Car Audio Explained, Part 7: His Step-by-Step Tuning Process and Measurement Tools (Andy Wehmeyer's Lessons)

4 days ago

40 min read

This is Part 7 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. Tuning and measurement

Andy treats tuning as measuring what is wrong and fixing it in a fixed order: polarity, distance-based delays, standard LR4 crossovers, then level and EQ to a target, with listening only at the end. His tools are cheap (a calibrated mic, free REW software, pink noise and a test CD); what he teaches is how to read the measurement, which errors are audible, and which are just reflections you should leave alone. Subsections 3.1–3.7 cover the process, the mindset, measurement, targets, EQ and crossovers; delay, polarity, level-matching and listening checks follow from 3.8.

Jargon used throughout: LR4 = Linkwitz-Riley 24 dB/octave crossover; Fs = a driver's free-air resonance (Fc = resonance in its box); Q = how narrow an EQ filter or a peak is (higher Q = narrower); RTA = real-time analyzer; REW = Room EQ Wizard (free measurement software); TA = time alignment (delay); mono / correlated pink noise = the same noise in both channels.

What this section is about: "Tuning" means adjusting the electronics that feed each speaker (usually a DSP, a digital signal processor) so the system sounds right from your seat. Andy's approach is to measure with a microphone, fix what the measurement shows in a fixed order, and only listen at the end. This part of the guide covers his step-by-step recipe, the thinking behind it, and the measuring tools he uses. The jargon line above is a glossary; each term is also explained where it first matters below.

3.1 His tuning process, step by step

What this part is about: His complete tuning recipe in the order he does it, from checking the equipment to setting the final volume and personal taste. After the recipe comes a short history of how that order changed between 2008 and 2023.

Master procedure (the order of his Tuning PDF v1 ~2018 and v2 ~2023, with his 2021–22 forum refinements; version differences flagged).

3.1 #1 · Check the prerequisites before touching anything.

The DSP needs delay on every output and separate left and right EQ; its input must be a flat, two-channel, in-phase signal, with input sensitivity matched so the A/D converter never clips. V1 also requires a protection capacitor on each tweeter, even when active. Tools: polarity checker (or the UMI-1 mic plus scope software), REW, the Audiofrog test CD. [Era: capacitor precondition not restated in v2] Sources: Tuning PDF v1 (~2018) · Tuning PDF v2 (~2023) · Audiofrog blog Mar 2021 (scopes and sine waves)

In plain English: Before any adjusting, he checks that the equipment can do the job. A DSP is the box that splits the music to each speaker and lets you adjust each one. He needs it to have delay on every output (each speaker can be held back by a tiny amount of time) and separate EQ for left and right (EQ is a set of very precise tone controls). His PDF says that without separate left and right EQ, "this process isn't going to work very well." The signal going into the DSP must be "flat" (no tone shaping already added), plain left and right, and in phase (both channels pushing the same way), like the output of an aftermarket radio. Its input sensitivity must be set so the loudest signal never overloads ("clips") the A/D converter, the part that turns the incoming signal into numbers. His 2018 PDF also asks for a protection capacitor on each tweeter (the small treble speaker) even when the DSP does the frequency splitting; a capacitor blocks low frequencies, and he treats it as a safety device. His 2023 version doesn't repeat that requirement. Tools: a polarity checker (or his UMI-1 mic with oscilloscope software), REW, and the Audiofrog test CD.

3.1 #2 · Set a safe starting gain structure.

His 2022 order: set only the sub amp, with a 0 dB sine and the DSP's sub and input levels at maximum, until it just clips; then turn every DSP input and output down 6 dB. The older gain guide found the head unit's maximum clean output first and set all DSP channels to −12 dB. Full gain detail is in section 4. [Changed view: −12 dB → −6 dB] Sources: DIYMA 2022-03-03 · DIYMA "Andy's guide to setting gains" (attributed)

In plain English: "Gain" is the input-sensitivity knob on an amplifier: how hard it amplifies the incoming signal. "Gain structure" is how the volume settings of the radio, DSP and amplifiers are set against each other so nothing distorts and hiss stays low. Clipping is what happens when a device is asked for more than it can give: the tops of the waveform get chopped off, which sounds harsh. His 2022 order: play a 0 dB sine wave (a pure test tone at the maximum digital level) with the DSP's sub output and input levels at maximum, and set only the subwoofer amp's gain so it just clips. Then turn every DSP input and output down 6 dB, and put the test tones away. His older gain guide instead started by finding the head unit's (radio's) maximum clean output, then set all DSP channels to −12 dB. So the starting cut changed from 12 dB to 6 dB over time. Section 4 has the full gain method.

3.1 #3 · Confirm every driver's polarity with a checker, the markings or the UMI-1 and a scope.

Never set polarity by listening for the center image. Sources: Tuning PDF v1 (~2018) · Tuning PDF v2 (~2023)

In plain English: Polarity is which way a speaker's cone moves for a given signal. Every speaker and amp has + and − terminals; if one speaker's wires are swapped, it pushes while the others pull, and their sound partly cancels. He confirms each driver ("driver" just means a single speaker unit) with a polarity checker (a small tool that shows whether the cone moves out or in), with the + and − markings, or with his UMI-1 mic and oscilloscope software, which shows the shape of the signal. His rule is firm: check it with a tool, never by listening to whether the center image sounds right. The center image is the illusion that a singer stands in the middle of the dashboard.

3.1 #4 · Strap the mic to the headrest at the listening position, then run the cable out and close the door.

He measures from outside the car (see 3.4 #1). Sources: Tuning PDF v1 (~2018) · UMI-1 video (Nov 2021) @11:02

In plain English: The measuring microphone goes where your head would be in the driver's seat, strapped to the headrest. He then runs the mic cable out of the car and shuts the door, and works the computer from outside. Section 3.4 #1 gives his reasons: his target curve assumes an empty car, and working from outside spares you a long session of pink noise and keeps the laptop from blocking speakers.

3.1 #5 · Tape-measure from each grille or dust-cap center to the mic and enter the delays; then leave them alone.

The sub is a straight line to the mic whichever way it faces; with passive crossovers measure to the midbass (3-way) or midrange (2-way). Check what your DSP wants: time, distance, or each path subtracted from the longest. Sources: Tuning PDF v2 (~2023) · OFCAP Ep. 19 @3:17:02 · DIYMA 2021-08-26

In plain English: In a car you sit much closer to the speakers on your side of the car than to the ones on the far side, so sound from the near side reaches you first and pulls the image toward it. A DSP can hold back (delay) the nearer speakers so that sound from every speaker arrives at your head at the same moment. This is often called time alignment. He sets it with a tape measure: measure from the center of each speaker's grille or dust cap (the dome in the middle of a woofer cone) to the mic, enter those numbers, and then leave them alone. Sound travels about 34 cm (a little over 1 foot) per millisecond, which is why distance converts directly into delay. The subwoofer is measured as a straight line to the mic, whichever way it faces. If the front speakers share a passive crossover (a small box of parts that splits one amp channel between a woofer and a tweeter), measure to the midbass in a 3-way system or the midrange in a 2-way. DSPs differ in what they want typed in: a time, a distance, or each path subtracted from the longest one. Check yours.

