Car Audio Explained, Part 12: Power, Clipping, Gain Setting, Amplifier Specs and Noise (Andy Wehmeyer's Lessons)

3 days ago

37 min read

This is Part 12 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.

4.7 How much power you really need, and what clipping does

What this part is about: How many watts each speaker in a car really uses, why only the subwoofer amplifier ever works near full power, and how "clipping" (an overdriven signal) is what actually burns speakers.

4.7 #1 · Power demand halves every octave up, so only the sub amp ever reaches rated power.

Pink noise (equal energy per octave) mimics music: calibrate 20 W at 20 Hz and a tweeter crossed at 2.5 kHz gets ~0.16 W (~0.02 W at 20 kHz), and a midbass crossed at 80 Hz needs at most a quarter of the sub's power. A tweeter needs "a watt or two at the most". [RARE] Sources: Audiofrog blog Dec 2020 · DIYMA 2020-10-14 · DIYMA 2020-12-23

In plain English: Sound is measured in hertz (Hz), vibrations per second: deep bass is around 20–80 Hz, treble is thousands of Hz. An "octave" is a doubling of frequency, such as 20 Hz to 40 Hz. Pink noise is a test hiss with equal energy in every octave, and he uses it as a stand-in for real music. In pink noise, the power at any one frequency halves each time the frequency doubles. So if an amp makes 20 W at 20 Hz, it makes 10 W at 40 Hz, 5 W at 80 Hz, and so on, down to about 0.16 W at 2.5 kHz and about 0.02 W at 20 kHz. A tweeter (the small treble speaker) that starts playing at 2.5 kHz therefore gets a tiny fraction of a watt. A midbass speaker that starts at 80 Hz (two octaves above 20 Hz) needs at most a quarter of what the sub gets. Many people believe tweeters need big amplifiers. His point is that only the sub amp ever reaches its rated power, and a tweeter needs "a watt or two at the most".

4.7 #2 · Measured: real music averages about 1/32 of its peak power, most of it bass.

An EDM track peaking near 0 dBFS (call it 400 W) averaged ~−16 dB (he rounds to −15 dB = 1/32 of the power, ~12.5 W), with peak energy near 50 Hz and 1 kHz about 20 dB down. Sources: DIYMA 2021-11-29

In plain English: In a digital recording, "0 dBFS" is the loudest level the format can hold. A decibel (dB) is a ratio scale for loudness or power: every 10 dB down is one tenth of the power. He measured an electronic dance music (EDM) track whose loudest moments hit about 0 dBFS. If those peaks drove an amp to 400 W, the track's average was about 16 dB lower, which he puts at about 12.5 W, roughly 1/32 of the peak. Most of that energy sat in the bass, near 50 Hz. At 1 kHz (the middle of the audio range) the level was about 20 dB lower, which is one hundredth of the power. The question he was answering was about an amp's average current draw. His point is that even loud music uses only a small slice of the amp's peak power on average, and most of what it does use is bass.

4.7 #3 · Rule of thumb: 50–100 W for mids and highs, at least ~800 W on the sub.

Cars need 6–12 dB more bass to sound natural (~9 dB takes ~8× the power), while extra watts elsewhere buy little: 80→125 W is 1.9 dB, and a 90 dB/1 W speaker turns 99.4% of it into heat. A speaker 3 dB more sensitive equals doubling the amp. Sources: DIYMA 2013-04-25 · DIYMA 2017-07-05 · DIYMA 2022-01-26

In plain English: His rough guide is 50–100 watts per channel for the midrange and treble speakers, and at least about 800 W for the subwoofer. The reason is that a car needs 6–12 dB more bass than the rest of the range to sound natural. Every 3 dB more takes double the power, so about 9 dB more takes about 8 times the power. Extra watts on the other speakers buy little: going from 80 W to 125 W is only 1.9 dB louder, a small change. A speaker's "sensitivity" is how loud it plays from 1 watt (here 90 dB). He notes that such a speaker turns 99.4% of the power into heat, not sound. So a speaker that is 3 dB more sensitive gives the same gain as doubling the amplifier's power.

4.7 #4 · You can't burn a coil with too little power; the extra high-frequency power of a clipped wave does it.

A clipped wave approaches a square, whose average power equals its peak (a sine's is half), so a 100 W amp clipped hard makes ~200 W, all the extra at high frequencies; clipping is not DC. Subs survive it (coil inductance and class-D output filters reject the HF); tweeters behind a passive crossover take it. [RARE] Sources: OFCAP 19 @20:07 · DIYMA 2021-07-01 · DIYMA 2021-07-08

In plain English: Many people believe a small, "underpowered" amp blows speakers. He says too little power can't burn a speaker's voice coil (the wire coil that drives the cone). What does damage is clipping. Clipping happens when an amp is asked for more than it can give, so the smooth tops of the sound wave get cut flat. A pure tone is a smooth "sine" wave; a hard-clipped one looks almost like a flat-topped "square" wave. A square wave's average power equals its peak, while a sine's average is half its peak. So a 100 W amp clipped hard puts out about 200 W, and all the extra is high-frequency content. He also says clipping is not DC (a steady one-direction current), another common belief. Subwoofers survive this because their coils naturally block high frequencies, and class-D amps have output filters that remove some of it. Tweeters fed through a passive crossover (a filter box of coils and capacitors at the speakers) receive that extra treble energy and burn.

4.7 #5 · Too much amp kills tweeters even without clipping; "overpowering for safety" works only if nothing can clip.

