. A "rainbow" is when different pitch bands seem to come from different places: bass low in the doors, voices somewhere else, cymbals at the tweeters. His cure is a good crossover. A baffle is the panel a speaker is mounted in. He says what counts is when each driver's sound arrives at your ears, not where the speakers sit on the baffle. For example, a midrange 6.8 inches (about 17 cm) farther away than the tweeter arrives 180° out of phase (exactly opposite) at 1 kHz. When the subwoofer-to-midbass crossover is in the right phase, the bass seems to come from in front. When midbass, midrange and tweeter are in the right phase, voices stay up at tweeter height.

1.1 #13 · In a car, left plus right sums to +6 dB only in the midbass; higher up, about +3 dB is normal.

The dense reflection comb averages to ~90° between sides, so leave anything within ±90° alone (true cancellation is 135–225°). Every peak or dip is a phase shift, so mismatched peaks mean mismatched phase. If the center is solid and pans move smoothly, ignore a small dip. [RARE] Sources: DIYMA 2021-08-26 · DIYMA 2021-08-27 · DIYMA 2022-03-24

In plain English: Two equal sounds that arrive exactly in step add up to 6 dB louder than one alone. If they are 90° apart, they add up to about 3 dB louder. He says that in a car, left plus right reaches +6 dB only in the midbass. Higher up, about +3 dB is normal. The reason is the dense pattern of reflections (the "comb"), which averages out to about 90° of difference between the two sides. So if the two sides are within ±90° of each other, leave them alone. Real cancellation happens only at 135–225°, around fully opposite. He says every peak or dip in a response is a phase shift, so peaks that don't match between left and right mean the phase doesn't match either. So in practice: if the center image is solid and sounds move smoothly across when panned, ignore a small dip.

1.1 #14 · Learn what imaging is with a $5 pair of identical speakers.

They needn't be good, only the same: set a thrift-store pair in an equilateral triangle, sit in the middle, then flip one's polarity to hear out-of-phase. That, not "knowing live music", is a tuner's first reference; a car needs extra steps only because you can't sit in the middle. [RARE] Sources: Absolute Electronix 2023 @41:56 · Absolute Electronix 2023 @46:15

In plain English: He suggests a cheap experiment to learn what a stereo image is. Buy any pair of identical speakers, for example from a thrift store for about $5. They don't need to be good, only the same. Set them up with your listening spot so that the two speakers and your head form an equilateral triangle (all three sides the same length), and sit centered. Listen to how voices appear in the middle. Then flip the polarity of one speaker (swap its + and − wires) to hear what out-of-phase sounds like. He says this, not "knowing live music", is a tuner's first reference. A car needs extra tuning steps only because you can't sit in the middle.

1.2 How we hear direction (the psychoacoustics he builds on)

What this part is about: How our ears and brain work out where a sound comes from, and which of those methods work at which pitches. He builds his placement and tuning advice on these facts, so this part explains why some errors matter a lot and others hardly at all.

1.2 #1 · His localization map by band.

Below ~60–80 Hz neither time nor level cues work; ~80–800 Hz we use arrival-time (phase) differences between the ears; ~800 Hz–1.6 kHz is a weak zone ("we just basically suck at it"); above that, level only. So arrival matters most for midbass and mids; phase hearing fades from ~1 kHz and is gone by ~6 kHz. [RARE] Sources: DIYMA 2013-11-20 · DIYMA 2016-04-27 · DIYMA 2016-03-17

In plain English: Localization means working out where a sound is coming from. He splits the range of pitches into zones. Below about 60–80 Hz (deep bass), neither timing nor loudness differences between your two ears help, so you can't tell where the sound is. From about 80 Hz to 800 Hz, we use differences in arrival time (phase) between the two ears. From about 800 Hz to 1.6 kHz is a weak zone where, in his words, "we just basically suck at it". Above that, we use only level (loudness) differences between the ears. So arrival timing matters most for the midbass and midrange. Our hearing of phase fades from about 1 kHz and is gone by about 6 kHz.

1.2 #2 · Only the lowest comb-filter notch really matters; head movement smears the rest.