3.1 #6 · Enter LR4 crossovers on every driver, with the sub low-pass half an octave below the midbass high-pass.

Typical: sub 60 Hz low-pass, midbass 80 Hz high-pass, tweeter at about 1.5–2× its Fs, nothing but the sub below its resonance (details in 3.7). Sources: Tuning PDF v1 (~2018) · Tuning PDF v2 (~2023)

In plain English: A crossover splits the music by frequency so each speaker gets only the range it handles well: lowest bass to the sub, upper bass to the midbass, treble to the tweeter. A low-pass filter lets lows through; a high-pass lets highs through. LR4 (Linkwitz-Riley, 24 dB per octave) is a standard, fairly steep filter shape whose two halves add back together smoothly. He uses LR4 on every driver. He sets the sub's low-pass half an octave below the midbass high-pass; his typical numbers are sub low-pass at 60 Hz and midbass high-pass at 80 Hz. He crosses the tweeter at about 1.5 to 2 times its Fs (the frequency where it naturally resonates), and no speaker except the sub is asked to play below its own resonance. Section 3.7 gives the details.

3.1 #7 · Calibrate the RTA to the target, then don't touch the volume again.

Play mono pink noise at normal listening level, offset REW so the midrange sits at 0 dB on the imported Audiofrog target. His 2021 video and the gain guide instead slide the target down onto the response's dips so EQ is mostly cutting (3.6 #3). REW settings and offset steps in 3.3 #5–#6. [Conflict: midrange-at-0 dB (PDFs) vs target-on-the-dips (video, forum)] Sources: Tuning PDF v2 (~2023) · UMI-1 video @16:00

In plain English: An RTA (real-time analyzer) is a live graph showing how loud each frequency band is at the mic. The target curve is the shape he wants that graph to follow; Audiofrog supplies it as a file you load into REW. Play mono pink noise (the same steady hiss in both channels) at your normal listening level, then shift REW's display so the midrange part of the measured curve sits at 0 dB on the target. After that, don't touch the volume, or the measurement and the target no longer line up. The [Conflict] tag notes two ways he has described this step. In the PDFs, the midrange is put at 0 dB. In his 2021 video and the gain guide, the target is slid down until it sits on the dips (low points) of the measured response, so that most EQ changes are cuts rather than boosts. Section 3.3 #5–#6 gives the exact REW settings.

3.1 #8 · Left side plus sub, mono pink noise: get close to the target with channel levels first, then EQ.

Save the measurement. His PDF example: sub up about 3.5 dB, midrange unchanged, one tweeter up about 6 dB, then EQ. Sources: Tuning PDF v1 (~2018) · DIYMA 2022-03-03

In plain English: Now play mono pink noise through only the left speakers plus the sub, and watch the RTA. First adjust each speaker's overall level (its volume) to get close to the target, and only then use EQ for the finer shape. The gain guide attributed to him says doing levels first gets you partway to the target and makes the EQ faster. A dB (decibel) is the unit for level differences. In his PDF example the sub went up about 3.5 dB, the midrange stayed where it was, and one tweeter went up about 6 dB, before any EQ. Save this measurement, because the right side will be matched to it.

3.1 #9 · Right side plus sub: match the saved left within "a couple of dB."

If one side has a dip you can't fix, copy it to the other side. Sources: Tuning PDF v1 (~2018) · Tuning PDF v2 (~2023)

In plain English: Repeat the same thing for the right speakers plus the sub, matching the right side to the saved left measurement within "a couple of dB." If one side has a dip that can't be fixed, he copies that same dip onto the other side. His reasoning comes from 3.2 #2: matching left to right is the priority, because a matched left and right response is part of his definition of correct stereo.

3.1 #10 · Both sides, mono pink noise: a check, not an EQ step.

Dips of 9 dB or more that appear only in the sum mean wrong delays or delay/phase EQ already in the source. Don't EQ the sum, except a peak above 6 dB cut equally on both sides or a global change such as more bass. Gain guide: set sub level here (about +6 dB if only one input channel feeds the sub). Sources: Tuning PDF v2 (~2023) · DIYMA 2020-07-08 · DIYMA gain guide (attributed)

In plain English: Next, play mono pink noise through both sides at once. This step is only a check, not a place to EQ. If the combined curve shows dips of 9 dB or more that weren't there on either side alone, he says the delays are wrong, or the signal coming into the DSP already has delay or phase-altering EQ in it. Don't EQ this combined curve. The only exceptions are a peak more than 6 dB high, which you cut by the same amount on both sides, and a global change such as more bass. His gain guide sets the sub level at this point; if only one input channel feeds the sub, expect to raise it about 6 dB.

3.1 #11 · Now listen: band-limited pink noise, then panned tracks.

Test-CD tracks 2–5 should each sound centered; tracks 6–20 (pink noise, bass guitar, piano) should move evenly across the dash. Sources: Tuning PDF v1 (~2018) · UMI-1 video @27:01

In plain English: Only now does he listen. Band-limited pink noise is hiss covering just one slice of the frequency range. On the Audiofrog test CD, tracks 2–5 are such slices, and each should sound like it comes from the center of the dash. Tracks 6–20 are pink noise, bass guitar and piano that are "panned," meaning moved gradually from one side to the other. They should move evenly across the dash.

3.1 #12 · Center the image with level only.

Turn down the whole channel on the side the image leans to; never use delay. If the center is smeared or off to one side after the objective steps, suspect a delay-entry error first. Sources: Tuning PDF v2 (~2023) · Absolute Electronix livestream (Mar 2023) @43:32

In plain English: If the center image leans to one side, he fixes it with level only. Turn down the whole channel, meaning all the speakers on the side the image leans toward, by the same amount. Never use delay for this. His 2015 forum advice (3.1 evolution #4) explains the risk: a level error pulls the image, and "fixing" it with delay breaks other frequencies. If, after all the measured steps, the center sounds smeared (spread out) or sits off to one side, his first suspect is a mistake when the delays were typed in.

3.1 #13 · If the bass isn't in front:

look for 60–160 Hz dips, then lower the sub slightly; only as a last resort delay all front speakers by exactly the same amount, gradually, with track 25. Sources: Tuning PDF v1 (~2018) · Tuning PDF v2 (~2023)

In plain English: Sometimes the bass seems to come from behind you, where the subwoofer is, instead of from the front with the rest of the music. His order of fixes: first look for dips between 60 and 160 Hz, around where the sub hands over to the front speakers. Then try lowering the sub's level slightly. Only as a last resort, delay all the front speakers by exactly the same amount, a little at a time, while listening to track 25 of the test CD.

