Car Audio Explained, Part 9: Crossovers, Polarity and Phase (Andy Wehmeyer's Lessons)

4 days ago

38 min read

This is Part 9 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.

3.7 Crossovers: points, slopes and filter types

What this part is about: Crossovers, the filters that decide which speaker plays which range of notes. It covers where he sets them, how steep they should be, which filter types he uses and why, and the common crossover "tricks" he thinks make cars sound worse.

3.7 #1 · LR4 on every crossover by default.

Reasons: best protection for the high-passed driver (it nearly removes power at resonance), a narrow transition band (what cars need, given high power and compromised placement), a target phase difference of 0° that phase-reading analyzers can check, and acoustic slopes that follow the electrical ones when drivers are flat about an octave either side. If an amp's built-in crossover is 12 dB/oct, disable it and do LR4 in the DSP. [RARE] Sources: Tuning PDF v2 (~2023) · DIYMA 2021-09-27 · DIYMA 2020-10-03

In plain English: A crossover is a filter that splits music by pitch, so each speaker only gets the notes it can handle: highs to the tweeter, lows to the woofer. A "high-pass" filter lets highs through and blocks lows; a "low-pass" does the opposite. "LR4" is one specific recipe: Linkwitz-Riley, 4th order, which cuts at 24 dB per octave (an octave is a doubling of pitch, such as 1 kHz to 2 kHz). He uses it everywhere by default, for four reasons. It protects the high-passed speaker best, because it almost removes power at that speaker's resonance (the pitch where it naturally wants to move the most). The zone where both speakers overlap is narrow, which suits cars, with their high power and poor speaker locations. Done right, the two speakers should be in step (0° phase difference), which phase-reading analyzers can check. And if the speakers are themselves flat for about an octave either side, the sound you hear follows the filter's slopes. So in practice: if an amplifier's built-in crossover is only 12 dB/oct, switch it off and set LR4 in the DSP (the digital sound processor).

3.7 #2 · Pick crossover points in this order: protect the smaller driver, cross where the larger one still disperses widely, then worry about in-car response.

Never cross below resonance except on a sub; tweeters at about 1.5–2× Fs. In small pods Fs becomes Fc, so measure it: with a DATS, or a multimeter on current in series with the speaker during a quiet sine sweep (minimum current = resonance). [RARE] [Conflict: 2009 rule of 2× Fs and ~3 kHz for 1-in tweeters] Sources: DIYMA 2021-07-12 · DIYMA 2021-08-26 · DIYMA 2009

In plain English: He chooses crossover points in a fixed order. First, protect the smaller speaker (for example, the tweeter) from notes too low for it. Second, cross where the bigger speaker still spreads its sound widely, before it starts to "beam" (send sound mostly straight ahead). Only then does he worry about how the response looks in the car. Every speaker has a resonance frequency, called Fs, where it moves most easily. He says never cross below it, except on a subwoofer, and put tweeters at about 1.5 to 2 times their Fs. In a small pod (a small sealed housing), the trapped air raises the resonance; the new value is called Fc, so you have to measure it. Two ways: a DATS (a speaker-testing tool), or a multimeter reading current, wired in series with the speaker while a quiet sine-wave sweep plays. The pitch where the current is lowest is the resonance. The [Conflict] tag: in 2009 his rule of thumb was never below 2× Fs, and about 3 kHz for 1-inch car tweeters; his later posts allow 1.5–2× Fs.

3.7 #3 · Don't play mids and midbass as low as they can stand.

"Resonance IS distortion": at and below it the spider (his "parking brake") dominates; above it, driver distortion is a "nothingburger." Each octave lower roughly doubles power and quadruples excursion. GB25 midrange: 200 Hz if you must, 250 Hz safer, above 300 Hz a waste; a 4-in dash mid at 100 Hz only makes sense with no midbass. [RARE] Sources: DIYMA 2022-03-10 · DIYMA 2021-12-14 · DIYMA 2017-08-17

In plain English: Many people set each speaker to play as low as it can without obvious trouble. He disagrees. "Resonance IS distortion" is his phrase. At and below resonance, the spider dominates the cone's motion. The spider is the springy ring that holds the cone centred; he calls it the "parking brake". Above resonance, he says, the speaker's own distortion is a "nothingburger": not worth worrying about. Each octave lower roughly doubles the power the speaker must take and quadruples how far the cone moves (its excursion). His numbers for the Audiofrog GB25 midrange (Audiofrog is his company): 200 Hz if you must, 250 Hz is safer, and above 300 Hz is a waste. Hz (hertz) measures pitch; a lower number is a lower note. A 4-inch dash midrange crossed at 100 Hz only makes sense if there is no midbass speaker.

3.7 #4 · His dispersion chart grades crossover points green, yellow, orange and red.

A 6-in crossed to a 1-in tweeter at 5 kHz is "in the red zone"; 1.5–2 kHz is much better; adding a 2.5-in midrange crossed at 2.5 kHz lands in the green. Crossing where the larger driver has narrowed leaves a dip in the reflected sound the RTA can't separate, so it's hard to tune. Protect the tweeter first; yellow only if unavoidable, never red. [RARE] Sources: Tuning PDF v2 (~2023) · DIYMA 2020-12-23

In plain English: Dispersion is how widely a speaker spreads its sound. As notes get higher, a cone spreads its sound less, and bigger cones narrow sooner. His tuning PDF has a chart that colour-codes crossover points like a traffic light: green is good, yellow and orange are worse, red is bad. A 6-inch woofer crossed to a 1-inch tweeter at 5 kHz is "in the red zone", because the 6-inch already beams there. Crossing at 1.5–2 kHz is much better, and adding a 2.5-inch midrange crossed to the tweeter at 2.5 kHz lands in the green. Why it matters: if the bigger speaker has already narrowed, the sound bouncing off the car's surfaces has a dip near the crossover. An RTA (real-time analyzer: a microphone plus a screen showing level at each pitch) can't separate that from the direct sound, so the car becomes hard to tune. His order: protect the tweeter first; accept yellow only if you can't avoid it; never red.

3.7 #5 · Cars need 3-ways more than studios do, because close reflections expose a 2-way's off-axis hole.

A 2-way crossed where the mid beams can be flat on axis with a big hole off axis; the mic sees the hole, and boosting it puts a peak in the direct sound (piercing highs near the crossover). Studio monitors get away with 2-ways at ~3 ft with walls ~10 ft away. Home-speaker crossover rules still apply in a car and make EQ easier. [RARE] Sources: DIYMA 2021-01-11 · DIYMA 2022-01-05 · DIYMA 2013

In plain English: A 2-way system has a woofer and a tweeter on each side; a 3-way adds a midrange between them. People often ask why cars need 3-ways when many top studio monitors are 2-ways. His answer is reflections. A 2-way crossed where the woofer has started to beam can measure flat straight in front but have a big "hole" in the sound going off to the sides. In a car, surfaces are so close that the microphone picks up that hole. If you boost EQ to fill it, you also boost the direct sound, which creates a peak: piercing highs near the crossover. Studio monitors get away with 2-ways because the listener sits about 3 feet (0.9 m) away, with walls about 10 feet (3 m) away. His point: home-speaker crossover rules still apply in a car, and following them makes EQ easier.