A comb filter is the row of dips two delayed copies of a sound create. The first notch is widest; higher ones are narrow, and narrow dips are far harder to hear than peaks. So judge with band-limited pink noise or music, never sine waves. [RARE] Sources: DIYMA 2013-11-20 · DIYMA 2017-07-12 · DIYMA MS-8 FAQ #21

In plain English: The point defines a comb filter. It is called that because, on a graph, the row of dips looks like the teeth of a comb. He says the first, lowest-pitched dip is the widest, and only that one really matters. The higher dips are narrow, and as your head moves they shift and smear out. Narrow dips are also far harder to hear than peaks. So in practice: judge the sound by listening to band-limited pink noise or to music, never to sine waves. Pink noise is a steady rushing sound containing all pitches; "band-limited" means it is cut down to one range of pitches. A sine wave is a single pure tone.

1.2 #3 · Above ~1 kHz the in-car response swings 25 dB inside a head-sized circle, and you don't hear it.

Five mic positions in a 7" circle agree below ~200 Hz, diverge from 200 Hz to 1 kHz and swing ≥25 dB above it. The success of kick-panel cars suggests path errors above ~1 kHz matter much less. [RARE] Sources: Audiofrog blog Jan 2021 (Measuring Frequency Response) · Time Alignment Part 2

In plain English: He measured with a microphone at five spots in a circle 7 inches (about 18 cm) across, about the size of a head. Below about 200 Hz the five measurements agreed. From 200 Hz to 1 kHz they started to differ. Above 1 kHz they swung by 25 dB or more, a huge difference, between spots only inches apart. Yet we don't hear the sound changing like that as we move our heads. He also points to cars with speakers in the kick panels (the panels low down in front of the doors, by your feet). Kick-panel mounting makes the left and right distances more nearly equal, which pushes the remaining errors up to higher pitches. He takes the success of such cars as a sign that path-length errors above about 1 kHz matter much less.

1.2 #4 · Precedence: sounds arriving within ~30 ms fuse and are heard at the first arrival.

That's why a rear sub can image in front (boost it enough and you hear it behind you), and why every in-car reflection, all within ~20 ms, fuses with the direct sound. [RARE] Sources: DIYMA 2014-03-22 · DIYMA 2009-12-06

In plain English: The precedence effect (also called the Haas effect) is how the brain handles echoes. When the same sound reaches you more than once within a short time, you hear a single sound, placed where the first arrival came from. He gives that window as about 30 ms (thirty thousandths of a second). He says this is why a subwoofer in the back of the car can seem to play from the front. But boost it enough and you will hear it behind you. It is also why every reflection inside a car, all of which arrive within about 20 ms, blends with the direct sound instead of sounding like an echo.

1.2 #5 · "HRTF EQ" can't raise the stage.

HRTF (head-related transfer function) is the head-and-ear filtering that cues direction; it works only with a moving head or source. EQ a "height" bump (e.g. 3.5 kHz) and the image rises only while you're adjusting, then settles back, leaving a bump. Height comes from where speakers are (1.7 #7). [RARE] [Era: unchanged 2008–2021] Sources: DIYMA 2008-07-27 · DIYMA 2016-04-27 · DIYMA 2021-11-11

In plain English: The point defines HRTF. The folds of your outer ear change the tone of a sound slightly depending on its direction, and that is one way we sense height. Some tuners try to fake a higher stage with EQ, for example by adding a bump around 3.5 kHz. He says the image rises only while you are adjusting the EQ. Then it settles back down, and all you are left with is a bump in the response. These cues, he says, only work when your head or the sound source is moving. Real height comes from where the speakers actually are (see point 1.7 #7). The tag notes that he held this view unchanged from 2008 to 2021.

1.2 #6 · We hear height worst, yet car audio prioritizes stage height.