A 200 W amp at full output feeds a "20 W" speaker's tweeter ~2 W, ten times its share. Gains set on a −10 dB track license a 300 W amp to try for 3,000 W, and a 250 W/ch amp bridged onto a 100 W speaker melts the coil. [RARE] Sources: DIYMA 2020-10-14 · OFCAP 19 @27:00 · OFCAP 19 @36:04

In plain English: Many people believe you should buy an amp bigger than the speaker's rating "for safety". He says that only works if the amp can never clip. Using his pink-noise math, the tweeter in a speaker rated "20 W" handles about 0.2 W. A 200 W amp at full output gives that tweeter about 2 W, ten times its share, with no clipping at all. The "gain" is the amp's input-level knob. If you set it using a test track recorded 10 dB below maximum, a full-level song asks for 10 times the power (10 dB is 10× power). A 300 W amp is then trying to make 3,000 W, so it clips. In the same talk he says a 300 W amp on a 100 W speaker is safe only if the gain is set with a full-level (0 dB) test track. His last example: a 250 W per channel amp "bridged" (two channels combined into one bigger one) onto a speaker rated 100 W melts the voice coil.

4.7 #6 · Head-unit or DSP clipping, not amplifier clipping, usually kills mids and tweeters.

About 70% of GB25 warranty claims are DSP input clipping (25% are broken terminal screws), and returned "crackling" units measure perfect. Clipping the A/D shows up in the mids and tweeters, not the sub, and "sounds like broken speakers". [RARE] Sources: Audiofrog blog Mar 2021 · DIYMA 2022-03-13 · Absolute Electronix 2023 @33:35

In plain English: The head unit is the radio in the dash. A DSP (digital signal processor) is a box that splits, delays and equalizes the signal before the amps. Its first stage, the A/D (analog-to-digital) converter, turns the incoming voltage into numbers, and if the voltage is too high it clips. He says this is usually what damages midranges and tweeters, not the amplifier clipping. His evidence: about 70% of warranty claims on Audiofrog's GB25 midrange are really DSP input clipping, and 25% are broken terminal screws. Speakers sent back for "crackling" measure perfect. This kind of clipping shows up in the mids and tweeters, not the sub, and "sounds like broken speakers". So a crackle may be a setup error, not a bad speaker.

4.7 #7 · Find where distortion lives by turning things down, starting at the source.

If it stays with the head unit turned down, it isn't after the head unit; if it stays with the phone turned down, it's the track or codec (streaming mixes are often clipped, and phone "volume leveler" features add more). Ask what changed (new phone, new service?); no bass over Bluetooth means a fault, since codecs strip only highs. [RARE] Sources: DIYMA 2021-10-13 · DIYMA 2022-01-15 · DIYMA 2021-07-08

In plain English: This is a way to track down where a distortion comes from by turning down one part of the chain at a time, starting at the source. If the distortion is still there with the head unit turned down, it isn't coming from anything after the head unit. If it is still there with the phone turned down, it is in the track itself or the codec. A codec is the compression method used to stream or send audio, for example over Bluetooth. He notes that streaming mixes are often already clipped, and a phone's "volume leveler" feature adds more. He also says to ask what changed, such as a new phone or a new streaming service. And if bass disappears over Bluetooth, something is faulty, because codecs strip only the highs, not the bass.

4.7 #8 · A power rating is a limit, not a requirement, and "RMS power" is a misnomer.

The right term is continuous average power, tested with pink noise at rated watts (full power only at 20 Hz, halving per octave), so a mid or tweeter rating depends on the test's high-pass. Compressed recordings raise the average power, the real danger; "3× RMS for headroom" is nonsense, and warranties cover defects, not abuse. [RARE] Sources: OFCAP 19 @39:01 · DIYMA 2013-08-01 · DIYMA 2021-09-20

In plain English: A speaker's power rating tells you how much it can take, like a speed limit, not how much it needs. "RMS power" is the term most people use. He says it is a misnomer and the right term is "continuous average power". The rating test plays pink noise at the rated watts, so full power occurs only at 20 Hz and halves every octave up. A midrange or tweeter is tested through a high-pass filter (one that blocks lower notes), so its rating depends on where that filter was set. He says the real danger is heavily compressed (always-loud) recordings, which raise the average power. Many people believe you should buy an amp three times the speaker's RMS rating "for headroom"; he calls that nonsense. He adds that warranties cover defects, not abuse.

4.7 #9 · The volume control is the protection, and a customer's bass boost can undo a correct install.

A competent user trades loudness against peak distortion ("carefully monitored", in pro-audio terms) instead of setting levels on −10 dB tones. His case: a 1970s-disco fan with no sub-level control boosted bass ~12 dB at the head unit, clipped, burned tweeters and wrecked midbass; the fix was a higher, pro-sound-style crossover. [RARE] Sources: DIYMA 2021-07-10 · CMA Connected Oct 2021 @43:34

In plain English: His point is that the person at the volume knob is the real protection. A competent listener turns up until the loud peaks start to distort, then backs off. Pro-audio people call that "carefully monitored" use. The alternative he rejects is setting levels on test tones recorded 10 dB below maximum. His example: a dealer installed a system "correctly", yet the customer kept blowing speakers. The customer played 1970s disco loud. Those recordings have little bass, and the car had no separate sub-level control. So he boosted bass about 12 dB at the head unit. That clipped the signal, burned the tweeters and wrecked the midbass drivers. The fix was to raise the crossover point and set the system up like a pro sound system. That meant accepting that the bass might no longer seem to come from the front, which this customer didn't care about.

4.7 #10 · Amp power isn't what makes a car "dynamic".

His 2010 Volvo S60 ran everything but the subs on 20 W chip amps; a 2013 24-channel car used ~30 W × 23 plus 500 W on subs; judges called his 18 W car "so dynamic". [RARE] Sources: DIYMA 2010-02-16 · DIYMA 2013-04-27

In plain English: "Dynamic" describes a system whose loud moments jump out with punch. Many people assume that takes big amplifiers. His own cars say otherwise. His 2010 Volvo S60 ran every speaker except the subs on small 20 W "chip" amps. A 2013 car with 24 channels used about 30 W on each of 23 channels, plus 500 W on the subs. And judges described his 18 W car as "so dynamic".