3.7 #6 · Don't lower the tweeter crossover to stop a 6-in from beaming; build a 3-way.

Crossing every driver as low as possible wears speakers faster and can make the target harder to hit; a 6-in plus a tweeter "is not a great combo." The tweeter point can sit a little below 2× Fs only with steep slopes and careful tuning. [RARE] Sources: OFCAP Ep. 19 @1:41:04 · OFCAP Ep. 19 @1:44:09

In plain English: "Beaming" means a speaker sends its sound mostly straight ahead, which large cones do at higher notes. One tempting fix is to cross the tweeter lower so the 6-inch stops sooner. He says build a 3-way instead. Crossing every speaker as low as possible wears speakers out faster and can make the target curve (the tonal balance you aim for) harder to hit. In his words, a 6-inch plus a tweeter "is not a great combo." He allows the tweeter crossover a little below 2× Fs (twice its resonance) only with steep slopes and careful tuning.

3.7 #7 · His actual crossover numbers.

PDF 3-way: sub 60 Hz LP, midbass 80 Hz HP, mid/midbass 300 Hz, mid/tweeter 2.2 kHz. 2022 Audiofrog 3-way: sub 60 Hz; GB60 80–300 Hz; GB25 300 Hz–3 kHz; tweeter from 3 kHz (2.5 kHz with 24 dB/oct; ~3 kHz with shallower; 1.8 kHz only with LR4, and a 3-way never needs it). GB10 tweeter: 2.2 kHz+ LR4 or 2.5 kHz+ at ≥12 dB/oct. A 3-way's midrange needn't play low. Sources: Tuning PDF v1 (~2018) · DIYMA 2022-03-10 · DIYMA 2017-10-29

In plain English: These are the specific settings he has published. His tuning PDF's 3-way: subwoofer low-pass at 60 Hz, midbass high-pass at 80 Hz, midrange-to-midbass crossover at 300 Hz, and midrange-to-tweeter at 2.2 kHz (2,200 Hz). For an Audiofrog 3-way in 2022: sub to 60 Hz; GB60 midbass from 80 to 300 Hz; GB25 midrange from 300 Hz to 3 kHz; tweeter from 3 kHz. The tweeter can go down to 2.5 kHz with 24 dB/oct slopes, about 3 kHz with shallower slopes, and 1.8 kHz only with LR4, and he adds that a 3-way never needs that. For the GB10 tweeter: 2.2 kHz and up with LR4, or 2.5 kHz and up with a slope of 12 dB/oct or steeper. Finally, the midrange in a 3-way doesn't need to play low notes.

3.7 #8 · Raising mid crossovers to cure "distortion" usually hides a different fault.

The real cause is typically DSP input clipping, clipping inside the DSP, or a leaky pod. [RARE] Sources: DIYMA 2022-03-10

In plain English: When a midrange or midbass sounds distorted, a common reaction is to raise its crossover so it plays fewer low notes. He says that usually just hides a different fault. Clipping is what happens when a signal is too big for the electronics: its peaks get chopped off, which sounds harsh. The usual real causes he names: clipping at the DSP's input, clipping inside the DSP, or a leaky pod (air escaping from the speaker's housing).

3.7 #9 · Set the sub low-pass half an octave below the midbass high-pass (60 Hz LP, 80 Hz HP, both LR4).

Most people raise the sub 9–12 dB to sound right; with 80/80, that lifts the sub's real acoustic crossover and puts sub and front out of phase, so people misuse delay to drag bass forward. Rule: drop the sub low-pass half an octave per 12 dB of extra sub level, or lift the bass with a 12 dB/oct low shelf at ~60 Hz. [RARE] [Changed view: MS-8 era used the same point, 80/80] Sources: DIYMA 2020-12-30 · DIYMA 2021-08-21 · DIYMA 2015

In plain English: The sub's low-pass (where it stops playing upward) and the midbass's high-pass (where it stops playing downward) are often set to the same number. He sets them about half an octave apart: sub low-pass 60 Hz, midbass high-pass 80 Hz, both LR4. His reason: most people turn the sub up 9–12 dB so bass sounds right (dB, decibels, measure loudness; 10 dB more sounds roughly twice as loud). With both at 80 Hz, that boost moves the sub's real, acoustic crossover higher, and the sub and front speakers end up out of phase (one pushing while the other pulls, so they work against each other). People then misuse delay to drag the bass forward. His rule: lower the sub low-pass by half an octave for every 12 dB of extra sub level, or instead lift the bass with a low shelf (an EQ setting that raises everything below a chosen pitch) at about 60 Hz with a 12 dB/oct slope. The [Changed view] tag: in the MS-8 era (a processor he worked on at Harman) he used the same point for both, 80/80.

3.7 #10 · Bass "in front" is sub–midbass phase at the crossover, not a low or big midbass.

Wrong high-pass/low-pass phase is the #1 reason bass won't come from the front, and delay can't fix it (the half-wavelength-delay-plus-flip trick works only at one frequency). Aim to stay far from 180°, not to hit exactly 0° (60° still sums louder); check by proxy: measured slopes and −6 dB point match the target, delays correct. "Impact" lives near the mid/tweeter crossover (~3 kHz), not at 40 Hz. [RARE] Sources: DIYMA 2021-08-19 · DIYMA 2021-08-21 · DIYMA 2021-12-02

In plain English: Many people think bass that seems to come from in front of you (not from the trunk) needs a big midbass playing very low. He says it depends on phase between the sub and midbass at their crossover. Phase describes where each speaker is in its push-pull motion; in phase means pushing together. A wrong match between the high-pass and low-pass is the number-one reason bass won't come from the front, and delay can't fix it. One trick (delay by half a wavelength, then flip polarity) only works at a single frequency. Aim to stay far from 180° (fully opposite) rather than hit exactly 0°: even 60° apart still adds up louder. Phase is hard to see directly, so check it by proxy: the measured slopes and the −6 dB points match the target, and the delays are correct. Separately, he says "impact" comes from around the midrange-to-tweeter crossover (about 3 kHz), not 40 Hz; in a 2021 post his examples are a bass guitar's thumb slap and the snap of toms.

3.7 #11 · Door midbass crossed at 40 Hz is a red flag.