We localize front sounds well, rear sounds poorly and height worst. Hanging mids in A-pillars optimizes what we hear worst while distorting what we hear best, like painting a tonally perfect picture for a color-blind viewer with every shape wrong. [RARE] Sources: DIYMA 2016-04-27 · DIYMA 2013-11-20

In plain English: He says we locate sounds in front of us well, sounds behind us poorly, and height worst of all. Many people believe the most important thing is a high soundstage, and they mount midrange speakers in the A-pillars, the posts on either side of the windshield. He says this optimizes the thing we hear worst while distorting what we hear best. His comparison is painting a picture with perfect colors for a viewer who is color-blind, while getting every shape wrong. You have perfected the part they can't see and spoiled the part they can.

1.2 #7 · Head movement matters more than individual ear shape.

"3-D" headphones fail mainly because they ignore head movement ("We're all basically shaped the same"); at Harman, open-back headphones with a head tracker and DSP were indistinguishable from a real speaker. [RARE] Sources: DIYMA 2016-05-22 · DIYMA 2016-06-23

In plain English: "3-D" headphones use processing to try to make sound seem to come from outside your head. Many people believe they fail because everyone's ears are shaped differently. He says they fail mainly because they ignore head movement: "We're all basically shaped the same." He worked at Harman, the company behind JBL. There, open-back headphones combined with a head tracker (a sensor that follows how you turn your head) and DSP processing could not be told apart from a real speaker.

1.2 #8 · What you see moves the center image.

Moving an iPod controller from left of center to the middle of his dash improved his IASCA center-image scores; ears work with eyes, and early reflections confirm what the eyes see. [RARE] Sources: DIYMA 2013-05-03 · OFCAP Ep. 38 @39:05

In plain English: IASCA is a car-audio sound-quality competition where judges score things like how well the center image is placed. He says that when he moved an iPod controller from the left of center to the middle of his dash, his center-image scores improved. His explanation is that our ears work together with our eyes, and the early reflections confirm what the eyes see. In other words, the brain uses what it sees as well as what it hears to decide where a sound is.

1.3 Reflections: what makes a car sound like a car

What this part is about: A car is a small box of glass, plastic and fabric, and sound bounces off all of it. This part explains what those bounces (reflections) do to the sound you hear and the image you see, why EQ and delay can't remove them, and how speaker placement can keep them from doing harm.

1.3 #1 · A car has no reverberant field, so you mostly hear the speaker's power response.

Every reflection arrives within ~20 ms and fuses with the direct sound into roughly the average of the speaker's output over all angles (its "power response"); there are no late reflections. So a speaker swap often changes little, and EQ is the main lever. [RARE] Sources: DIYMA 2009-12-06 · DIYMA 2017-10-28 · DIYMA 2017-08-27

In plain English: In a large room or concert hall, sound keeps bouncing around for a while after the music stops. That is reverberation, and it includes late reflections that arrive well after the direct sound. He says a car is far too small for this. Every reflection arrives within about 20 ms (twenty thousandths of a second) and blends with the direct sound. What you hear is roughly the average of everything the speaker sends out in all directions, called its "power response". So swapping one speaker for another often changes less than people expect, and EQ (turning particular frequencies up or down) is the main tool for shaping the sound.

1.3 #2 · Cross-car reflections near your ears are what make a car "sound like a car".

Right-channel sound bouncing off surfaces by your left ear (and vice versa) is acoustic crosstalk that narrows the image; worst are the far speaker off the near side glass and both speakers off the windshield. Home listeners call car systems "mono" for this; neither horns nor EQ remove it. [RARE] Sources: DIYMA 2010-01-03 · DIYMA 2011-11-17 · CMA Networks 2023 @38:04

In plain English: Crosstalk means one channel leaking into the other. Here, sound from the right speaker bounces off surfaces near your left ear, and the reverse, so each ear gets some of the wrong side's sound. He says this narrows the stereo image. The worst cases are the far speaker's sound bouncing off the side window next to you, and both speakers' sound bouncing off the windshield. He says this is why people used to home stereo call car systems "mono", meaning everything seems to come from one place. Horn speakers (which steer sound in a controlled direction) don't remove it, and neither does EQ.

1.3 #3 · Reflections show up as image spread and wandering, not as echoes.