3.1 #14 · Set final volume and noise without changing the curve.

Silent track 38: if hiss bothers you, lower all amp input sensitivities by the same amount; for more volume raise all equally until hiss appears. His 2022 order: raise all DSP outputs equally, then inputs, then every amp gain equally, watching the RTA so the curve doesn't move. Sources: Tuning PDF v1 (~2018) · DIYMA 2022-03-03

In plain English: Track 38 of the CD is silence, used to listen for hiss. If the hiss bothers you, lower all the amplifier input sensitivities (gains) by the same amount. If you need more volume, raise them all equally until hiss appears. Moving everything by the same amount keeps the balance you just tuned. His 2022 order for adding volume is: raise all DSP outputs equally first, then the DSP inputs, then every amplifier gain equally, watching the RTA the whole time so the curve doesn't change shape.

3.1 #15 · Preference changes last, through the target or the DSP's input EQ, so both sides change identically.

Preferences change the target, never the process. Sources: Tuning PDF v2 (~2023) · DIYMA gain guide (attributed)

In plain English: Personal taste, such as wanting more bass, comes last. He makes those changes through the target curve or through the DSP's input EQ, which acts on the signal before it is split to the speakers, so both sides change identically. His principle: your preference changes the target you aim for, never the steps you follow.

How his procedure evolved

3.1 evolution #1 · 2008–2009: delay first (with low-passes off), then an RTA to find each driver's usable bandwidth in place, then one channel to the target.

This manual method was written to copy what the MS-8 automated; 2008 tips add that a measurement 20 dB above background noise is loud enough. [Era: Harman/MS-8] Sources: DIYMA 2009 · DIYMA 2008

In plain English: In 2008–2009 his manual order was: set the delays first (with the low-pass filters switched off), then use an RTA to find the range each speaker can actually play well where it is mounted, then EQ one channel to the target. He wrote it to copy, by hand, what the MS-8 did automatically. The MS-8 was a JBL car processor from his Harman years that tuned itself with a microphone. His 2008 tips add that a measurement is loud enough when it is 20 dB above the background noise. This item belongs to that Harman/MS-8 period.

3.1 evolution #2 · 2010–2011: EQ each side alone; in 2011 he even put EQ before delay.

2010: EQ right+sub with left off, then left+sub, and largely ignore the both-sides RTA above about 150 Hz. 2011: EQ left, match right (attenuating the nearer speaker), then delay by tape measure. [RARE] [Changed view: EQ-before-delay, later reversed] Sources: DIYMA 2010 · DIYMA 2011

In plain English: In 2010 he EQ'd each side on its own: the right side plus sub with the left side off, then the left plus sub. He said to largely ignore the both-sides RTA above about 150 Hz. In 2011 he even put EQ before delay: EQ the left side, match the right side to it (turning down the nearer speaker), and only then set delays by tape measure. The tags flag this as unusual and as a view he later reversed; his later process sets delays first. Many people believe you should EQ with both sides playing together; his 2010 method of EQ'ing each side alone went against that.

3.1 evolution #3 · 2013–2014: delay, crossovers, EQ, recheck, "drink beer."

Delay comes first in multi-amped systems because it changes the crossover sums; flip a polarity only for a big hole at the crossover. Sources: DIYMA 2013 · DIYMA 2014

In plain English: By 2013–2014 his order was: delay, crossovers, EQ, recheck, then "drink beer" (his joke for: you're done). His reason for doing delay first in multi-amped systems (one amplifier channel per speaker, split by the DSP) is that delay changes how neighboring speakers add up around their crossover frequency. He would flip a speaker's polarity only if there was a big hole (deep dip) at the crossover.

3.1 evolution #4 · 2015: two valid orders.

Standard: delays by distance, crossovers, then EQ and levels. If you insist on delay by ear, do crossovers and exact EQ first and delay last, because a level error pulls the image and "fixing" it with delay breaks other frequencies. Sources: DIYMA 2015 · DIYMA 2015 (shape drivers to LR4)

In plain English: In 2015 he described two acceptable orders. The standard one: set delays by measured distance, then crossovers, then EQ and levels. If someone insists on setting delay by ear, he said to do crossovers and exact EQ first and delay last. The reason: a level error pulls the image to one side, and if you "fix" that with delay, you break other frequencies.

3.1 evolution #5 · MS-8 bring-up (2010–2015): tune a 2-way stereo front first, then add pieces.

Baseline: subsonic 20 Hz 12 dB/oct, sub 80 Hz 24 dB/oct, front 2-way at 3.5 kHz 24 dB/oct, amp crossovers off, Logic 7 off. [Era: MS-8] Sources: DIYMA MS-8 FAQ · DIYMA 2015

In plain English: This is how he brought up an MS-8 system (2010–2015): get a simple 2-way stereo front working first, then add the other pieces. His starting settings were a subsonic filter at 20 Hz, 12 dB per octave (a filter that removes the very lowest bass), the sub at 80 Hz, 24 dB per octave, and the front 2-way crossed at 3.5 kHz, 24 dB per octave. Crossovers on the amplifiers were switched off so only the MS-8 did the filtering, and Logic 7 (Harman's surround feature that creates extra channels from stereo) was off. This applies to that one product and period.

3.1 evolution #6 · 2018 rule for which comes first.

One EQ per channel: delays first. One EQ per driver: EQ each driver to its crossover target, then delay. The 2018 PDF's process is its first two pages; the next ~60 explain why. Sources: DIYMA 2018-05-22 · Tuning PDF v1 (~2018)

In plain English: His 2018 rule depends on the DSP. If it has one EQ per channel (one EQ shared by all the left speakers, for example), set the delays first. If it has one EQ per driver, first EQ each speaker so its response matches its intended crossover shape, then set the delays. He also notes that his 2018 PDF's actual process fits on its first two pages; the next roughly 60 pages explain why it works.

3.1 evolution #7 · 2021–2023: same steps, refined.

His REW guide was written for REW 5.19 (a new one was in progress in 2021); 2022 added the sub-clip-first gain order (step 2); v2 (2023) keeps the steps. By 2023 he said everyone now talks about tuning, so he would teach system design instead (3.2 #15). Sources: DIYMA 2021-07-15 · Tuning PDF v2 (~2023) · Absolute Electronix livestream (Mar 2023) @1:00:14

In plain English: From 2021 to 2023 the steps stayed the same, with small refinements. His REW guide was written for REW version 5.19, and a new guide was in progress in 2021. In 2022 he added the "set the sub amp to clip first" gain order (step 2 above). The 2023 PDF keeps the same steps. By 2023 he said that everyone now talks about tuning, so he would rather teach system design (see 3.2 #15).