Wishful-thinking crossovers (fronts at 40 Hz, sub at 60–70 Hz) were the biggest flaw in championship cars he helped. His six-woofer Harman experiment gave smooth bass in every seat but "no image in any seat," so he minimizes sub/midbass overlap. A high-passed midbass needs no box: the filter does the box's job. [RARE] Sources: DIYMA 2018-02-09 · DIYMA 2009 · DIYMA 2021-09-11

In plain English: He treats door midbass speakers set to play down to 40 Hz as a warning sign. He calls such settings wishful thinking: fronts down to 40 Hz with the sub crossed at 60–70 Hz, so both play the same bass. That overlap was the biggest flaw in championship competition cars he helped with. At Harman he tried six woofers: bass was smooth in every seat, but there was "no image in any seat" (the image is the sense that instruments sit at particular spots in front of you). So he keeps sub and midbass overlap small. He adds that a midbass with a high-pass filter needs no box (enclosure): the filter does the box's job. Background: a sealed box acts on a speaker much like a high-pass filter, reducing its lowest notes and how far the cone moves; an electronic high-pass filter does that job instead.

3.7 #12 · A low-pass on the sub alone is not a crossover.

With full-range mains, the sub drifts out of phase with them about an octave above a 12 dB/oct cutoff (half an octave for 24 dB/oct). Crossovers work because the high-pass leads and the low-pass lags in a complementary way. [RARE] Sources: DIYMA 2022-02-10

In plain English: Some systems leave the main speakers full range (no high-pass) and only put a low-pass on the sub. He says that is not a real crossover. Every filter shifts timing (phase). In a true crossover the high-pass shifts its speaker ahead ("leads") and the low-pass shifts its speaker behind ("lags"), and the two shifts complement each other so the speakers add up correctly. With only a low-pass, nothing balances it: the sub drifts out of phase with the mains about an octave above a 12 dB/oct cutoff, or half an octave above with 24 dB/oct.

3.7 #13 · "The real crossover is the acoustic output of our speaker."

What counts is the measured point, slope, Q and phase (driver response × filter), not the DSP setting. Either EQ the driver flat first, or filter and then EQ until the result matches the target slope; then set delay by distance. [RARE] Sources: Audiofrog blog Jan 2021 (crossovers) · DIYMA 2017-06-25 · DIYMA 2020-10-19

In plain English: The number you type into the DSP is the electrical filter. What you actually hear is that filter combined with the speaker's own natural response in the car. He says the second is what counts: the measured crossover point, slope, Q (how sharp or gentle the bend is) and phase. There are two ways to get there: EQ the speaker flat first and then apply the filter, or apply the filter and then EQ until the measured result matches the target slope. Then set the delay by distance.

3.7 #14 · The crossover point is where the measured curves cross, and channel level moves it.

"Overlap" and "underlap" mean nothing unless measured; raising the low-passed channel's gain shifts the crossing (his example: sections meeting at 73 Hz with the dial at 63 Hz). Re-EQ after any crossover change, since crossovers are for safety and dispersion and EQ is for response. [RARE] Sources: DIYMA 2014 · DIYMA 2014 (73 Hz example) · DIYMA 2011

In plain English: He says the true crossover point is where the two speakers' measured curves cross. Words like "overlap" (both speakers play the same range) and "underlap" (a gap between them) mean nothing unless you measure. Turning up the gain (level) of the low-passed channel moves the crossing point. His example: the dial said 63 Hz, but the measured sections met at 73 Hz. So redo the EQ after any crossover change. His reasoning: crossovers are for safety and dispersion (how widely sound spreads); EQ is for the frequency response (the tonal balance).

3.7 #15 · How filter types sum.

Each 6 dB/oct adds 45° of shift. First order: 90° apart, flat on paper, poor in practice. 2nd-order Butterworth: +3 dB hump with polarity reversed (deep dip without); LR2: flat with reversal (180°). 3rd-order Butterworth: 270° apart, each −3 dB, flat in either polarity, so no polarity check. LR4: each −6 dB, 360° apart, flat with correct polarity. Sources: Audiofrog blog Jan 2021 (crossovers) · DIYMA 2018-01-21 · DIYMA 2022-01-14

In plain English: This point explains how different crossover recipes add together. Background: the "order" of a filter is its steepness: 1st order is 6 dB/oct, 2nd is 12, 3rd is 18, 4th is 24. Butterworth and Linkwitz-Riley (LR) are two families of filter shape. Polarity is which way the cone moves for a positive signal; swapping a speaker's + and − wires reverses it. He says each 6 dB/oct adds 45° of phase shift. First order: the two speakers are 90° apart; flat on paper, poor in practice. 2nd-order Butterworth: a +3 dB hump at the crossover with one speaker's polarity reversed, and a deep dip without the reversal. LR2: flat with the reversal, since the two are 180° apart. 3rd-order Butterworth: 270° apart, each 3 dB down at the crossover, flat in either polarity, so it gives you no way to check polarity. LR4: each 6 dB down, 360° apart (a full turn, so back in step), and flat with correct polarity.

3.7 #16 · Steeper slopes don't protect more, and are no excuse to cross lower.

Each 6 dB removes 75% of what's left (from 100 W the steps remove 75, 18.75, 4.69, then 1.17 W), so most protection comes early; a 48 dB/oct filter passes about 1.5× more power near the knee than LR4. Moving a tweeter from 4 to 2 kHz roughly doubles its power. 6 dB slopes are cheaper, not better; beyond 24 dB/oct wastes processing. [RARE] Sources: Audiofrog blog Oct 2021 (slopes and protection) · DIYMA 2021-07-15 · DIYMA 2013

In plain English: Many people believe a steeper filter protects a speaker better, so they can cross it lower. He says no. Each 6 dB of reduction removes 75% of the power that's left. From 100 W (watts), the first 6 dB step removes 75 W, the next 18.75 W, then 4.69 W, then 1.17 W, so most of the protection comes early. Near the "knee" (the corner where the slope begins), a 48 dB/oct filter actually passes about 1.5 times more power than LR4. The crossover point matters more: moving a tweeter from 4 kHz down to 2 kHz roughly doubles the power it receives. He says 6 dB/oct slopes are cheaper, not better, and going beyond 24 dB/oct wastes processing power.

3.7 #17 · Tweeters see very little power.

With 1000 W of pink noise in the 20 Hz octave, a mid crossed at 80 Hz gets at most ~250 W and a tweeter crossed at 2.5 kHz ~8 W; EDM sends even less up top, jazz more. A 24 dB/oct high-pass at 2.5 kHz is −24 dB at 1.25 kHz, so an extra series cap only acts where the tweeter contributes nothing. [RARE] Sources: Audiofrog blog Dec 2020 (pink noise) · DIYMA 2018-05-04

In plain English: Pink noise is a test signal with equal energy in every octave. He uses it as a stand-in for music and explains that, as a result, the power an amp delivers halves with each octave up: 1000 W at 20 Hz means 500 W at 40 Hz and 250 W at 80 Hz. So a midrange crossed at 80 Hz gets at most about 250 W, and halving all the way up to 2.5 kHz leaves a tweeter about 8 W. EDM (electronic dance music) sends even less to the top; jazz sends more. Also, a 24 dB/oct high-pass at 2.5 kHz is already 24 dB down at 1.25 kHz, one octave lower. So an extra capacitor in series (a part that blocks low notes) only acts where the tweeter contributes nothing anyway.