Each reflection acts as another source; if its response differs from the direct sound, images land in different places per frequency. Below 1 kHz an out-of-phase reflection can weaken one side so pans stick near the speakers (a classic stuck drumbeat in pan tests); an A-pillar mid acts like four mids. [RARE] Sources: DIYMA 2013-11-20 · DIYMA 2022-02-23 · DIYMA 2020-10-07

In plain English: Many people expect reflections to cause echoes. He says that in a car they show up instead as an image that spreads out or wanders. Each reflection acts like another speaker. If the reflected sound has a different tonal balance from the direct sound, images land in different places at different pitches. Below 1 kHz, a reflection that arrives out of phase can weaken one side. Then, when a test track pans a sound from one side to the other, it sticks near the speakers instead of moving smoothly. His classic example is a drumbeat that gets stuck in a pan test. Since each strong reflection acts as another source, he says a midrange in an A-pillar (the post beside the windshield) acts like four midranges.

1.3 #4 · No EQ or delay can fix a reflection comb.

The processor sits before the speaker, so any EQ or delay applies to the reflection too ("that's the law"), and reflections don't change delay settings. Box peaks are EQ-able, reflection dips aren't. If a dip can't be moved, copy it into the other side: the image improves though the dip stays. [RARE] Sources: Time Alignment Part 2 · DIYMA 2021-03-26 · DIYMA 2021-11-06

In plain English: The processor (DSP) sits before the speaker in the signal chain. Any EQ or delay you apply changes what the speaker plays. The reflection is just a bounced copy of that sound, so it gets exactly the same change: "that's the law". The reflection still arrives later by the same amount, so the dip it causes stays. He says peaks caused by a speaker's box can be fixed with EQ, but dips caused by reflections can't. His workaround applies when you can't move the dip by moving the speaker. Copy the same dip into the other side with EQ. The dip remains, but the image improves because left and right now match.

1.3 #5 · Predict reflection dips with a tape measure before you build.

A dip falls where the reflected path is half a wavelength longer than the direct one: a mid 6" from a surface (~1 ft extra) dips near 500 Hz; a 2-ft path near 250 Hz; a 600 Hz null means ~11" extra. Never put a null inside a driver's passband: push mids into corners (null above their range) and move tweeters away (null below theirs). [RARE] Sources: Audiofrog blog Jun 2021 (Dispersion and reflection) · DIYMA 2020-10-07 · DIYMA 2021-11-06

In plain English: A wavelength is the length of one full sound wave: low notes have long waves and high notes short ones. He says a dip appears at the pitch where the bounced path is half a wavelength longer than the direct path. At that pitch the reflection arrives exactly opposite to the direct sound. His examples: a midrange 6 inches (15 cm) from a surface gets about 1 foot (30 cm) of extra path and dips near 500 Hz. A 2-foot (60 cm) extra path dips near 250 Hz. A dip at 600 Hz means about 11 inches (28 cm) of extra path. His rule is never to put a dip inside the range a speaker plays. So in practice: push midranges into corners, so the dip lands above their range, and move tweeters away from surfaces, so the dip lands below theirs.

1.3 #6 · Mount a driver in a surface, not in front of it.

A surface used as the baffle stops being a troublesome reflector: in-door mounting removes the door panel as a null source, in-dash removes the dash, and deep in a corner the return paths are so short you get only gain. A pod standing off a baffle adds exactly the dip you're avoiding. [RARE] Sources: DIYMA 2011-04-15 · DIYMA 2020-10-07

In plain English: A baffle is the panel a speaker is mounted in. He says that if a surface itself is the baffle, with the speaker set into it, it stops being a reflector that causes trouble. Mounting in the door removes the door panel as a cause of dips. Mounting in the dash removes the dash. Deep in a corner, the bounced paths are so short that you only get extra loudness (gain), not dips. A pod is a housing that holds a speaker out in front of a surface. He says a pod standing off a baffle adds exactly the kind of dip you were trying to avoid.

1.3 #7 · Distance is the lever: reflections drop 6 dB per doubling, and one next to your ear hurts as much as one next to the speaker.