4.7 #11 · Efficiency vs power.

2021: efficiency matters more (at ~88 dB/1 W/1 m only ~1% becomes sound; 16,000 W systems need heavy coils that cut efficiency further). [Conflict: 2013 "Efficiency is overrated. Power is cheap." — said about using separate amp channels rather than resistors to level-match drivers.] Sources: OFCAP 19 @33:07 · DIYMA 2013-11-28

In plain English: Efficiency (sensitivity) is how much sound a speaker makes from each watt. "88 dB/1 W/1 m" means it plays 88 dB from one watt, measured one metre away. In 2021 he said efficiency matters more than power. At that sensitivity only about 1% of the power becomes sound and the rest becomes heat. Huge 16,000 W systems also need heavy voice coils, which lower efficiency further. The tag flags an apparent conflict. In 2013 he wrote "Efficiency is overrated. Power is cheap." That line was about something narrower: giving each speaker its own amp channel to balance levels, instead of wasting power through resistors. Both statements are reported here as he made them.

4.8 Gain structure: DSP input, digital levels, amplifier gains

What this part is about: "Gain structure" means setting the level controls along the signal chain (radio, then DSP, then amplifiers) so the system plays loud without clipping and without audible hiss. This part covers the DSP's input setting, digital levels and amp gain knobs, his full step-by-step method, and how his advice changed over the years.

4.8 #1 · The DSP input-sensitivity control has "ONE function": keep the A/D converter from clipping at maximum volume.

Set it at or above the source's real maximum (a 4 V radio → 4–5 V, never lower); a 4 V radio into a 1 V setting allows ~12 dB of clipping nothing downstream can remove. Set it for full volume, not your usual level, above all in customers' cars. [RARE] Sources: Audiofrog blog Mar 2021 · DIYMA 2021-08-04 · DIYMA 2021-08-03

In plain English: A DSP's input-sensitivity setting tells it how strong a voltage to expect from the radio, so it can fit that signal into its A/D converter (the part that turns the voltage into numbers). He says this control has only one job: stop that converter from clipping when the radio is at full volume. So set it at or above the radio's real maximum output. If the radio makes 4 volts, set 4–5 V, never lower. If you set 1 V instead, the top of the signal (about 12 dB of it, since 4 V is four times 1 V) gets clipped. He explains that this distortion gets digitized permanently, so nothing later in the chain can remove it. Set it for full volume, not the level you usually listen at, above all in a customer's car.

4.8 #2 · Find the source's true maximum by measuring its response first.

Aftermarket radio with tone controls flat: a 1 kHz tone is fine. Factory radio or amp: measure the response (sound card and REW), find the loudest frequency and read the voltage there with a 0 dB sine and a true-RMS meter; a USB interface with balanced inputs plus an attenuator works. Sources: DIYMA 2021-08-04 · DIYMA 2021-08-05 · DIYMA 2013-09-17

In plain English: To set the DSP input correctly (point #1), you need to know the most voltage your source can put out. With an aftermarket radio and its bass and treble controls set flat, a 1 kHz test tone gives a fair reading. A factory radio or factory amp is different: it applies its own equalization, so it may be loudest at some other frequency. There, first measure its frequency response (how loud it plays at each frequency) using a computer sound card and REW (Room EQ Wizard, a free measurement program). Find the loudest frequency. Then play a full-level (0 dB) test tone at that frequency and read the voltage with a true-RMS multimeter, a type that reads AC voltage accurately. He says a USB audio interface with balanced inputs, plus an attenuator (a part that safely lowers the voltage), also works as the measuring tool.

4.8 #3 · In a digital system err toward never clipping, and leave headroom for bass EQ.

A little analog clipping is inaudible; a little digital clipping "sounds terrible". Two inputs mixed at 50% sum to 0 dBFS, and with every level at 100% the bass boosts clip the DSP; start DSP levels around −6 to −12 dB. 16-bit/44.1–48 kHz is enough if levels are right. [RARE] Sources: DIYMA 2021-08-04 · Audiofrog blog Dec 2020

In plain English: Inside a DSP the signal is numbers, and there is a hard ceiling called 0 dBFS ("full scale"). He says a little clipping in an analog circuit is inaudible, but a little digital clipping "sounds terrible". So in digital gear, lean toward never clipping. Levels add up quickly: two inputs each mixed in at 50% already add up to the full-scale maximum. If every level is at 100% and you then boost bass with the equalizer (EQ), the DSP clips. His advice is to start DSP levels around 6 to 12 dB below maximum. He adds that 16-bit audio at 44.1–48 kHz (CD-type quality) is enough, as long as levels are set right.

4.8 #4 · His step-by-step gain procedure (2022 version), and how it evolved.