3.7 #18 · Keep a protection capacitor on tweeters even in active systems, but remove it if you go back to a passive crossover.

Pick a value putting −3 dB about an octave below Fs (e.g., 68 µF for 4 Ω behind an LR4 2.5 kHz high-pass, ~0.5 dB change from 1.8–2.5 kHz); a ~$1.75 non-polar electrolytic is fine. It's a "parking brake" against pops, amp failures and wrong DSP entries; in a passive network it alters slope and Q. [RARE] Sources: Audiofrog blog Dec 2020 (to cap or not) · DIYMA 2021-09-22

In plain English: A capacitor wired in line with a tweeter blocks low notes. He says keep one on tweeters even in active systems (where the DSP does the crossover), but remove it if you go back to a passive crossover (the box of parts that often comes with component speakers). Choose a value that puts its −3 dB point about an octave below the tweeter's resonance (Fs). His example: 68 µF (microfarads) for a 4 Ω (ohm) tweeter behind an LR4 high-pass at 2.5 kHz, which changes the response by only about 0.5 dB between 1.8 and 2.5 kHz. A cheap non-polar electrolytic capacitor, about $1.75, is fine. He calls it a "parking brake" against pops, amplifier failures and wrong settings typed into the DSP. In a passive network, though, it changes the filter's slope and Q (the shape of its bend).

3.7 #19 · Crossover summation checks only work anechoically; in a car, highs sum to about +3 dB.

Reflections mean nothing is truly in phase up high, so L+R or driver sums won't reach +6 dB. Don't obsess over mid/tweeter phase (it can't be measured in a way that matches hearing); sub/midbass and midbass/mid phase matter and can be measured. [RARE] Sources: DIYMA 2021-03-26 · DIYMA 2021-08-25 · DIYMA 2021-07-12

In plain English: "Summation" means playing two sounds together (left plus right, or the two speakers either side of a crossover) and checking how much louder the result is. Background: two identical sounds perfectly in phase add up to +6 dB. He says this check only works anechoically, meaning in a space with no echoes. In a car, reflections mean nothing is truly in phase at high frequencies, so highs add up to only about +3 dB, and left-plus-right or driver sums won't reach +6 dB. His advice: don't obsess over midrange-to-tweeter phase, because it can't be measured in a way that matches what you hear. Phase at the sub/midbass and midbass/midrange crossovers does matter, and it can be measured.

3.7 #20 · Group delay and FIR crossovers are overrated in cars.

Stop-band group delay doesn't matter because the driver isn't used there; front-bass failures are crossover phase, not group delay (he mocks rejecting vented boxes for group delay, then using a 72 dB/oct low-pass at 40 Hz). FIR isn't inherently "frequency-dependent delay"; he questions FIR crossovers. [RARE] Sources: DIYMA 2017-09-12 · DIYMA 2022-01-14 · DIYMA 2021-08-20

In plain English: Group delay describes how much a filter or speaker delays some frequencies relative to others. FIR (Finite Impulse Response) is a type of digital filter often promoted for its phase behaviour. He says both are overrated in cars. Group delay in the stop-band (the range a filter is blocking) doesn't matter, because the speaker isn't being used there. When bass won't come from the front, the cause is crossover phase, not group delay. He mocks people who reject vented (ported) boxes because of their group delay, then use a 72 dB/oct low-pass at 40 Hz, a very steep filter. He says FIR isn't inherently "frequency-dependent delay", and he questions FIR crossovers.

3.7 #21 · Keep left and right crossover slopes identical.

Different L/R midbass slopes act like an all-pass on one side, shifting relative phase about 45° an octave above the crossover and ~90° two octaves above. [RARE] Sources: DIYMA 2018-01-27

In plain English: Different crossover slopes on the left and right midbass act like an all-pass filter on one side. An all-pass filter leaves the level alone but shifts the timing (phase). The result: the left and right midbass drift about 45° apart one octave above the crossover, and about 90° apart two octaves above. So in practice: use the same slope on both sides.

3.7 #22 · Sometimes you must move the crossover rather than EQ.

Occasionally a reflection or cancellation forces it: in a Dodge Durango's OE Alpine system he moved door mids from ~4 kHz down to ~1 kHz, before they beamed and before their first big peak. Avoid bands under an octave wide. The popular "tricks" (big overlap, flipped mids, 6 dB slopes for "blend") always made his car worse. Sources: DIYMA MS-8 FAQ · DIYMA 2014 · DIYMA 2011

In plain English: Normally he fixes response problems with EQ, but occasionally a reflection or cancellation forces a crossover change instead. His example is a Dodge Durango with the factory (OE, original equipment) Alpine system. He moved the door midranges' crossover from about 4 kHz down to about 1 kHz, so they stopped before they began to beam and before their first big peak in response. He also says to avoid bands under an octave wide, meaning a speaker given a range of less than one octave to play; in a 2014 post he gives 80–150 Hz as an example of such a bad crossover. The popular "tricks" (big overlap between speakers, midranges with flipped polarity, and 6 dB slopes for "blend") always made his car worse.

3.7 #23 · For a customer who only wants loud, raise the crossovers like a pro sound system.

His 1970s-disco example: a "correct" 60–80 Hz system kept getting blown because the owner boosted bass ~12 dB at the head unit, past clipping, cancelling the midbass high-pass. Raise the crossover and accept less front bass; get to correct, then back off it. [RARE] Sources: CMA Connected (Oct 2021) @43:34

In plain English: Not every customer cares about imaging. His example is a customer who played 1970s disco at full volume. The dealer had built a "correct" system crossed at 60–80 Hz, but it kept getting blown. The owner boosted bass by about 12 dB at the head unit (the radio), pushing it past clipping (overloading so the signal's peaks get chopped off), and that boost cancelled the midbass high-pass. His fix: raise the crossover and set it up like a pro sound system, accepting less bass from the front. His wider point: "get to correct" first, then back off from correct to suit the listener.

3.7 #24 · (MS-8 manual) Factory defaults, for reference.

24 dB/oct; sub/front 80 Hz; tweeters at 3.5 kHz or higher; subsonic 20 Hz 12 dB/oct; front high-pass without a sub 20–40 Hz for 6-in+ and 40–60 Hz for 5-in and smaller; center 80 Hz; sides and rears ≥100 Hz. Product documentation, not verified as his words. [Era: MS-8] Sources: MS-8 manual (mirror)

In plain English: This point is from the manual of the JBL MS-8, a car-audio processor made by Harman, where he worked on it. These are its factory starting settings, listed for reference: 24 dB/oct slopes; sub-to-front crossover at 80 Hz; tweeters at 3.5 kHz or higher; a subsonic filter (which cuts the very lowest notes) at 20 Hz, 12 dB/oct; front high-pass, when there is no sub, at 20–40 Hz for 6-inch speakers and larger and 40–60 Hz for 5-inch and smaller; center speaker 80 Hz; side and rear speakers 100 Hz or higher. As the point says, this is product documentation, not confirmed as his own words.