Carpet absorbs the top octaves and glass reflects them; the driver's own window causes as much response error as the windshield, and no pattern control changes that. [RARE] Sources: DIYMA 2012-09-06 · DIYMA 2016-04-23

In plain English: He says a reflection gets about 6 dB quieter each time the distance it travels doubles, so distance is the main way to weaken it. A surface right beside your ear does as much harm as one right beside the speaker. Carpet absorbs the top octaves, the highest treble, while glass reflects them. His example: the window next to the person driving, beside their ear, causes as much error in the frequency response as the windshield next to the speaker. No pattern control, meaning speaker designs that aim sound in a set direction, changes that.

1.3 #8 · Dash mats, foam and films don't control midrange reflections.

Absorbers start working at about a quarter wavelength, so ~200 Hz would need ~1.5 ft of absorber; a dash mat "does NOTHING" in the lower midrange and foam just stops rattles. Real absorbers are membranes (what mass-loaded vinyl does in walls) or fiber stuffing. [RARE] Sources: DIYMA 2013-09-14 · DIYMA 2022-01-14

In plain English: An absorber is soft material that soaks up sound instead of bouncing it back. He says an absorber starts working only once it is about a quarter of a wavelength thick. At 200 Hz that means about 1.5 feet (about 45 cm) of material. So a dash mat "does NOTHING" in the lower midrange, and foam just stops rattles. Real absorbers, he says, are membranes or fiber stuffing. A membrane absorber is a heavy, flexible sheet, which is what mass-loaded vinyl does inside walls.

1.3 #9 · A 250 Hz "resonance" is usually a reflection peak to EQ, not a panel to deaden.

A true resonance is mass and spring trading energy; a lone peak with dips either side is usually a cabin dimension reinforcing that frequency. Play a 250 Hz tone: if nothing buzzes, EQ it down. [RARE] Sources: DIYMA 2021-09-07

In plain English: A true resonance, he says, is mass and spring trading energy back and forth, like a panel vibrating. Many people hear a peak around 250 Hz, decide a panel is resonating, and add sound deadening (material stuck to panels to stop vibration). He says a single peak with dips on either side is usually a cabin dimension reinforcing that frequency: sound arriving from two distances and adding up. His test: play a 250 Hz tone. If nothing buzzes or rattles, EQ the peak down.

1.3 #10 · Two strategies for reflections, and he picks "spread the chaos around".

Either minimize reflections with horns or big diaphragms (pattern control), or use each driver only where its dispersion is wide, so reflections share the direct sound's response and one EQ fixes both. He prefers the second (drivers exist, mounting is easier, EQ works); neither fixes reflections near the ears. [RARE] Sources: DIYMA 2010-01-03 · DIYMA 2013-08-10

In plain English: He describes two ways to deal with reflections. The first is to minimize them using pattern control, meaning speakers that send sound only in a chosen direction, such as horns or large diaphragms (big speakers beam more). The second is to use each speaker only in the range where it still spreads sound widely. Then the reflections have the same tonal balance as the direct sound, and one EQ setting fixes both. He prefers the second, which he calls "spread the chaos around", because suitable speakers exist, mounting is easier, and EQ works. He notes that neither approach fixes reflections near your ears.

1.3 #11 · Kill the reflections and the car sounds like headphones: aim for a pleasing room, not no room.

The least reflective car he heard was "very accurate" but "missed the sense of space"; stereo in an anechoic chamber was horrible. Inverting the car's impulse response works only for a head that never moves, but it is a good base for adding a virtual room. [RARE] Sources: DIYMA 2013-09-15 · DIYMA 2017-07-05 · DIYMA 2017-07-04

In plain English: Many people think the ideal car would have no reflections at all. He says the least reflective car he has heard sounded "very accurate" but "missed the sense of space", much like headphones. He has also heard stereo in an anechoic chamber, a test room with no echoes at all, and says it was horrible. So the aim is a pleasing room, not no room. An impulse response is a measurement of how the car changes a single sharp click. Some people propose "inverting" it with a processor to cancel the reflections. He says that works only for a head that never moves, but it is a good starting point for adding a virtual room, an artificial sense of space created by the processor.