The aim is to balance hiss, output and clipping, without scopes and meters on every channel. [RARE] History: 2008 "go buy a scope" (Vmax = √(P × R), e.g. ~26.5 V for 175 W into 4 Ω, on a −10 dB sine). 2009–2015: deliberate overlap (point 5). 2014: no scope in 15 years; by ear. 2016: radio at ¾, raise the tweeter amp until distortion, back off, then lower until hiss vanishes at zero volume. 2018: amp gains per the DSP manual, tune, raise all gains until a little tweeter hiss, back off. Title-attributed DIYMA guide: DSP levels start at −12 dB; sub ~+6 dB if one input feeds it. 2021: with factory sources use the DSP's high-level input and skip the "sine at ¾ volume" routine. [Changed view: scope and −10 dB tones (2008–2015) → sine on the sub only, the rest by pink noise and RTA (2018–2022).] Sources: DIYMA 2022-03-03 · DIYMA 2008-10-27 · DIYMA 2018-05-04

Set the DSP's analog input sensitivity at or above the source's maximum (4 V radio → 4 V or higher).
Sub only: 0 dB sine, DSP sub level and input at 0 dB, raise the sub amp gain until it just clips; then put the sine and scope away. Tweeters never get a sine (disconnect them for any sine-wave step).
Turn every DSP input and output level down 6 dB.
Crossovers and delays set by the numbers; with pink noise, bring the other amps' gains toward the target curve using level only, tweeter amp lowest.
EQ to the target and listen.
Raise all DSP outputs equally until one reaches 0 dB; check for clipping or noise.
Raise the DSP inputs toward 0 dB.
Raise every amp gain by the same amount while watching the RTA; at the first noise or distortion you've found the maximum.

In plain English: This is his 2022 recipe for setting levels in a system with a DSP and separate amplifiers. The goal is a balance between hiss, loudness and clipping, without putting an oscilloscope (a screen that draws the signal's wave shape) and meters on every channel.

Step 1. Set the DSP's input sensitivity at or above the most the radio can put out, so the DSP's converter can never clip (see #1 and #2).

Step 2. This step is for the subwoofer only. Play a full-level (0 dB) sine test tone with the DSP's sub level and input level at maximum (0 dB). Turn up the sub amp's gain until it just starts to clip, then stop. After that, put the test tones and scope away. Tweeters must never be fed a sine tone, so disconnect them during any step that uses one; he says setting tweeter and midrange levels with sine tones puts the speakers at unnecessary risk.

Step 3. Turn every DSP input level and output level down by 6 dB. In line with #3, this leaves headroom inside the DSP for EQ boosts.

Step 4. Set the crossovers (which frequencies go to which speaker) and the delays "by the numbers", meaning to planned values (for the delays, measured distances), not by ear. Then play pink noise and set the other amps' gains so the measured balance gets close to the target curve (the tonal balance you're aiming for), using level alone, with the tweeter amp's gain set lowest.

Step 5. Use EQ to match the target curve, then listen.

Step 6. Raise all the DSP output levels by the same amount until one of them reaches 0 dB (its maximum). Check for clipping or noise. He says to stop here if it's already loud enough.

Step 7. If not, raise the DSP input levels toward 0 dB, backing off if noise or clipping appears.

Step 8. If it is still not loud enough, turn every amp's gain up by the same amount while watching an RTA (real-time analyzer, a live display of level at each frequency), so the tonal balance stays the same. When you first hear noise or distortion, you've found the system's maximum. He says going louder than that means accepting some noise or distortion.

The "History" lines show how his method changed. In 2008 he told people to buy a scope and set gains on a −10 dB sine, working out the amp's full-power voltage with a formula (about 26.5 V for 175 W into 4 Ω). From 2009 to 2015 he advised deliberate overlap (see #5). In 2014 he said he hadn't used a scope in 15 years and set gains by ear. In 2016: radio at three-quarter volume, raise the tweeter amp until it distorts, back off, then lower it until hiss disappears with the volume at zero. In 2018: set amp gains per the DSP's manual, tune, then raise all gains until a little tweeter hiss appears and back off. A DIYMA forum guide whose thread title credits him (the research couldn't confirm who actually posted it) starts DSP levels at −12 dB, with the sub about 6 dB higher if one input feeds it. In 2021 he said that with factory sources you should use the DSP's high-level (speaker-level) input and skip the "sine at ¾ volume" routine. The [Changed view] tag sums it up: scope and −10 dB tones (2008–2015) gave way to a sine on the sub only, with everything else set by pink noise and an RTA (2018–2022).

4.8 #5 · Gain overlap (deliberate clipping headroom) was once his advice; he dropped it for DSP systems.

2009–2015: set amps to clip on a −10 dB tone ("YOU WANT SOME CLIPPING!"), ~9–10 dB of overlap (2 V head unit → ~600 mV), since transient clipping is hard to hear; JBL's GTO "EZ" −10 dB disc built this in. Constant: never clip the A/D or head unit. [Changed view: 2021, overlap "was never appropriate" at the DSP input, and −10 dB tones on HF amps license tweeter-killing clipping.] Sources: DIYMA 2009-08-13 · DIYMA 2015-10-15 · Audiofrog blog Mar 2021

In plain English: "Overlap" means setting the amp so it reaches full power, and clips, before the signal feeding it reaches its own maximum. From 2009 to 2015 he recommended it on purpose: set the amps to clip on a test tone recorded 10 dB below maximum, which gives about 9–10 dB of overlap. For example, with a radio that puts out 2 V, set the amp's sensitivity to about 600 mV (0.6 volts). His reasoning then was that brief clipping on short peaks is hard to hear, and quieter recordings still play loud. JBL's GTO "EZ" amps came with a −10 dB test disc that built this in. One rule never changed: never clip the A/D converter or the head unit. In 2021 he changed his view. He said overlap "was never appropriate" at the DSP input. He also said −10 dB tones on treble amps allow clipping that kills tweeters.

4.8 #6 · Starting point: amp gains around 2 V.

With MS-8 or any DSP he repeatedly says ~2 V (JBL MS amps: "Low" switch, pot fully counter-clockwise); an 8 V setting guarantees the amp never clips and wastes ~12 dB. Sources: DIYMA 2015-04-24 · DIYMA 2013-05-25

In plain English: An amp's gain knob is usually marked in volts: the input voltage that drives the amp to full power. With JBL's MS-8 (a specific JBL car processor) or any other DSP, he repeatedly suggests starting around 2 V. On JBL MS-series amps, that means the "Low" switch position with the knob turned fully counter-clockwise. Setting 8 V means the amp needs 8 volts in to reach full power. Fed a signal of around 2 V, it never gets there, so it never clips, but you throw away about 12 dB of its output (8 V is four times 2 V).