3.8 Polarity and phase

What this part is about: Polarity (which way a speaker's cone moves when the signal pushes) and phase (whether two speakers push and pull in step at a given pitch). It covers how he checks both, what phase targets he aims for, and how all-pass filters, including ones hidden inside factory systems, change the picture.

3.8 #1 · Set polarity with markings or a test tool, never "whichever sounds better."

Positive voltage should push every cone toward the listener. If a flip sounds better, find out why: a left/right distance error needs delay, not a flip. With LR4 (Linkwitz-Riley 24 dB/oct) and correct polarity the crossover sums flat, so a deep, narrow dip there flags a reversed driver; flip it on both sides. [RARE] Sources: Tuning PDF v2 (~2023) · DIYMA 2014 · DIYMA 2014

In plain English: Polarity is which way a speaker's cone moves when the signal goes positive; swapping a speaker's + and − wires reverses it. He says the right setting is that positive voltage pushes every cone toward the listener, and you set it from the wire markings or with a test tool, never by picking whichever sounds better. If a flip sounds better, find out why. For example, if the left and right speakers sit at different distances from you, the fix is delay (holding back the nearer speaker), not a flip. With LR4 (Linkwitz-Riley 24 dB/oct) and correct polarity, the crossover adds up flat. So a deep, narrow dip right at the crossover tells you a speaker is reversed; flip it on both sides.

3.8 #2 · Never fix a vocal-range center problem by flipping one midrange, and never flip one and add delay too.

His worked example: a flip "fixed" 250 Hz but worsened the bass sum; adding correct delays then left the midbasses 180° out at all frequencies and destroyed the center. Use delay or a flip, "one or the other, not both"; flipping is also what forum advice often tells people to do. [RARE] Sources: Tuning PDF v2 (~2023) · Audiofrog blog Dec 2020, Time Alignment Part 2 · OFCAP Ep. 19 @3:17:02

In plain English: The centre image is the sense that a singer stands in the middle, between the speakers. When it's wrong in the vocal range, forum advice often says to flip one midrange's polarity. He says never do that, and never flip one and add delay as well. In his worked example, a flip "fixed" 250 Hz but made the bass sum worse. Adding correct delays on top then left the midbasses 180° out of phase (fully opposed) at all frequencies, which destroyed the centre. His rule: use delay or a flip, "one or the other, not both".

3.8 #3 · Most polarity-test tracks can't check woofers or subs; use a long, all-positive pulse.

Short clicks lack the bass that passes a low-pass filter ("If there's no bass, then we can't check a subwoofer"). Use a right-triangle pulse with a vertical leading edge (Audiofrog's is about 9 ms) or a long train of positive pulses; read only the first motion (later negative swings are normal) and mute the other speakers. Alternative: a DC-offset track and a voltmeter at the speaker leads. [RARE] Sources: Audiofrog blog 2021-08-17, polarity · DIYMA 2021-08-16 · DIYMA 2015

In plain English: A polarity test track plays a pulse, and a polarity checker (a small tool held near the speaker) reads which way each cone first moves. He says most such tracks can't check woofers or subs, because their short clicks have little bass to pass through a low-pass filter: "If there's no bass, then we can't check a subwoofer." Instead, use a long pulse that is all positive: a right-triangle shape with a vertical leading edge (Audiofrog's lasts about 9 ms, nine thousandths of a second), or a long train of positive pulses. Read only the first motion, because later negative swings are normal, and mute the other speakers while you test. Another option: a DC-offset track (a steady positive voltage) and a voltmeter on the speaker's wires.

3.8 #4 · A reversed tweeter in a passive crossover is usually deliberate.

Only Linkwitz-Riley orders that are multiples of 4 sum in phase; 2nd order is 180° apart and needs a reversal. For a passive tweeter right next to the woofer (1.5" offset ≈ half a wavelength at 4.5 kHz), a deep narrow dip at the crossover means reverse the tweeter, then tune. Sources: Audiofrog blog 2021-08-17, polarity · DIYMA 2018-04-25 · DIYMA 2021-09-26

In plain English: A passive crossover is the set of parts (often in a small box) that comes with many component speaker sets. In these, the tweeter is often wired reversed, and he says that is usually deliberate. Only Linkwitz-Riley filters whose order is a multiple of 4 add up in phase; a 2nd-order filter leaves the two speakers 180° apart, so one needs reversing. For a passive tweeter right next to the woofer, he notes that even a 1.5-inch (about 38 mm) offset is about half a wavelength at 4.5 kHz (at that pitch, the two sounds arrive half a cycle out of step). His test: if there's a deep, narrow dip at the crossover, reverse the tweeter, then tune.

3.8 #5 · Polarity is not delay.

A flip starts the sound at the same moment and multiplies phase by −1 at every frequency; a fixed delay adds phase that grows with frequency (there are 360° in one cycle and f cycles in one second, so the same delay is a different number of degrees at each frequency). So a flip can never substitute for distance correction. [RARE] Sources: DIYMA 2008 · DIYMA 2016

In plain English: Polarity and delay are different tools, and he says one cannot stand in for the other. A polarity flip doesn't make the sound start later: it starts at the same moment, but every push becomes a pull (phase is multiplied by −1 at every frequency). A fixed delay holds the sound back, and the phase shift it adds grows with frequency. Every cycle has 360°, and higher notes complete more cycles each second, so the same delay amounts to a different number of degrees at different pitches. So in practice: a flip can never substitute for correcting a speaker's distance.

3.8 #6 · The goal is "far from 180°," not 0°.

Phase only matters where two speakers play the same frequency within about 12 dB; under 90° apart is hard to hear, so treat ±45° as in phase and leave anything within ±90° alone. Sub and midbass need only be within ±90°; if one midbass is in phase with the sub and the other isn't, shift the sub about 90° (+3 dB, not a null). [RARE] Sources: DIYMA 2012 · DIYMA 2016 · DIYMA 2021-02-04

In plain English: Phase difference is measured in degrees: 0° means two speakers push and pull exactly together, and 180° means exactly opposite, so they cancel. He says aim to be "far from 180°", not to hit 0°. Phase only matters where two speakers play the same frequency within about 12 dB of each other. Differences under 90° are hard to hear, so he treats ±45° as in phase and leaves anything within ±90° alone. Sub and midbass only need to be within ±90°. If one midbass is in phase with the sub and the other isn't, shift the sub's phase about 90°; that gives roughly +3 dB rather than a null (a deep cancellation dip).

3.8 #7 · Read phase from the left+right sum of mono pink noise.