1.3 #12 · Windshield reflections are not ambience: a car has "spread", not early and late reflections.

They arrive far too early to suggest a stage or room, so they just unfocus the image and leave a jagged comb; cabin reflections can't add depth either. There is "no ambience in a car unless you create it" with rear speakers and delay. [RARE] Sources: DIYMA 2013-12-15 · DIYMA MS-8 FAQ p3 · DIYMA 2022-02-23

In plain English: In a concert hall, early reflections and late reflections arriving at different times give a sense of the room and of space. That sense of space is called ambience. He says a car's windshield reflections arrive far too early to suggest a stage or a room. They just blur the image, which he calls "spread", and leave a jagged comb of dips. Reflections inside the cabin can't add depth either. There is "no ambience in a car unless you create it", which you do with rear speakers and delay.

1.4 Dispersion and directivity: the property that decides placement

What this part is about: Dispersion means how widely a speaker spreads its sound. This part explains why that depends mainly on the speaker's size compared with the length of the sound waves it plays. He then uses that to decide which size of speaker should play which range, where the crossovers should go, and why many cars need three speakers per side rather than two.

1.4 #1 · Speakers are floodlights, not flashlights: dispersion is set by diameter versus wavelength.

Every driver radiates everywhere at wavelengths long relative to its size and beams at short ones. Cone material, phase plugs, ring radiators and "diffusers" barely change it: two 1" tweeters differ off-axis only above ~3 kHz, and a 15 kHz dip at 60° doesn't matter in a car. [RARE] Sources: DIYMA 2010-07-11 · DIYMA 2014-06-01 · DIYMA 2021-11-23

In plain English: A floodlight lights up everything around it; a flashlight throws a narrow beam. He says every speaker sends sound in all directions at pitches whose waves are long compared with its size, and "beams" (sends sound mostly straight ahead) at pitches whose waves are short. So the speaker's diameter compared with the wavelength sets its dispersion. Many people believe cone material, phase plugs (the bullet shape in the middle of some speakers), ring-radiator tweeters or "diffusers" widen a speaker's spread a lot. He says they barely change it. "Off-axis" means heard from an angle rather than straight in front. Two different 1-inch tweeters, he says, differ off-axis only above about 3 kHz. A dip at 15 kHz heard 60° off to the side doesn't matter in a car.

1.4 #2 · His dispersion rule in numbers: driver circumference ÷ wavelength.

Below 1: same response on- and off-axis; ~2: narrowing starts; ~3: matches only to ~30°; ~4: ~15°; ~10: on-axis only. Working figures: a 6" is omnidirectional below ~300–500 Hz, a 3" below ~3 kHz, a 4" below ~6 kHz; a 1" tweeter beams only above ~10 kHz. [RARE] Sources: DIYMA 2014-03-20 · Audiofrog blog Jun 2021 · DIYMA 2013-12-15

In plain English: Circumference is the distance around the speaker's cone, about 3.14 times its diameter. His rule divides that by the wavelength of the sound being played. Below 1, the speaker sounds the same straight in front and off to the side. Around 2, the spread starts to narrow. Around 3, the sound matches only within about 30° of straight ahead. Around 4, only within about 15°. Around 10, it is good only straight ahead. His working figures follow from this. A 6-inch speaker spreads evenly in all directions (omnidirectional) below about 300–500 Hz. A 3-inch does so below about 3 kHz, and a 4-inch below about 6 kHz. A 1-inch tweeter beams only above about 10 kHz.

1.4 #3 · Off-axis response matters more in a car than in a room.

Surfaces are closer, so reflections are louder and you hear on- and off-axis output combined; Harman's room-directivity research applies even more strongly in cars. [RARE] Sources: DIYMA 2009-07-10 · DIYMA 2015-05-07

In plain English: "On-axis" is the sound straight in front of a speaker; "off-axis" is the sound it sends out at an angle. In a car, the surfaces are much closer to the speakers than the walls of a room are. So, he says, the reflections are louder, and what you hear is the on-axis and off-axis output combined. Harman, the company he worked for, is known for research on how a speaker's off-axis sound affects how it sounds in a room. He says that research applies even more strongly in cars.