4.8 #7 · The amp gain knob is a level-matching control, and low gain isn't automatically low noise.

If noise arises after the pot, low gain worsens the signal-to-noise ratio; test with RCAs unplugged and gain at max (hiss means the amp's own preamp). A tweeter that plays 3–6 dB louder than the woofer (Audiofrog's do) lets that channel's gain, and its hiss, come down. [RARE] Sources: DIYMA 2010-09-11 · DIYMA 2016-06-06 · DIYMA 2017-10-01

In plain English: Many people believe turning an amp's gain down always means less noise. He says the gain knob is really for matching levels between channels, and a low setting isn't automatically quieter. If an amp's hiss is created after the gain control (the "pot"), turning the gain down lowers the music but not the hiss, so the signal-to-noise ratio gets worse. His test: unplug the RCA cables (the signal leads) and turn the gain all the way up. If you hear hiss, it comes from the amp's own preamp stage, and a lower setting will help. He adds that a tweeter that plays 3–6 dB louder than the woofer, as Audiofrog's tweeters do, lets you turn that channel's gain down, and its hiss with it.

4.8 #8 · Line drivers are "possibly the most unnecessary component in all of car audio", and fat RCAs are useless too.

They made sense with ~100 mV head units and single-ended amps; into a differential input, an 8 V line driver is just 6 dB of overlap. When one "fixed" noise, its ground break did it; amp input impedance makes heavy-gauge RCAs pointless. [RARE] Sources: DIYMA 2010-06-01 · DIYMA 2011-12-27 · DIYMA 2021-07-10

In plain English: A line driver is a small box that boosts the radio's signal voltage before it reaches the amps. He calls it "possibly the most unnecessary component in all of car audio". It made sense when radios put out only about 100 mV (a tenth of a volt) and amps had "single-ended" inputs, the older kind that use the cable's outer shield as the ground return. A "differential" input instead measures the difference between two wires and ignores noise common to both. Into a differential input, he says, an 8 V line driver just adds 6 dB of overlap (signal beyond what the amp needs; see #5). When a line driver seemed to cure noise, he says the cure came from its ground isolation, not its extra voltage. He also says thick, heavy-gauge RCA cables are pointless: amp inputs have a high input impedance (resistance to current), so they draw almost no current and wire thickness doesn't matter.

4.8 #9 · DSP amplifiers have no analog gain after the DAC, so EQ cuts cost output you can't win back.

The input is attenuated to fit the A/D (~1–4 V), costing SNR; the D/A can't exceed 0 dBFS, so after tuning cuts a "1,000 W" DSP amp may deliver ~10 W. For high-end systems use a separate DSP and amps with analog input-sensitivity controls; with a DSP amp, EQ dips up as well as cutting peaks. [RARE] Sources: Absolute Electronix 2023 @33:35 · DIYMA 2022-03-03

In plain English: A "DSP amplifier" is one box containing both a processor and an amplifier. The signal goes in, is shrunk to fit the A/D converter, whose ceiling is about 1–4 V, and that shrinking costs some signal-to-noise ratio (SNR, how far the music sits above the hiss). Then a DAC, or D/A (digital-to-analog converter), turns the numbers back into a voltage, and it can never go above 0 dBFS. In these boxes there is no adjustable analog gain after that point. So every EQ cut you make during tuning lowers the maximum output for good. In his example, after tuning cuts a "1,000 W" DSP amp may deliver about 10 W. For high-end systems he recommends a separate DSP plus amps that have analog input-sensitivity controls. If you do use a DSP amp, he says to EQ by boosting dips as well as cutting peaks.

4.8 #10 · "All-digital" signal chains waste amplifier power.

Nothing exceeds 0 dBFS and cars need 6–12 dB more bass, so everything else must be attenuated; multi-thousand-watt digital systems end up sounding like 300 W ones. An analog chain with balanced outputs and differential inputs is "just fine". [RARE] Sources: DIYMA 2017-07-05

In plain English: In an "all-digital" chain the signal stays as numbers until the very end, with no analog level controls along the way. Nothing can go above the digital ceiling (0 dBFS). Cars need 6–12 dB more bass than the rest of the range, so everything else has to be turned down to make room for that bass boost. The result, he says, is that systems with thousands of watts end up sounding like 300 W systems. Many people assume all-digital is automatically better. He says an analog chain using balanced outputs and differential inputs (wiring that cancels noise picked up along the cable) is "just fine".

4.8 #11 · Know your volts and decibels.

Scope readings are peak (4 V peak ≈ 2.83 V RMS); ~4.24 V RMS is the most a 12 V head unit makes without a switching supply. Use 20 log for voltage and 10 log for power (a 6 dB cut is 75% less power); splitting a preamp signal doesn't lower its level; non-true-RMS meters keep rising into clipping. Sources: DIYMA 2013-07-21 · DIYMA 2013-09-17 · DIYMA 2017-10-11

In plain English: These are the basic numbers he wants installers to get right. An oscilloscope shows a wave's peak voltage. A multimeter shows "RMS" voltage, a kind of effective average; for a sine wave, RMS is about 0.707 times the peak, so 4 V peak is about 2.83 V RMS. He says about 4.24 V RMS is the most a head unit running on 12 volts can make without a switching power supply (a circuit that raises the voltage). To turn voltage ratios into decibels you use "20 log"; for power ratios you use "10 log". So a 6 dB cut is half the voltage but 75% less power. Splitting one preamp signal to feed several amps doesn't lower its level. And a meter that isn't "true-RMS" keeps showing rising numbers even after the signal starts clipping; he says this makes setting gains with an ordinary multimeter imprecise.