3.2 How he thinks about tuning

What this part is about: The mindset behind his process: what "correct" means, why he measures before he listens, why he distrusts tricks and endless ear-tuning, and how he judges whether a tuning method is any good.

3.2 #1 · Measure and fix everything objectively first; listen once, at the end.

He never listens until the measurements say it's worth it, and won't listen after each step ("you don't taste raw chicken after salting it"). He does listen at the end; if something is wrong he goes back to the same instruments, not to random tweaks. Hours of touch-up mean a wrong measurement, a misunderstood one, or a useless process. [RARE] Sources: DIYMA 2017-06-28 · DIYMA 2021-07-09 · UMI-1 video @24:03

In plain English: His rule is to measure and fix everything that can be measured first, and to listen only once, at the end. He doesn't listen after each step, and compares that habit to tasting raw chicken right after salting it: the dish isn't finished, so the taste tells you nothing useful yet. He does listen at the end. If something sounds wrong, he goes back to the same measuring instruments to find the cause, instead of making random tweaks. If someone needs hours of touch-up by ear, he says the measurement was wrong, was misunderstood, or the process itself is useless. Many people believe tuning is mainly done by ear, listening after every change, which is why this point is tagged as unusual.

3.2 #2 · "Correct" stereo has just three conditions: left and right match in level, in frequency response, and in arrival time (phase) at the seat.

Everything else (tools, targets, tricks) is only useful insofar as it gets you there; a windshield rake or dash mat calls for an adjustment, not a new method. Sources: Mobile Electronics 2017 (tag page) · DIYMA 2017-06-28 · DIYMA 2020-10-03

In plain English: For him, "correct" stereo comes down to three things at your seat: the left and right sides match in loudness, in frequency response (how loud each pitch is, from deep bass to high treble), and in arrival time (phase). Background: when two identical speakers play the same sound at equal loudness and timing, your brain hears it coming from midway between them, which is how a stereo "image" works. In a car you sit off to one side, so tuning restores that match. Tools, target curves and tricks only matter if they get you to those three conditions. A windshield's rake (its slope) or a dash mat calls for an adjustment within the method, not a new method.

3.2 #3 · You don't "tune" a system; you measure its deficiencies and fix them.

He repeats an acoustician friend's line; if the last 20% of your process is choosing which errors you prefer, the objective part needs work. Shops should bill for an efficient, predictable troubleshooting process. [RARE] Sources: 361 VR Cadillac video @3:01 · DIYMA 2018-05-04 · Mobile Electronics Aug 2017 (DSP Tuning Part 3)

In plain English: He repeats a line from an acoustician friend: you don't "tune" a system, you measure what's wrong with it and fix that. If the last 20% of your process is choosing which errors you prefer to keep, the measured, objective part of your process needs more work. For shops, his view is that they should charge for an efficient, predictable troubleshooting process. Many people believe tuning is a personal art of taste; he treats it more like diagnosing and repairing faults.

3.2 #4 · There are no tuning "tricks," only a process; pick one you understand and don't splice in other people's steps.

A barrel of tips from a contest winner isn't a process, and if someone can't say why a trick works he doesn't know that it works (he posted a 29-step parody of ear-tuning). Mixing teachers gives an "amalgam"; never experiment on customer cars. [RARE] Sources: DIYMA 2018-07-02 · DIYMA 2021-07-09 · OFCAP Ep. 19 @3:19:02

In plain English: He says there are no tuning "tricks," only a process. Pick one process you understand and follow it, rather than splicing in steps from other people. A collection of tips from a contest winner is not a process, and if someone can't explain why a trick works, he says they don't actually know that it works. (He once posted a 29-step parody of ear-tuning to make the point.) Mixing several teachers' methods gives an "amalgam," a mixture that isn't any one coherent method. And never experiment on customer cars.

3.2 #5 · When "the process didn't work," find the one thing done wrong.

Every failure he has seen came from a skipped or inserted step: a flipped speaker, a 1 kHz peak in one channel, distances typed where the DSP wanted delays. "Cars do strange things" is a cop-out; if tape-measured delays don't give the expected phase, something else is changing phase. [RARE] Sources: OFCAP Ep. 19 @3:17:02 · DIYMA 2021-08-21 · Absolute Electronix livestream @44:23

In plain English: When someone says "the process didn't work," he looks for the single thing that was done wrong. Every failure he has seen came from a step that was skipped or one that was added. His examples: a speaker wired with flipped polarity, a peak at 1 kHz left in only one channel, or distances typed in where the DSP wanted delay times. He calls "cars do strange things" a cop-out. If delays set by tape measure don't give the expected phase (timing) result, something else in the system is changing the phase, and that is what you should look for.

3.2 #6 · Ear-only tuning never converges, because each track excites different frequencies.

You cut what annoys you on one song, then the next, until "there's no more midrange"; if you keep retuning per track, look for an unfixed high-Q peak. He calls setting things by ear "nuts" and ear-tuning aimless fiddling. [RARE] Sources: DIYMA 2014 · DIYMA 2015 · DIYMA 2021-05-12

In plain English: "Converge" means settling on a final answer. He says tuning by ear alone never does, because every song emphasizes different frequencies. You cut whatever annoys you on one song, then on the next, and the next, until "there's no more midrange." If you keep re-tuning for each track, he says to look for a narrow (high-Q) peak you haven't fixed yet. He calls setting things by ear "nuts" and describes ear-tuning as aimless fiddling. Many people believe you can tune well by ear alone; this point is his answer to that.

3.2 #7 · Judge a method by whether it reliably ends at "correct," then by how teachable and fast it is.

Repeatability against a defined target is the engineering meaning of quality; for a shop booked six weeks out, a 1-hour process beats a 16-hour one. Tuners who insisted on 10 hours finished his way in about 1 h 20 min; the objective steps take him 30–40 minutes. Sources: DIYMA 2011 · OFCAP Ep. 19 @3:19:02 · Absolute Electronix livestream @43:32

In plain English: He judges a tuning method first by whether it reliably ends up at "correct," then by how easy it is to teach and how fast it is. In engineering, he says, quality means repeatably hitting a defined target. For a shop booked six weeks ahead, a 1-hour process is better than a 16-hour one. He reports that tuners who insisted they needed 10 hours finished his way in about 1 hour 20 minutes, and that the measured, objective steps take him 30–40 minutes.

3.2 #8 · Professional tuning needs an agreed endpoint, and you stop when you know what's achievable.

Billing hours with no defined end is a painter who can't name the color; a sane front-plus-sub system takes him an hour or two. Stop at diminishing returns, not when ideas run out; the threat is "imagine how much better it would sound if…". Don't retune for each recording. [RARE] Sources: DIYMA 2015 · DIYMA 2022-03-03

In plain English: A professional tune needs an end point agreed in advance. Billing hours with no defined end is, in his comparison, like a painter who can't name the color he's painting. A sensible system of front speakers plus a sub takes him an hour or two. Stop when you reach diminishing returns (more work gives very little improvement), not when you simply run out of ideas. The trap is the endless thought "imagine how much better it would sound if…". And don't re-tune for each recording.