In phase sums about +6 dB over one side; ~90° apart +3 dB; ~120° equals one side; less than one side means near 180°. Real cancellation is 135–225°. Use mono (correlated) noise; he sees no use for uncorrelated noise here. Every peak or dip carries a phase shift, so where L and R peaks don't line up, they aren't in phase. [RARE] Sources: DIYMA 2021-08-26 · DIYMA 2021-11-02 · DIYMA 2021-05-09

In plain English: This is his way to read phase from simple level measurements. Play mono pink noise (the same noise in both channels, called "correlated") and measure left alone, right alone, then both together. If the sum is about 6 dB louder than one side, they're in phase. About +3 dB means roughly 90° apart. If the sum equals one side, they're about 120° apart. If it's quieter than one side, they're near 180°. Real cancellation only happens between 135° and 225°. He sees no use here for uncorrelated noise (different noise in each channel). Also, every peak or dip in a response carries a phase shift, so where the left and right peaks don't line up, the two sides aren't in phase.

3.8 #8 · In a car, highs only sum to about +3 dB, so the "summation" crossover method only works anechoically.

Dense reflections average high-frequency phase out to about 90°; in 30 years he has never seen a car differ. Expect ~+6 dB below about 1 kHz and ~+3 dB above. Sources: DIYMA 2021-03-26 · DIYMA 2021-08-25 · DIYMA 2021-08-26

In plain English: Two identical sounds perfectly in step add up to +6 dB. In a car, he says, the highs only add up to about +3 dB, because dense reflections average the high-frequency phase out to about 90°. In 30 years he has never seen a car behave differently. So expect about +6 dB below roughly 1 kHz and about +3 dB above it. That's why the "summation" method of setting crossovers (adjusting until the two speakers add up to the full +6 dB) only works anechoically, meaning in a space with no echoes.

3.8 #9 · Phase matters much less above about 1 kHz; don't chase mid-to-tweeter phase.

Those dips are narrow, an inch of head or mic movement changes them, and a spatial average has no phase at all. Set mid/tweeter delay by tape measure, EQ both to the target slopes, and check a few spots around your head. Sub/midbass and midbass/mid are the crossovers where phase is both important and measurable. [RARE] Sources: Audiofrog blog Dec 2020, Time Alignment Part 2 · DIYMA 2021-07-12 · OFCAP Ep. 19 @2:56:00

In plain English: Above about 1 kHz, he says, phase matters much less, so don't chase phase between the midrange and tweeter. The dips you see there are narrow, and moving your head or the microphone by an inch (25 mm) changes them. A spatial average (several measurements around the head combined into one) has no phase information at all. His method instead: set the midrange/tweeter delay with a tape measure, EQ both speakers to the target slopes, and check a few spots around your head. The crossovers where phase is both important and measurable are sub/midbass and midbass/midrange.

3.8 #10 · When a crossover pair is in phase, you hear both drivers at the higher-frequency one.

So correct sub/midbass phase pulls the bass forward, and correct midbass/mid and mid/tweeter phase keeps vocals at tweeter height ("the cure for the rainbow is a good crossover"). That is also why club-style listeners who want impact need a DSP: impact is midbass-to-sub phase. [RARE] Sources: DIYMA 2021-12-02 · DIYMA 2017-08-17 · CMA Connected 2021 @46:57

In plain English: When the two speakers either side of a crossover are in phase, you hear them as one sound, located at the higher-frequency speaker. So correct sub/midbass phase pulls the bass forward to the front speakers. Correct midbass/midrange and midrange/tweeter phase keeps vocals at tweeter height. In his words, "the cure for the rainbow is a good crossover". The sources don't define "rainbow" exactly; it appears to mean different frequency ranges seeming to come from different places, instead of all from one point above the dash. He adds that this is why club-style listeners, who want impact rather than imaging, still need a DSP: impact is midbass-to-sub phase.

3.8 #11 · Only the acoustic crossover counts, and you can't flatten magnitude first and fix phase later.

The target is an acoustic LR4, flat in magnitude and phase; EQ and electrical filters add together, and an all-pass that corrects phase also changes the summed level. For minimum-phase behaviour, matching one channel's magnitude to the other also matches phase. Sources: DIYMA 2013 · DIYMA 2017-06-25 · DIYMA 2015

In plain English: The "acoustic crossover" is what the speakers actually produce in the car, as opposed to the filter settings typed into the DSP. His target is an acoustic LR4: flat in both magnitude (level) and phase. He warns that you can't make the level flat first and fix the phase later. EQ and electrical filters add together, and an all-pass filter used to correct phase also changes the level of the combined sound. "Minimum-phase" is a technical term for behaviour where a speaker's phase is tied to its level response. Where that holds, he says, matching one channel's magnitude to the other's also matches its phase.

3.8 #12 · What an all-pass filter is: phase rotation with flat level, not a frequency-dependent delay.

A 1st-order all-pass rotates 0→180° (90° at its frequency); a 2nd-order 0→360° (180° at its frequency); Q sets how fast. Its phase curve is a low-pass filter's. Degrees per inch depend on frequency: 1° ≈ 0.48" at 80 Hz, 1.86" at 20 Hz, 0.04" at 1 kHz. An "AP" crossover setting on an amp just means no filter. [RARE] Sources: Dean & Fernando live @10:07 · DIYMA 2021-12-03 · DIYMA 2021-05-13

In plain English: An all-pass filter lets every frequency through at full level but rotates its phase (shifts its timing) around a chosen frequency. A 1st-order all-pass rotates phase from 0° to 180°, reaching 90° at its set frequency; a 2nd-order one rotates from 0° to 360°, reaching 180° at its frequency. Q sets how fast the rotation happens. He notes its phase curve is a low-pass filter's, and stresses that it is not a frequency-dependent delay. How many inches one degree equals depends on frequency: 1° is about 0.48 inch (12 mm) at 80 Hz, 1.86 inches (47 mm) at 20 Hz, and 0.04 inch (1 mm) at 1 kHz. And an "AP" crossover setting on an amplifier just means no filter at all.

3.8 #13 · A DSP's "phase" slider (e.g., Helix) is an all-pass; its degrees just move the filter frequency.

At the sub/midbass crossover he'd rather get the crossover right; if he uses one, it goes on the high side of the crossover, where its effect on the sum shrinks as that speaker rolls off. All IIR filters (high-, low-, all-pass) delay frequencies around and below their frequency, never above. Sources: DIYMA 2017-07-07 · DIYMA 2021-12-03 · DIYMA 2021-03-05

In plain English: Many DSPs, such as the Helix, have a "phase" slider marked in degrees. He says it is really an all-pass filter, and changing the degrees just moves the filter's frequency. At the sub/midbass crossover he would rather get the crossover itself right. If he does use one, he puts it on the "high side" of the crossover; his reason is that where it rotates the phase, that speaker is getting quieter, so its effect on the combined sound shrinks. He adds a general rule: all IIR filters (the standard kind of DSP filter: high-pass, low-pass and all-pass) delay frequencies around and below their set frequency, never above.

3.8 #14 · The 250 Hz insight: an all-pass gives both front seats a center, which delay can't.