4.9 Amplifier specs and design: what matters, what doesn't

What this part is about: Which numbers on an amplifier's spec sheet actually matter, which are marketing, and what really makes one amp sound different from another. It ends with details of his own company's amplifiers.

4.9 #1 · Damping factor is a marketing spec; above ~10 it's meaningless.

The voice coil's own resistance (~3.2–4 Ω) sits in series with the amp's output impedance and sets the current, so DF 1000 moves Qts from 0.707 to 0.708. Output impedance matters only when it nears the coil's DCR, where it tilts response; that is tube "warmth". [RARE] Sources: Audiofrog blog Dec 2020 · DIYMA 2017-07-12 · DIYMA 2021-05-08

In plain English: "Damping factor" (DF) is a spec that claims to show how tightly an amp controls a speaker's cone. Many people believe a huge number like 1000 means tighter bass. He calls it a marketing spec that is meaningless above about 10. Resistance is measured in ohms (Ω). The voice coil's own wire resistance, its DCR (about 3.2–4 Ω), sits in line with the amp's tiny internal resistance (its "output impedance"), and it is the coil's resistance that sets the current. So going to DF 1000 changes the speaker's Qts (a number describing how damped, or controlled, its resonance is) only from 0.707 to 0.708, a negligible change. The amp's output impedance matters only when it gets close to the coil's DCR. Then it tilts the frequency response, and he says that is the "warmth" people hear from tube amps.

4.9 #2 · Slew rate is already covered by the power-bandwidth (THD) spec; compare distortion at ~1 W.

A slew limit shows up as HF distortion in the THD figure. Tweeters live around a watt, and some amps distort more at very low power than at rated power. [RARE] Sources: DIYMA 2017-07-07 · DIYMA 2021-05-18

In plain English: "Slew rate" is how fast an amp's output voltage can change. Some buyers treat it as a separate quality measure. He says it is already covered by the THD (total harmonic distortion) spec measured across the full frequency range at rated power. If an amp can't change its voltage fast enough, that shows up as extra distortion at high frequencies in the THD number. The more useful thing to compare is distortion at about 1 watt. That's because tweeters live around a watt, and some amps distort more at very low power than at full rated power.

4.9 #3 · "Class D is only for bass" is backwards; class D's real car problem is FM interference.

In Harman's 1996 CES blind demo (Infinity Beta class D vs a Krell) only 3–4 trained listeners heard a difference, in long bass notes. Class AB and D differ only when audibly clipped, so judge implementations ("all marinara is terrible because you tried Ragu"). [RARE] [Conflict: ~800 demos/3 listeners (2022) vs ~1,500 demos/4 listeners (2023).] Sources: CMA 2023 @25:04 · CMA 2022 @29:01 · DIYMA 2022-03-26

In plain English: "Class AB" and "class D" are two ways of building an amplifier. Class D switches its output transistors on and off very fast, which makes it small and efficient. Many people believe class D is only good enough for subwoofers. He says that's backwards, and that class D's real problem in cars is interference with FM radio reception. His evidence is a 1996 blind demo by Harman (the company behind JBL and Infinity) at the CES trade show. It compared an Infinity Beta class-D amp with a Krell, a high-end amp. Only 3–4 trained listeners heard a difference, and only on long bass notes. He says class AB and class D sound different only when audibly clipping, so judge each amp on its design, not its class. Hence his line "all marinara is terrible because you tried Ragu". The [Conflict] tag notes that he gave different counts in two interviews: about 800 demos and 3 listeners (2022), versus about 1,500 demos and 4 listeners (2023).

4.9 #4 · Class-D treble depends on the load.

The passive output filter is designed for one impedance: with a ~20 kHz filter, 2 Ω loads roll off highs and 8 Ω or rising tweeter impedance peak. Audiofrog puts it at ~40 kHz, matched to its low-inductance tweeters (flat almost to 30 kHz); big HF differences between amps usually come from output filters. [RARE] Sources: Absolute Electronix 2023 @29:58 · CMA 2022 @35:07 · DIYMA 2021-05-18

In plain English: A class-D amp has a small filter on its output to remove the switching noise. That filter is designed to behave correctly with one particular load impedance (the speaker's resistance to the signal). If the filter is set around 20 kHz, a 2 Ω load makes the highest treble roll off (drop), while an 8 Ω load, or a tweeter whose impedance rises at high frequencies, makes the treble peak (rise). He says Audiofrog places this filter at about 40 kHz and matches it to its own low-inductance tweeters, which stay flat almost to 30 kHz. His point is that when two amps sound very different in the treble, the output filters are usually the reason.

4.9 #5 · You can't audition an amp by itself.

What people hear is output impedance interacting with the speaker, or clipping; "one unclipped watt is one unclipped watt". Buy expensive amps for features, longevity or clipping behavior; GaN brings smaller size and lower HF distortion, but silicon is fine. [RARE] Sources: CMA 2023 @19:05 · DIYMA 2013-05-19 · DIYMA 2022-03-26

In plain English: When people compare amps by ear, he says they are really hearing the whole system: the amp's output impedance interacting with that particular speaker, or the amp clipping. Apart from that, "one unclipped watt is one unclipped watt". So the reasons to pay more for an amp are its features, how long it lasts, or how it behaves when it clips. GaN (gallium nitride, a newer type of output transistor) brings smaller size and lower distortion at high frequencies, but he says silicon is fine.

4.9 #6 · Low noise and low-level distortion matter more than headline watts, and noise is board layout.