1.4 #4 · Big woofer + small tweeter = an off-axis hole at the crossover = far-side bias.

A 6.5–8" with a 1" dome can be flat on-axis yet leave "a big hole" off-axis; aiming only moves the hole between direct and reflected sound. Aimed on-axis, leave the dip; flat in the door, EQing it makes the passenger seat worse. A 6" + 1" is "a bad deal"; a 3-way fixes it. [RARE] Sources: Crutchfield · DIYMA 2013-03-10 · DIYMA 2021-11-30

In plain English: A 6.5–8 inch woofer is already narrowing its spread by the frequency where it hands over to a 1-inch dome tweeter. So the pair can measure flat straight in front but leave "a big hole" in the sound sent off to the sides around the crossover. Far-side bias means the image leaning toward the speaker on the far side of the car. He says aiming the speakers only moves the hole between the direct sound and the reflected sound. If the speakers are aimed at you, leave the dip alone. If they sit flat in the door, you can EQ the dip, but that makes the passenger seat worse. His reason: the near speaker needs more boost, which is the opposite of what the passenger seat needs. He calls a 6-inch plus 1-inch pairing "a bad deal" and says a 3-way (adding a midrange) fixes it.

1.4 #5 · The target is a smoothly narrowing pattern, not identical on- and off-axis response.

Directivity index means how much more a speaker sends forward than its average; it should rise gradually, with no holes or peaks near crossovers (off-axis equal to a flat on-axis sounds too bright). Home crossover rules apply in cars because they make reflected sound resemble direct sound. [RARE] Sources: DIYMA 2009-07-10 · DIYMA 2013-09-17 · DIYMA 2021-09-02

In plain English: Many people believe a speaker should sound exactly the same straight in front and off to the side. He says the target is different: a pattern that narrows smoothly as the pitch rises. The point defines directivity index. He says it should rise gradually with frequency, with no holes or peaks near the crossovers. If the off-axis sound were equal to a flat on-axis sound, the result would sound too bright. The usual rules for designing home-speaker crossovers apply in cars too, he says, because they make the reflected sound resemble the direct sound. He adds that this makes EQ easier, because the measuring microphone can't tell direct sound from reflected sound.

1.4 #6 · Smooth directivity needs a waveguide or a 3-way, and cars need the 3-way.

Studio monitors can be 2-ways because the engineer sits ~3 ft away with walls ~10 ft off; in a car the reflectors touch the speakers. A waveguide big enough to match a 6" (his home GB60/GB15 one is ~5" effective) won't fit a car. [RARE] Sources: DIYMA 2022-01-05 · DIYMA 2008-12-02 · DIYMA 2015-12-21

In plain English: A waveguide is a shallow horn shape around a tweeter that narrows its spread at the low end, so it can match a bigger woofer. A 3-way splits the sound among three speakers per side (woofer, midrange and tweeter) instead of two. He says studio monitors can be 2-ways because the recording engineer sits about 3 feet (about 1 m) away, with walls about 10 feet (about 3 m) off. In a car, the reflecting surfaces are right against the speakers. A waveguide big enough to match a 6-inch woofer won't fit in a car. For scale, the one he uses at home with the GB60 woofer and GB15 tweeter (Audiofrog products) is about 5 inches effective. So, in a car, he says the answer is the 3-way.

1.4 #7 · A waveguide should narrow the tweeter only at the crossover.

Narrowing its low end to meet the mid is fine if the curve stays smooth; narrowing its top end to match a beaming woofer leaves a power-response hole that "NEVER sounds good". The JBL 660GTi waveguide was designed for ~35° off-axis door mounting. [RARE] Sources: DIYMA 2013-03-22 · DIYMA 2014-04-12

In plain English: He says a waveguide should narrow the tweeter's spread only around the crossover. Narrowing its low end so it meets the midrange is fine, as long as the change stays smooth. Narrowing its top end, to match a woofer that is already beaming, leaves a hole in the power response (the total sound sent in all directions, which is mostly what you hear in a car; see 1.3 #1). He says that "NEVER sounds good". He notes that the JBL 660GTi waveguide was designed for mounting in a door at about 35° off-axis.