With door speakers the path difference first puts L and R out of phase around 250 Hz (lower in big cars, higher in small), mid-vocal range; the error is equal and opposite for the two seats at the same frequency, so reversing one side's phase in that band helps both. Driver's seat: a real but larger center (he likens delay-tuned cars to a diorama of tiny army men on the dash); passenger: diffuse but spread. [RARE] Sources: Dean & Fernando live @21:00 · Dean & Fernando live @23:05 · DIYMA 2016

In plain English: This is one of his key ideas for cars with two front seats. With door speakers, each front seat is closer to one side. That difference in path length first puts left and right out of phase around 250 Hz (lower in big cars, higher in small ones), in the middle of the vocal range. The error is equal and opposite for the two seats at the same frequency, so reversing one side's phase in that band, with an all-pass filter, helps both seats at once. Delay can't do that. The result in the driver's seat is a real but larger centre image; he likens delay-tuned cars to a diorama of tiny army men on the dash. In the passenger seat the sound is diffuse but spread out.

3.8 #15 · How he applies all-pass "phase EQ" for two seats.

Use it only between about 80 Hz and 800 Hz–1 kHz, never higher. For a 250–500 Hz L/R mismatch: a 2nd-order all-pass at 250 Hz, Q about 1; in his own car, one on the sub plus one half an octave higher on one midbass centered the bass for both seats. To shift L/R phase across a band, put a 1st-order all-pass on each side at spaced frequencies. In a true two-seat system, drop the L/R delays. [RARE] [Changed view: 2010 he had never made an all-pass work for sub/midbass blending ("all about amplitude"); 2014–15 Harman's sub-band phase EQ made "no audible difference"; from 2016 on it is his preferred two-seat tool.] Sources: DIYMA 2010 · DIYMA 2016 · Dean & Fernando live @29:05

In plain English: "Phase EQ" here means using all-pass filters to adjust phase in chosen bands. He uses it only between about 80 Hz and 800 Hz–1 kHz, never higher. For a left/right mismatch between 250 and 500 Hz: a 2nd-order all-pass at 250 Hz with a Q of about 1. In his own car, one all-pass on the sub plus one half an octave higher on one midbass centred the bass for both seats. To shift left/right phase across a band, put a 1st-order all-pass on each side at spaced-apart frequencies. In a true two-seat system, drop the left/right delays. The [Changed view] tag: in 2010 he had never made an all-pass work for sub/midbass blending, which he called "all about amplitude" (level); in 2014–15 Harman's sub-band phase EQ made "no audible difference"; from 2016 on it has been his preferred two-seat tool.

3.8 #16 · Before using an all-pass on an L/R dip, check whether it matters.

If band-limited pink noise sits centered and the pan tracks move evenly, leave a small dip alone (his case: 650–1800 Hz, ~120° apart) or EQ both fronts identically. A dip that sums to the level of one side is ~120°, not 180°; with one mic, re-measure around a ~7" circle, and only if it persists nudge the all-pass frequency. An all-pass can fix one band and break another, and one on the driver's side helps only the driver. [RARE] Sources: DIYMA 2021-05-07 · DIYMA 2021-08-27 · DIYMA 2022-03-24

In plain English: Before using an all-pass on a left/right dip, he first checks whether the dip matters. Play band-limited pink noise (noise covering just the range in question) and pan tracks (test tracks that move a sound from side to side). If the noise sits centred and the pans move evenly, leave a small dip alone (his case: 650–1800 Hz, about 120° apart), or EQ both fronts the same way. A dip where the sum equals one side alone means about 120°, not 180°. With one microphone, re-measure at several points around a circle of about 7 inches (18 cm); only if the dip persists, nudge the all-pass frequency a little. Note that an all-pass can fix one band and break another, and one on the driver's side helps only the driver.

3.8 #17 · Two quick tests tell you whether a factory system uses all-pass filters or delay.

(1) Listen to a centered vocal from both front seats: all-pass gives a broad center from the driver's seat and both sides audible from the passenger seat; delay collapses everything into the passenger door. (2) Feed mono pink noise and view L+R in the DSP's input analyzer: one big dip between 200 and 500 Hz means all-pass; a dip followed by evenly spaced dips (a comb) means delay. [RARE] Sources: DIYMA 2021-10-26

In plain English: Factory systems sometimes use all-pass filters to give both front seats a centre image, while others use delay. He gives two quick tests to tell which. First, listen to a centred vocal from both front seats. With all-pass, the driver hears a broad centre and the passenger can hear both sides. With delay, everything collapses into the passenger door; he compares it to a one-seat competition tune heard from the wrong side. Second, feed mono pink noise and look at the left-plus-right sum in your DSP's input analyzer. One big dip between 200 and 500 Hz means all-pass; a dip followed by evenly spaced dips (a "comb", like a comb's teeth) means delay.

3.8 #18 · What factory all-pass layouts look like, and why a polarity checker won't see them.

Typical: one channel reversed plus a 1st-order all-pass that brings the bass back into phase, leaving the vocal band near 250 Hz centered and out-of-phase highs for width; or a 2nd-order all-pass near 250 Hz on one side; on one plot he read two 1st-order all-passes near 250 and 500 Hz. Bose chains several per channel. A pulse checker reads an all-pass like a crossover, so it shows no reversal. [RARE] Sources: DIYMA 2021-05-16 · DIYMA 2021-05-23

In plain English: Here he describes what factory all-pass setups typically look like. One common layout: one channel's polarity reversed, plus a 1st-order all-pass that brings the bass back into phase. That leaves the vocal band around 250 Hz centred, while the highs stay out of phase, which makes the sound wider. Another: a 2nd-order all-pass near 250 Hz on one side. On one measurement plot he read two 1st-order all-passes near 250 and 500 Hz. Bose chains several per channel. A polarity checker reads an all-pass the same way it reads a crossover filter, so it shows no reversal.

3.8 #19 · Measure factory outputs against the original pink noise, not channel against channel.

Comparing L with R can't show where a filter sits (X+Y=10 has many answers; 1+Y=10 has one). In REW an all-pass is easy to spot on a flat response with the start time set to zero. Check for all-pass filters, EQ and bass limiting before blaming the factory D/A. Sources: DIYMA 2021-05-20 · DIYMA 2021-05-16 · DIYMA 2021-02-03

In plain English: When investigating a factory system, he says to measure each output against the original pink noise signal, not left against right. Comparing two outputs with each other can't show which one holds a filter. His analogy: X + Y = 10 has many answers, but 1 + Y = 10 has only one; knowing the original signal is like knowing X. In REW (Room EQ Wizard, a free measurement program), an all-pass is easy to spot on a flat response with the start time set to zero. And before blaming the factory D/A (the converter that turns digital audio into an electrical signal), check for all-pass filters, EQ and bass limiting.

3.8 #20 · Living with factory all-pass filters: keep the factory polarity and add no delays.