3.2 #9 · EQ is the main lever; swapping speakers is "applying a nonadjustable EQ."

About 90% of problems are EQ-fixable (not distortion); cheap speakers with great EQ beat great speakers with poor EQ, and most complaints about a brand's "voicing" vanish with a DSP. Replace speakers only if broken, underpowered or distorting. [RARE] Sources: DIYMA 2008 · DIYMA 2016 · DIYMA 2022-01-14

In plain English: EQ (precise tone adjustment in the DSP) is his main tool. He describes swapping speakers as "applying a nonadjustable EQ": a different speaker mostly gives you a different tonal balance, which you can't adjust. He says about 90% of problems can be fixed with EQ (the exception being distortion). Cheap speakers with great EQ beat great speakers with poor EQ, and most complaints about a brand's "voicing" (its typical tonal character) disappear once a DSP is used. He would replace speakers only if they are broken, underpowered or distorting. Many people believe new speakers are the main way to better sound, so this is tagged as unusual.

3.2 #10 · The speaker does what it does; everything else you measure at the seat is the car (or a filter you added).

A two-channel RTA proves it: amp output and the speaker at 1 inch look like the source; the driver's seat with the door shut looks very different. He teaches techs to separate "speaker," "filters" and "car" so they can predict any location. [RARE] Sources: DIYMA 2011 · DIYMA 2021-08-06 · DIYMA 2021-08-06 (troubleshooting)

In plain English: His point is that the speaker itself does what it does; everything else you measure at the seat comes from the car (or from filters you added in the DSP). He says a two-channel RTA, which compares two signals at once, proves it. The amplifier's output, and the speaker measured 1 inch (about 25 mm) away, both look like the source signal. At the driver's seat with the door shut, the curve looks very different. He teaches technicians to separate three things, "speaker," "filters" and "car," so they can predict what will happen in any mounting location.

3.2 #11 · A car is not a living room or a concert hall, so imported home and pro methods mislead.

Speakers and "room" are EQ'd together from one measurement set because seats are fixed and no reflection is late. Imports he mocks: hundredth-of-an-inch delays, 1/48-octave EQ, thousands of averages, first-order "phase-coherent" crossovers, no tweeters, aiming to "eliminate" reflections. [RARE] Sources: DIYMA MS-8 FAQ · DIYMA 2020-07-08

In plain English: He says a car is nothing like a living room or a concert hall, so methods imported from home audio or professional sound mislead. In a car, the speakers and the "room" (the cabin) are EQ'd together from one set of measurements, because the seats are fixed and no reflection arrives late. (Background: in a large room, echoes from distant walls arrive noticeably after the direct sound; in a car, every surface is close, so reflections arrive almost together with it.) The imported habits he mocks include: delays set to a hundredth of an inch, EQ with 1/48-octave precision, thousands of averaged measurements, gentle first-order "phase-coherent" crossovers, systems with no tweeters, and trying to "eliminate" reflections.

3.2 #12 · Cars are more alike than different, so expertise is knowing the general case.

"Every car is different, just try things" isn't expertise; knowing a DSP's menus isn't knowing audio. Cover "the 95%" and outliers stand out, which is how he predicts a car's problems before listening. Simplicity beats sophistication: complexity isn't useful until it can be explained simply. [RARE] Sources: DIYMA 2021-12-22 · DIYMA 2017-06-28 · Audiofrog blog Jan 2021 (crossovers)

In plain English: He says cars are more alike than different, so real expertise is knowing the general case. "Every car is different, just try things" is not expertise, and knowing a DSP's menus is not the same as knowing audio. If you understand what is true for "the 95%" of cars, the unusual cases stand out, and that is how he predicts a car's problems before he even listens. He prefers simplicity to sophistication: complexity isn't useful until it can be explained simply.

3.2 #13 · His process is a one-seat tune; two-seat all-pass tunes work but he won't teach them.

The default centers the image for the driver. He sometimes does all-pass two-seat tunes, but they take fiddling and listening, so they aren't a repeatable process. On a later podcast he called "forget the passenger" a cop-out (details from 3.8). [Changed view: one-seat (PDFs) → two-seat where tools allow (2016 onward)] Sources: Tuning PDF v1 (~2018) · DIYMA 2021-09-07 · OFCAP Ep. 38 @52:04

In plain English: His written process tunes for one seat, the driver's, so the image is centered for the driver. "Two-seat" tuning tries to make the image work for driver and passenger at once, often with all-pass filters (DSP filters that change the timing of some frequencies without changing their loudness). He says these can work and he sometimes does them, but they need fiddling and listening, so they are not a repeatable process and he won't teach them. The [Changed view] tag records two positions: his PDFs present a one-seat tune, while from 2016 onward, where the DSP's tools allow it, he moved toward tuning for both seats, and on a later podcast he called "forget the passenger" a cop-out. Section 3.8 has the details.

3.2 #14 · "Correct" is the beginning of expertise, not the end.

Learn to reach correct, then learn to back off it for customers who want something else; a contest win only proves judges liked it. Spatial errors bother him more than tonal ones: a missing center image is worse than a peak or dip. [RARE] Sources: CMA Connected (Oct 2021) @42:16 · DIYMA 2016 · SQOLOGY P2 @46:07

In plain English: Reaching "correct" is where expertise starts, not where it ends. First learn to get a system to correct; then learn how to move away from it on purpose for a customer who wants something different. Winning a contest only proves that the judges liked it. He says errors in where sounds appear (spatial errors) bother him more than errors in tone: a missing center image is worse to him than a peak or a dip in the response.

3.2 #15 · Tuning begins with system design; by 2023 he taught mapping the car with a mic and speaker before choosing locations.

Measure every candidate spot (dash top, A-pillar, kick, door top, trunk) instead of guessing rules like "avoid the windshield corner"; he did this for his son's car and the Volvo C30 demo car. [RARE] [Era: 2021–2025] Sources: CMA Connected (Oct 2021) @41:52 · Absolute Electronix livestream (Mar 2023) @1:00:14 · OFCAP Ep. 38 @36:00

In plain English: He says tuning begins with system design: deciding where the speakers go. By 2023 he taught mapping the car first, by placing a speaker in each candidate spot (dash top, A-pillar, kick panel, door top, trunk) and measuring it with a mic before choosing locations. (The A-pillar is the post beside the windshield; the kick panel is the side panel near your feet.) He prefers measuring to following rules of thumb such as "avoid the windshield corner." He did this for his son's car and for the Volvo C30 demo car. This reflects his teaching from about 2021–2025.

3.2 #16 · Train your ears on two speakers in phase vs out of phase, not on concerts.