Tiny class-D amps hiss or whine because the supply transformer's field sits near preamp parts; big amps "solve" it with distance. Audiofrog uses a 4-layer board with a ground-shield layer, audio parts where the transformer's field is weakest, and a chassis that shields FM. [RARE] Sources: CMA 2023 @22:05 · Absolute Electronix 2023 @26:01 · CMA 2022 @44:00

In plain English: He says a quiet amp with low distortion at low volume matters more than a big wattage number. Car amps contain a power-supply transformer that raises the car's 12 volts, and it gives off a magnetic field. In tiny class-D amps that field sits right next to the delicate preamp (small-signal) parts, which causes hiss or whine. Big amps "solve" this simply by putting distance between them. His company, Audiofrog, instead uses a circuit board with four layers, one of them a ground layer that acts as a shield. It places the audio parts where the transformer's field is weakest, and uses a metal case that shields against FM radio interference. So in his view, noise comes down to how the circuit board is laid out.

4.9 #7 · Channel separation beyond ~60 dB is irrelevant, and separate left/right amps buy nothing.

60 dB is a millionth of the power, while acoustic crosstalk in a car is only 6–12 dB; for separation "a car wash is more likely" to help. A full-range bridged amp causes no image-damaging phase shift. [RARE] Sources: DIYMA 2022-03-26 · DIYMA 2021-11-05 · DIYMA 2021-09-15

In plain English: "Channel separation" measures how little of the left channel leaks into the right inside the amp. Some people buy separate amps for left and right to improve it. He says anything beyond about 60 dB is irrelevant: 60 dB of separation means the leak is a millionth of the power. Meanwhile, in a car the sound itself leaks. Sound from one side of the car reaches the other side, directly and by reflecting off the windows, only 6–12 dB quieter. That acoustic "crosstalk" dwarfs anything the amp does, so he jokes "a car wash is more likely" to help. He also says a full-range bridged amp (two channels combined to drive one speaker) causes no phase shift that would damage the stereo image (the sense of where instruments are placed).

4.9 #8 · Car amps are voltage sources: bridging quadruples power, and halving the load does not double it.

One channel at 10 V peak into 4 Ω is 25 W; bridged, the speaker sees 20 V, 100 W. Amps "know" only current; past their limit modern amps sag the supply rails rather than shut down, so "set gains under load" matters only deep into clipping. Sources: DIYMA 2012-05-12 · DIYMA 2022-02-10 · DIYMA 2021-11-14

In plain English: A "voltage source" holds its output voltage steady and lets the current change with the load. Power depends on voltage squared, so doubling the voltage gives four times the power. One channel swinging 10 V peak into a 4 Ω speaker makes 25 W. Bridged, the two channels push in opposite directions, so the speaker sees 20 V and gets 100 W. Many people believe that halving the speaker's impedance (say 4 Ω to 2 Ω) doubles an amp's power. He says it does not, and puts this down to real-world inefficiencies. He also says amps "know" only current. Pushed past their limit, modern amps let their internal supply voltage (the "rails") sag rather than shut down. So the common advice to "set gains under load" matters only when the amp is driven deep into clipping.

4.9 #9 · Sanity-check power claims with the fuse and the heatsink.

A 20 A fuse and a 2,000 W rating don't go together; some amps' thermal rollback kicks in within minutes, so "150 W" acts like 75 W. Designers now save aluminum by rolling back supply voltage when hot. [RARE] Sources: DIYMA 2011-12-16 · CMA 2022 @1:09:04 · DIYMA 2022-01-14

In plain English: You can check whether an amp's power claim is believable. An amp running on about 12 volts can draw only as much current as its fuse allows, and a 20 A (amp) fuse simply can't support a 2,000 W rating. The heatsink (the metal body that sheds heat) is another clue. Some amps have "thermal rollback": when they get hot, they cut their own power, sometimes within minutes. So an amp rated "150 W" may behave like a 75 W amp. He says designers now save on aluminum (smaller heatsinks) by lowering the supply voltage when the amp gets hot.

4.9 #10 · Built-in amp crossovers are "broccoli on the buffet": defeat them under a DSP.

Set the A150.4D filter to "None" and open the A600.1D fully (HP 5 Hz, LP 500 Hz). Its filters are 12 dB/oct because 24 dB pot filters need 8-gang pots that never track; with 12 dB amp filters, alternate polarity between bands and keep an octave between high- and low-pass. Sources: DIYMA 2013-10-31 · A-Series manual · DIYMA 2022-02-28

In plain English: Many amps have crossover filters built in. He calls them "broccoli on the buffet": cheap to include, and you can take them or leave them. When a DSP is doing the crossovers, switch the amp's filters off. On Audiofrog's A150.4D, set the filter to "None". On the A600.1D, open it fully: high-pass (HP) at 5 Hz and low-pass (LP) at 500 Hz, its widest settings, so the DSP does the filtering. He explains that these amps' filters are 12 dB per octave (a gentle slope). A steeper 24 dB filter adjusted by knobs would need "8-gang" pots, eight controls on one shaft, which never stay matched to each other. If you do use 12 dB amp filters, he says to reverse the speaker polarity (wiring +/−) on alternate bands, and to keep an octave between the high-pass and low-pass points.

4.9 #11 · Audiofrog A-series (product-specific): two small models, no DSP inside, build any system.