"Reversed" OEM polarity is usually intended (connectors can't be swapped). An all-pass can't be inverted (that would break causality), and delay on top makes it work poorly; you can only match it with the same all-pass on the other side, or remove it with an FIR (Finite Impulse Response) filter that delays everything, adding latency that hurts Bluetooth calls. The driver still gets a broad image, and the passenger is spared a driver's-seat delay tune (see 3.13 #10). [RARE] Sources: DIYMA 2021-05-20 · DIYMA 2021-05-13 · DIYMA 2021-03-26

In plain English: If a factory system uses all-pass filters, he says keep the factory polarity and add no delays. "Reversed" factory polarity is usually intended: the factory connectors can't be swapped, so it isn't an installer mistake. An all-pass can't be inverted (undone by an opposite filter), because that would break causality: the filter would have to respond before the sound arrives. Adding delay on top makes it work poorly. Your only options are to match it with the same all-pass on the other side, or remove it with an FIR (Finite Impulse Response) filter, which delays everything and adds latency (lag) that hurts Bluetooth calls. The upside of leaving it: the driver still gets a broad image, and the passenger is spared a tune made with delays for the driver's seat (see 3.13 #10).

3.8 #21 · "Pressure on the ears" usually means a low-frequency phase problem

, especially in the driver's seat. Sources: DIYMA 2016

In plain English: Some listeners describe certain systems as putting "pressure on the ears". He says that usually means a phase problem in the bass range, especially in the driver's seat.

3.8 #22 · Polarity markings on subs are almost never wrong.

Claims that half are mismarked are implausible on a documented line; of ~4,000 subs built with reversible coloured binding posts, he knows of 3 wrong. With amp crossovers, make sure all are engaged at the same frequency. Sources: DIYMA 2018-04-07

In plain English: Polarity markings are the + and − labels or colours on a speaker's terminals. Some people claim half of subwoofers are mismarked. He says that's implausible for a documented product line: of about 4,000 subs built with reversible coloured binding posts (terminals whose colours could be fitted either way round), he knows of 3 that were wrong. If you use the amplifier's own crossovers, make sure they are all switched on at the same frequency.

3.8 #23 · Different left and right slopes act like an all-pass on one side.

Mismatched L/R midbass slopes shift relative phase about 45° an octave above the crossover and about 90° two octaves above. Alignment targets also differ by filter type: 0° at the crossover is right only for LR4; 1st-order should be 90° apart, LR2 −180°, and a 2nd-order Butterworth aligned at 0° peaks. [RARE] Sources: DIYMA 2018-01-27 · DIYMA 2018-01-21

In plain English: If the left and right midbass use different crossover slopes, one side behaves as if it had an all-pass filter (one that shifts timing without changing level). Their relative phase moves about 45° one octave above the crossover and about 90° two octaves above. He also says the right phase target depends on the filter type: 0° at the crossover is correct only for LR4. For 1st order the two speakers should be 90° apart, for LR2 −180°, and a 2nd-order Butterworth aligned at 0° gives a peak.

3.8 #24 · Hearing midbass and tweeter as two separate sources means the crossover phase is wrong.

Sources: DIYMA 2020-12-23

In plain English: If you can hear the midbass and the tweeter as two separate places the sound comes from, rather than one blended source, he says the crossover phase between them is wrong.

Top rarely-discussed points in 3.1–3.7

Measure from outside the car with the door shut; his target assumes an empty car (3.4 #1)
Never wave the mic around your head from the seat; your head eats the treble and no target allows for it (3.4 #2)
Matching left to right beats hitting the target; copy an unfixable dip to the other side (3.6 #2)
Slide the target onto the dips like a router template and cut the peaks down to it (3.6 #3)
"Does it respond to EQ?" separates fixable minimum-phase problems from reflection or two-speaker interference (3.6 #10)
The "Harman curve" began as his installer curve, and popular targets differ by only ~2 dB (3.5 #2, #3)
One mic with 1/6-octave smoothing gets about 90% of a spatial average; five or six positions are plenty (3.4 #5, #7)
Lower the sub low-pass half an octave for every 12 dB of sub boost, keeping 60/80 Hz LR4 (3.7 #9)
Bass "in front" is sub–midbass crossover phase; door midbass crossed at 40 Hz is a red flag (3.7 #10, #11)
Steeper slopes don't protect tweeters more; 48 dB/oct passes more power near the knee than LR4 (3.7 #16)
Swapping speakers is just a non-adjustable EQ; EQ is the main lever (3.2 #9)
A 250 Hz "resonance" is usually a cabin reflection to EQ, not a panel to deaden (3.6 #14)

Changed views and conflicts in 3.1–3.7

Order of steps: 2011 EQ-before-delay → delay first from 2013; 2015 "two valid orders"; 2018 rule: one EQ per channel = delay first, one EQ per driver = EQ to the crossover target first (3.1 evolution #2–#6).
Starting gains: gain guide −12 dB on all DSP channels → 2022 sub-amp clip point first, then −6 dB (3.1 #2).
Where the target sits: PDFs put the midrange at 0 dB; his 2021 video, forum posts and the gain guide put the target on the dips (3.1 #7, 3.6 #3).
Sweeps vs RTA: 2012 sweeps "much better" → a live pink-noise RTA for production from 2018 (3.3 #1).
Uncorrelated noise: 2014 suggested it for 31-band work → 2020–22 useless except for upmixers, L−R rears or gaming an RTA (3.3 #4).
REW settings: PDFs FFT 16384, rectangular window, "dB not dBFS" → 2021 32k, window and units don't matter (3.3 #5).
Single mic vs spatial average: 2013 one mic worse than an MS-8-style average → 2017–2021 a smoothed single mic is ~90% as good and fine for production (3.4 #5).
Smoothing numbers: 2012–13, 2017, 2020 and 2022 give different splits; the rule (coarser above ~1 kHz) holds (3.4 #11).
Target numbers: +6–9 dB bass (2008) → +9 dB (MS-8, 2011) → +10 dB (2018 PDF); the shape never changed (3.5 #4).
Narrow EQ: blog says avoid Q above ~6; both PDFs still show a Q 10 cut at 1.6 kHz (3.6 #7).
Tweeter floor: 2009 "2× Fs, ~3 kHz for 1-in tweeters" vs PDFs 1.5–2× Fs at 2.2–2.5 kHz; 2022 forum puts the Audiofrog tweeter at 3 kHz (3.7 #2, #7).
Sub/midbass points: MS-8 era same point (80/80) → half-octave offset (60/80) from 2018 (3.7 #9).
Seats: one-seat PDF process → two-seat tuning where tools allow (2016 onward), but he won't teach all-pass two-seat tunes as a process (3.2 #13).
Focus: by 2023 he moved from teaching tuning to teaching system design by mapping the car (3.2 #15).
Tweeter capacitor: a stated precondition in PDF v1, not restated in v2, still recommended in his 2017 forum posts and Dec 2020 blog (3.1 #1, 3.7 #18).

The full series

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