Many tuners have never sat between two speakers and flipped one's polarity. Practice with any matched $5 thrift-store pair in an equilateral triangle: imaging is just what two identical speakers do for a centered listener. [RARE] Sources: Absolute Electronix livestream @41:56 · Absolute Electronix livestream @46:15

In plain English: He says to train your ears on two speakers playing in phase versus out of phase, not by going to concerts. Many tuners, he says, have never sat between two speakers and flipped the polarity of one. Practice is cheap: any matched pair of $5 thrift-store speakers, set up so you and the two speakers form an equilateral triangle (all three sides the same length). His point is that imaging is simply what two identical speakers do for a listener sitting in the center, so hearing that effect directly teaches you what to listen for. Many people believe the best ear training is listening to live concerts; that is the common advice he disagrees with.

3.2 #17 · An auto-tune is a spec plus a hit rate.

Judge any automatic optimizer by how often it hits its target across systems; the MS-8 worked within its 48 supported layouts and broke on edge cases, and no auto-tune beats every manual process every time. [Era: Harman/MS-8] Sources: DIYMA 2016

In plain English: An auto-tune is a processor that measures and tunes the system by itself. He says to judge any such optimizer by its specification plus its hit rate: how often it reaches its target across many different systems. The MS-8 (the JBL self-tuning processor from his Harman years) worked well within the 48 speaker layouts it supported and broke on unusual edge cases. He adds that no auto-tune beats every manual process every time. This point relates to the MS-8 era.

3.3 Measurement tools and setup

What this part is about: The measuring kit and how to set it up: which test signal to use, which microphone, how to set up the free REW software, and the common ways a measurement can mislead you.

3.3 #1 · For production tuning use a live RTA on mono pink noise, not sweeps or sine tones.

An RTA shows the curve move as you EQ; a sweep or a tone excites each narrow comb peak and dip one at a time and misleads in a car ("Pure tones aren't good tools"). He still uses REW sweeps for research and phase. [Changed view: in 2012 he thought REW sweeps worked "much better"] Sources: Audiofrog blog Jan 2021 (single vs multi-mic) · DIYMA 2021-09-03 · DIYMA 2012

In plain English: "Production tuning" means everyday tuning of customer cars in a shop. For that he uses an RTA (a live graph of loudness across frequencies) fed with mono pink noise (steady hiss, the same in both channels). The RTA updates live, so you see the curve move as you change the EQ. A sweep (a tone that glides from low to high pitch) or single test tones hit the narrow peaks and dips one at a time and give a misleading picture in a car. He sums it up as "Pure tones aren't good tools." Background: those narrow peaks and dips come from comb filtering, which happens when the same sound reaches the mic twice with a tiny time gap (directly and off a nearby surface), so some frequencies add up and others cancel. He still uses REW sweeps for research and for measuring phase (timing). The [Changed view] tag notes that in 2012 he thought REW sweeps worked "much better."

3.3 #2 · Pink noise, not white.

Pink noise has equal power per octave, sounds even from 20 Hz to 20 kHz and resembles music's energy spread; the RTA shows it flat by applying the inverse 3 dB/octave slope. White noise sounds bright and isn't used for car tuning. Sources: Audiofrog blog Dec 2020 (pink noise)

In plain English: Pink noise is a hiss with equal power in every octave (an octave is a doubling of frequency, such as 100 to 200 Hz or 1,000 to 2,000 Hz). It sounds even from 20 Hz to 20 kHz and resembles how music's energy is spread. On many analyzers it would show as a line sloping downward, so the RTA tilts the display the opposite way (the "inverse 3 dB/octave slope") to make pink noise read flat. White noise has equal power at every single frequency, which piles up energy in the treble, so it sounds bright. He says white noise isn't used for car tuning.

3.3 #3 · Never tune or center the image with 31 sine waves; if you must use your ears, use band-limited pink noise or music.

Above 1 kHz the comb filtering is so dense that a tone easily lands in a null, and the nulls differ left to right; that's why the 1980s pure-tone centering method fails ("really stupid"). [RARE] Sources: DIYMA 2021-07-08 · DIYMA 2017-07-12 · DIYMA 2020-07-08

In plain English: An old method was to play 31 pure tones (one for each band of a classic 31-band EQ) and adjust until each sounded centered. He says never tune or center the image this way. If you must use your ears, use band-limited pink noise (hiss covering only a slice of the range) or music. His reason: above 1 kHz the comb filtering (narrow cancellations from reflections) is so dense that a single tone easily lands in a null, a spot where it almost cancels, and the nulls fall at different frequencies on the left and right. That is why the 1980s pure-tone centering method fails; he calls it "really stupid." Many old-school competitors believed tone-by-tone centering was the right way, which is why this point is tagged as unusual.

3.3 #4 · Use mono (correlated) noise; uncorrelated noise has no center, so it can't show the summed image.

With one side playing the noise type doesn't matter; with both, you're measuring their sum. Uncorrelated noise is only for making everything but the center play in an upmixer, driving an L−R rear feed, or "gaming" an IASCA RTA. To measure L−R rears, play one channel (or pan mono noise fully). [RARE] [Changed view: in 2014 he suggested stereo noise for 31-band adjustments] Sources: DIYMA 2021-10-08 · DIYMA 2020-07-08 · DIYMA 2014

In plain English: Correlated (mono) noise is the identical signal in both channels, so with both sides playing it forms a sound in the center. Uncorrelated noise is a different random hiss in each channel, so it has nothing in common between the sides and no center. He says to use mono noise: with one side playing, the type of noise doesn't matter, but with both sides playing you are measuring their sum, and only correlated noise shows that. Uncorrelated noise has only a few uses: making every speaker except the center play in an upmixer (a processor that creates center and rear channels from stereo), driving rear speakers fed with L−R (the difference between left and right), or "gaming" the RTA test in IASCA competitions. To measure L−R rears, play one channel only, or pan mono noise fully to one side. The [Changed view] tag notes that in 2014 he suggested stereo noise for adjusting a 31-band EQ.

3.3 #5 · His REW RTA settings.

PDFs: 1/3-octave RTA (1/12 to measure a peak's width); FFT 16384, 16 averages, rectangular window, 50% overlap, update interval 2; 5 dB per division; 20 Hz–20 kHz; bars on. In 2021 he said use 32k, window and 64k make no difference, and dB vs dBFS doesn't matter. [Changed view: FFT 16k → 32k; "dB not dBFS" → either] Sources: Tuning PDF v2 (~2023) · DIYMA 2021-12-05 · DIYMA 2021-07-15

In plain English: These are his settings for REW's RTA. In his PDFs: 1/3-octave display (switch to 1/12-octave when you need to measure how wide a peak is); FFT length 16384; 16 averages; rectangular window; 50% overlap; update interval 2; 5 dB per division on the graph; range 20 Hz to 20 kHz; bars turned on. Background: FFT length is how big a chunk of sound the analyzer studies at once (bigger means finer detail); averages smooth out the jumpy display; dB per division is the vertical scale of the grid. In 2021 he said to use 32k, that the window type and a 64k length make no difference, and that showing dB or dBFS (decibels relative to digital full scale) doesn't matter. Those are the changes the [Changed view] tag records.