1.4 #8 · Design crossovers for the angle each driver is heard at, in the car.

His 4"-plus-tweeter passive (~2.7 kHz) was designed from the mid measured ~30° off-axis; an anechoic design moved into a car is not the same thing. Same plane matters more than same axis. [RARE] Sources: Time Alignment Part Three · DIYMA 2014-06-03

In plain English: A passive crossover is a small circuit of capacitors and coils that splits the music between speakers. He designed his crossover for a 4-inch midrange plus tweeter (crossing at about 2.7 kHz) using the midrange measured about 30° off-axis, the angle at which it is heard in the car. He says a crossover designed in an anechoic chamber (a test room with no reflections) and then moved into a car is not the same thing. So in practice: design for the angle at which each speaker is actually heard. He adds that mounting the two drivers in the same plane matters more than putting them on the same axis. Same plane means their faces sit on the same surface; same axis means they are lined up along one line.

1.4 #9 · Choose crossover points by dispersion and driver safety, not "as low as possible".

First keep the smaller driver safe (tweeter at ~2× its resonance), then hand off where the larger one still disperses widely, below any cone breakup peak. His pairings: 1.5" tweeter at ~1.5 kHz with a 6" ("the killer app" 2-way); 1" near 2.5 kHz, steep; 6" to small mid at 400 Hz–1 kHz; 3" to tweeter at 3.5–4 kHz. [RARE] Sources: OFCAP Ep. 19 @1:41:04 · DIYMA 2014-03-04 · DIYMA 2012-09-06

In plain English: Many people believe you should cross over to the tweeter "as low as possible". He says to choose the point by dispersion and driver safety instead. First, keep the smaller speaker safe: cross a tweeter at about twice its resonance, the pitch at which it naturally vibrates most easily. Then hand off where the larger speaker still spreads sound widely, and below any cone breakup peak, a spike that appears where the cone stops moving as one piece. His pairings: a 1.5-inch tweeter at about 1.5 kHz with a 6-inch woofer, which he calls "the killer app" 2-way. A 1-inch tweeter near 2.5 kHz, with a steep filter. A 6-inch to a small midrange at 400 Hz–1 kHz. A 3-inch midrange to a tweeter at 3.5–4 kHz.

1.4 #10 · Real pattern control is impractical in a car below ~1 kHz.

Using natural beaming would take ~15" midbass, 10" mids and 4" tweeters, or horns "and a school bus to put them in"; it "completely falls apart in the midbass". [RARE] Sources: DIYMA 2013-02-04 · DIYMA 2010-02-16

In plain English: Pattern control means steering where a speaker's sound goes, so less of it hits nearby surfaces. To do that using speakers' natural beaming, he says you would need about 15-inch midbass speakers, 10-inch midranges and 4-inch tweeters. The alternative is horns "and a school bus to put them in". He says the approach "completely falls apart in the midbass". This ties in with 1.3 #10, where he prefers using each speaker only where it spreads sound widely.

1.4 #11 · The point of a small midrange is crossover placement, not "all vocals from one driver".

A 2.5" bridges a 6" and a 1" so every crossover sits where dispersion is wide; running a tiny mid down to 200 Hz only costs distortion and power handling. Keeping vocals in one driver is "audio mythology at its worst". [RARE] Sources: DIYMA 2008-11-30 · DIYMA 2014-03-21

In plain English: A small midrange, such as a 2.5-inch, sits between the woofer and the tweeter. Many people believe its purpose is to play all of the vocals from one speaker. He calls that idea "audio mythology at its worst". He says its real job is to bridge a 6-inch woofer and a 1-inch tweeter, so that every crossover sits where both speakers still spread sound widely. Running a tiny midrange down to 200 Hz to cover more of the voice only costs distortion and power handling (how much power it can take without damage).

The full series

Download the complete guide as one PDF (351 pages, every point with its explanation).