3.3 #6 · Display calibration in REW.

Read the pink-noise level midway between the loudest and quietest bands (his example 75 dB), enter it under Measurement Offsets "All SPL," click "Add Offset to Data," then import "Audiofrog Target Curve for REW.csv" (answer "No") and check the midrange sits at 0 dB. For "whack-a-mole" hunting he once preferred 3 dB per division. Sources: Tuning PDF v1 (~2018) · DIYMA 2008

In plain English: This sets REW's display so your measurement can be compared with the target curve. Play pink noise and read the level midway between the loudest and quietest bands; his example is 75 dB. Type that number into Measurement Offsets, "All SPL," and click "Add Offset to Data," which shifts the measurement so it sits around 0 dB. Then import the file "Audiofrog Target Curve for REW.csv," answer "No" to the prompt that appears, and check that the midrange sits at 0 dB. "Whack-a-mole" tuning is his name for the problem where fixing the sound on one track breaks another. To hunt down the serious peaks and dips behind it, he once (around 2008–2009) preferred a finer graph scale of 3 dB per division instead of the default 10.

3.3 #7 · You need a known mic, not an expensive one.

Software applies the mic's calibration file, so a cheap calibrated capsule equals a precision mic (his first Harman rig was a ~$6,000 DOS "lunchbox"). His 11 mics, from ACO Pacific and Earthworks to UMIK-1, ECM8000 and UMI-1, read within about 1 dB except at the very top; avoid the APEX 220. A "true RTA" is just an RTA. [RARE] Sources: DIYMA 2021-08-16 · UMI-1 video @9:01 · DIYMA 2016

In plain English: He says you need a microphone whose errors are known, not an expensive one. Each measurement mic can come with a calibration file listing its own small errors, and the software corrects for them, so a cheap calibrated capsule measures as well as a precision mic. For contrast, his first measurement rig at Harman was a roughly $6,000 DOS-computer "lunchbox." He owns 11 mics, from ACO Pacific and Earthworks to the UMIK-1, ECM8000 and UMI-1, and they read within about 1 dB of each other except at the very top of the range. He says to avoid the APEX 220. He adds that a "true RTA" is just an RTA. Many people believe good measurements need an expensive mic or special analyzer, so this is tagged as unusual.

3.3 #8 · No 90° calibration file is needed in a car, and pointing direction hardly matters.

A car is a pressure field at low frequencies and random-incidence (diffuse) at high ones, so an omni pointed up or forward is fine; 90° files are for free-field work. Most such mics use similar 1/4-in electret capsules, and max SPL is set by bias voltage (5 V USB gives a bit over 100 dB). [RARE] Sources: DIYMA 2020-10-03 · DIYMA 2021-10-26

In plain English: Some calibration files come in a 0° version (mic pointed at the speaker) and a 90° version (mic pointed sideways or up). He says the 90° file isn't needed in a car and the direction the mic points hardly matters. His reason: at low frequencies a car cabin is a "pressure field," where the sound pressure is about the same everywhere in the small space. At high frequencies it is "random-incidence" (diffuse), meaning sound arrives from all directions at once after bouncing around. So an omni mic (one that picks up from all directions) pointed up or forward is fine; 90° files are for "free-field" work, where sound comes from one direction in open space. Most such mics use similar 1/4-inch (about 6 mm) electret capsules, and the loudest sound they can handle (max SPL) depends on their bias voltage: 5 V from USB gives a bit over 100 dB.

3.3 #9 · A phone's built-in mic is usable at low and mid frequencies.

Around 300 Hz it is close enough to omnidirectional; a real measurement mic matters mainly at high frequencies. [RARE] Sources: DIYMA 2021-11-02

In plain English: He says even a phone's built-in microphone is usable for low and middle frequencies. Around 300 Hz it picks up from all directions closely enough to work. A real measurement mic matters mainly for the high frequencies.

3.3 #10 · Your playback device can corrupt the test signal; use an external sound card and the right calibration.

Many laptops apply speaker protection in the OS that can't be disabled, Bluetooth may tap the signal after it, and outputs can clip near full volume. REW's soundcard calibration covers input and output together, so clear it when the pink noise comes from the CD or a phone. [RARE] Sources: DIYMA 2021-12-22 · DIYMA 2018-05-17 · Testgear site

In plain English: The test signal can be spoiled before it ever reaches the speakers. Many laptops apply "speaker protection" processing in the operating system that can't be turned off, Bluetooth may take its signal after that processing, and outputs can clip (distort) near full volume. So he recommends an external sound card (a separate audio interface). REW's soundcard calibration corrects the computer's input and output together, so if the pink noise is coming from the CD or a phone instead of the computer, clear that calibration, since the computer's output isn't part of the chain then.

3.3 #11 · "Nothing above 10 kHz" is usually the source.

A 128 kbps rip, Bluetooth SBC (cuts around 12–15 kHz) or MP3 drops the top octave, so don't EQ or judge highs from them. Lossy codecs shouldn't touch bass, so missing bass over Bluetooth means another fault. Sources: DIYMA 2021-12-05 · Testgear site

In plain English: If your measurement shows nothing above 10 kHz, he says the cause is usually the source, not the speakers. A 128 kbps rip, Bluetooth's basic SBC codec (which cuts off around 12–15 kHz) or an MP3 can drop the top octave of treble. So don't EQ or judge the highs using those sources. These "lossy" formats save space by discarding sound, but he says they shouldn't affect bass, so if bass goes missing over Bluetooth, look for a different fault.

3.3 #12 · Gating and frequency-dependent windows are useless in a car.

Reflections arrive too early to be heard as echoes (no Haas effect, no reverberant field, no late reflections), so they act as frequency-response errors you hear. A window short enough to exclude them loses all low-frequency resolution; gated data show nothing useful below about 1 kHz. [RARE] Sources: DIYMA 2009 · DIYMA 2014 · DIYMA 2017-10-28

In plain English: "Gating" or "windowing" means measuring only the first few thousandths of a second after the sound leaves the speaker, to cut out reflections. It is common in home speaker design. He says it is useless in a car. In a big room, a reflection can arrive late enough to be heard as a separate echo or as room sound; the Haas (precedence) effect and the reverberant field are the names for how the ear handles those later arrivals. In a car, every reflection arrives so soon after the direct sound that none of that happens; instead, reflections simply change the tone, as frequency-response errors you hear. A window short enough to exclude them is too short to capture slow bass waves, so it loses all low-frequency detail; he says gated data show nothing useful below about 1 kHz. Many people believe gated measurements, a standard home-speaker technique, also work in cars, which is why the point is tagged as unusual.

3.3 #13 · Scopes and 0 dB sine waves are of little use for setting DSP-system levels.

Setting mid and tweeter amp levels with a sine and scope is "a colossal waste of time" and can kill tweeters; doing it before the acoustic tune is riskier than setting gains by ear. Use them only for clipping points (gain detail in section 4). [RARE] Sources: Audiofrog blog Mar 2021 (scopes and sine waves) · Audiofrog blog Dec 2020 (pink noise)

In plain English: Many installers believe the proper way to set amplifier gains is with an oscilloscope (a tool that draws the waveform) and a 0 dB sine wave (a full-level test tone). He says this is of little use in a DSP system. Setting the midrange and tweeter amplifiers this way is "a colossal waste of time" and can kill (destroy) tweeters. Doing it before the acoustic tune is, in his view, riskier than simply setting gains by ear. He says to use scopes and sine waves only to find clipping points; section 4 has the gain details.

3.3 #14 · Phase and timing measurements need a two-input interface (loopback) or an acoustic reference.

Put the loopback on one input and the mic on the other; USB-only mics can't do it. Without loopback, drive the nearest tweeter as REW's acoustic timing reference; Audiofrog's PAA-1 preamp lets a UMI-1 capsule run on a 2-channel interface. Sources: DIYMA 2009 · DIYMA 2017-06-28 · DIYMA 2021-10-26

In plain English: Measuring timing and phase needs a reference for when the signal was sent. One way is "loopback": a two-input sound interface where one input receives the test signal straight from the output by cable, and the other input gets the mic. USB-only mics can't do this, because they have just one input. Without loopback, REW can use an "acoustic timing reference" instead: play a signal through the nearest tweeter first and time everything from it. Audiofrog's PAA-1 preamp lets the UMI-1's mic capsule plug into a 2-channel interface, so you can use loopback.

3.3 #15 · To measure a radio's outputs, use a balanced USB interface plus an attenuator, never a headset adapter on both channels.

Tying both channels' negatives to computer ground can trip the radio's protection. Put test leads on Hot and Cold with ground unconnected and divide the voltage to the interface's limit (his example takes ~1 V); a line-output converter ahead of an adapter also works. [RARE] Sources: DIYMA 2021-11-01 · DIYMA 2021-11-01 (attenuator)

In plain English: To measure what a radio actually sends out, he says to use a balanced USB interface plus an attenuator (a simple voltage divider that turns the signal down), never a headset adapter across both channels. The danger: a headset adapter ties both channels' negative wires to the computer's ground, which can trip the radio's protection circuit. Instead, put the test leads on each output's Hot (+) and Cold (−) wires, leave ground unconnected, and divide the voltage down to what the interface can accept; his example interface takes about 1 V. A line-output converter (a device that turns speaker-level signals into low-level ones) placed ahead of an adapter also works.

3.3 #16 · Near-field measurements can't show cone breakup or door-panel contributions.

Near field turns into far field at about 3× the driver's largest dimension; you must be that far away for cone and panel to sum. So a near-field anomaly (say at 2 kHz) is unlikely to be breakup, and sibilance or bad imaging is more likely EQ or reflections. [RARE] Sources: DIYMA 2022-03-16 · DIYMA 2017-07-30

In plain English: A near-field measurement puts the mic very close to the speaker, about an inch away. He says it can't show cone breakup (when parts of the cone flex unevenly at high frequencies) or what the door panel adds. The near field turns into the far field at about 3 times the driver's largest dimension, and you must be at least that far away for the sound from the cone and the panel to add together. (By that rule, for a 6.5-inch woofer that works out to roughly 20 inches, about half a meter.) So an odd bump in a near-field measurement, say at 2 kHz, is unlikely to be breakup. If you hear sibilance (harsh "s" sounds) or poor imaging, the cause is more likely EQ or reflections.

3.3 #17 · Sanity-check odd results before blaming the speaker.

A flat ~6 dB jump above 250 Hz between measurements may be the car cutting radio volume because it thinks a door is open; two identical tweeters measuring differently calls for a DC-resistance check first; claimed DSP "transient" differences can be settled with a null test (align, invert, sum). [RARE] Sources: DIYMA 2022-01-30 · DIYMA 2021-11-21 · DIYMA 2021-01-04

In plain English: Before blaming a speaker for a strange measurement, he checks the simple explanations first. His examples: a flat jump of about 6 dB above 250 Hz between two measurements may be the car itself turning the radio down because it thinks a door is open. If two identical tweeters measure differently, first check their DC resistance with a meter. And when people claim one DSP handles "transients" (sudden sounds) differently from another, a null test settles it: line the two recordings up in time, flip one upside down (invert it), and add them. If they cancel, they are the same.

3.3 #18 · "The RTA doesn't lie," but you can misread it.

A calibrated mic is correct; the screen misleads through single-mic combing, distortion it can't show, and stereo noise with no center. Cardboard near the mic only "invalidates" a measurement if you EQ its comb without understanding it, so always ask whether a measurement makes sense. Sources: Mobile Electronics 2017 (tag page) · DIYMA 2017-10-28 · DIYMA 2017-06-28

In plain English: His view is that "the RTA doesn't lie": a calibrated mic tells the truth about the sound at that spot. The danger is misreading the screen. Three ways it misleads: comb-filter peaks and dips that only exist at that one mic position; distortion, which the RTA can't show; and stereo noise, which has no center. A piece of cardboard near the mic only "invalidates" a measurement if you EQ away the comb it creates without understanding it. So his habit is to always ask whether a measurement makes sense.

3.3 #19 · The UMI-1 kit and binaural heads have different jobs.

UMI-1: serialized omni mic with its own calibration file on a headrest stand, USB-Duo interface usable with REW, Open Sound Meter or SMAART, ambient noise no problem. The BHM binaural mics are for recording and comparing cars, not EQ; tuning with "ears" gains nothing and needs a new target (3.5 #9). Company copy partly. Sources: UMI-1v2 product page · Audiofrog binaural page · DIYMA 2021-11-23

In plain English: This contrasts two Audiofrog products. (Andy co-founded Audiofrog, and the note "Company copy partly" means some of this comes from the company's product pages rather than his own words.) The UMI-1 is a measurement kit: an omni mic with its own serial number and calibration file, a stand that mounts on the headrest, and a USB-Duo interface that works with REW, Open Sound Meter or SMAART software; background noise doesn't spoil its measurements. The BHM binaural mics are tiny mics that sit in the ears of a listener or dummy head to record what a person would hear; they are for recording and comparing cars, not for EQ. He says tuning with "ears" mics gains nothing and would need a different target curve (see 3.5 #9).

The full series

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