A150.4D (~8 × 4 × 1.25 in): 4×100 W at 4 Ω, 4×150 W at 2 Ω, 2×300 W bridged (dyno ~175 W × 4 at 2 Ω); A600.1D: 450 W at 4 Ω, 600 W at 2 Ω (dyno ~750 W); 60 A fuse each. An active 3-way front is one A150.4D on tweeters and mids plus one bridged for midbass; RLC-Sync makes all sub amps follow the lowest-set one. Sources: A150.4D page · A600.1D page · Absolute Electronix 2023 @23:06

In plain English: This point is about specific products from Audiofrog, Andy's company. The A-series has just two small amps with no DSP inside, meant to be combined into any system. The A150.4D is about 8 × 4 × 1.25 inches (roughly 20 × 10 × 3 cm). It is a 4-channel amp rated 100 W per channel at 4 Ω, 150 W per channel at 2 Ω, or 300 W × 2 bridged; on a dyno (a test bench) it made about 175 W × 4 at 2 Ω. The A600.1D is a 1-channel amp rated 450 W at 4 Ω and 600 W at 2 Ω, and measured about 750 W. Each uses a 60 A fuse. An "active 3-way" front (tweeter, midrange and midbass each on their own amp channel) takes one A150.4D for the tweeters and mids plus a second, bridged, for the midbass. RLC-Sync is a feature of its remote level control: with several sub amps, all of them follow the one set lowest.

4.10 Noise: hiss, whine, ground loops and pops

What this part is about: How to tell the different kinds of unwanted noise apart (steady hiss, engine-speed whine, turn-on pops), how to find which box is making them, and which fixes actually work.

4.10 #1 · Ground loops make whine; hiss is almost never a ground problem.

Hiss that stays with the RCAs unplugged is the amp; hiss that rises with head-unit volume is the head unit; noise that moves sides when you swap both RCAs and speaker leads is the amp. Tying RCA shields together fixes whine, not hiss. [RARE] Sources: DIYMA 2015-09-16 · DIYMA 2015-09-27 · DIYMA 2022-03-01

In plain English: Hiss is a steady "shhh" sound. Whine is a tone, often one that rises and falls with engine speed. A ground loop happens when two connected pieces of equipment see slightly different "ground" (zero-volt reference) voltages, and the difference leaks into the audio. Many people blame every noise on grounding. He says ground loops make whine, but hiss is almost never a ground problem. His checks: if the hiss stays when the RCA signal cables are unplugged from the amp, the amp is making it. If the hiss gets louder as you turn up the head unit, the head unit is making it. If you swap both the left/right RCAs and the left/right speaker leads at the amp and the noise moves to the other side, the amp is the cause. Tying the RCA cables' shields (outer conductors) together can fix whine, but not hiss.

4.10 #2 · Ground loops are designed into equipment, and differential inputs largely ended the ground-loop era.

The loop is a difference between chassis and audio ground inside the boxes, so chasing ground points is often wasted effort. Threads list re-grounding, 0-gauge cable, the "Big 3" and new RCAs with the noise unchanged: "that's not the problem anymore". [RARE] Sources: DIYMA 2013-03-09 · DIYMA 2021-09-24 · DIYMA 2021-11-03

In plain English: He says a ground loop is built into the equipment itself: it comes from a voltage difference between the metal case (chassis) ground and the audio ground inside the boxes. So moving ground wires to new bolts in the car is often wasted effort. A "differential" input measures the difference between two signal wires and ignores noise that appears on both, and he says such inputs largely ended the ground-loop era. He points to forum threads where people re-grounded everything, fitted 0-gauge (very thick) power cable, did the "Big 3" (a common upgrade of the car's three main charging cables) and bought new RCA cables, and the noise didn't change. His verdict on grounding points: "that's not the problem anymore".

4.10 #3 · Ground-loop triage.

2008: unplug the RCAs at the amp (if noise stays, a grounding block won't help); touch the amp's RCA shield to the radio case (if noise goes, make that wire permanent); otherwise fit transformer isolators. 2015: if an amp's input shields read 0 Ω to its power ground it is single-ended and built to make loops, so avoid buying it. 2021: on a Helix, try the 200 Ω input-ground setting. [RARE] Sources: DIYMA 2008-12-06 · DIYMA 2015-04-18 · DIYMA 2021-11-03

In plain English: "Triage" means sorting out the problem in order. These are his tips from three different years. 2008: first unplug the RCA signal cables at the amp. If the noise is still there, a grounding block (a terminal block for sharing ground wires) won't help. Next, touch the outer shield of the amp's RCA cable to the radio's metal case. If the noise goes away, make that connection permanent with a wire. If neither works, fit transformer isolators (small boxes that pass the audio through a transformer and break the ground connection). 2015: use an ohmmeter (a meter that measures resistance). If an amp's input shields read 0 Ω to its own power ground, it has "single-ended" inputs that are built to make loops, so he says avoid buying it. 2021: on a Helix (a brand of DSP), try its 200 Ω input-ground setting.

4.10 #4 · The best cure: a floating speaker-level output into a differential input — no ground connection, no loop.

Its higher voltage also rejects radiated noise; a high-to-low converter (LOC) that cuts voltage tenfold often creates the noise, and shops then sell a line driver to win it back. Optical's one real advantage, radiated-noise immunity, is matched "damn close". [RARE] Sources: DIYMA 2010-06-01 · DIYMA 2009-08-23 · DIYMA 2013-11-22

In plain English: A radio's speaker outputs are often "floating", meaning not tied to ground. He says the best cure for ground loops is to feed such a speaker-level output straight into an amp or DSP input that is "differential" (it reads the difference between two wires). With no ground connection, there is no loop. The speaker-level signal's higher voltage also helps it stand out against noise picked up through the air ("radiated noise"). A common alternative is a high-to-low converter, or LOC (line output converter), which cuts the voltage to RCA level. He says an LOC that cuts the voltage tenfold often creates the noise problem, and shops then sell a line driver to boost the voltage back up. Some people prefer optical (light-based) digital connections. He says their one real advantage, immunity to radiated noise, is matched "damn close" by this floating-output method.

4.10 #5 · Match cable to input: twisted pair for differential inputs, coax only for single-ended.

Ohmmeter test: RCA shield to the amp's power ground near 0 Ω means single-ended (keep any LOC); hundreds to thousands of ohms means differential (feed it directly). Either mismatch amplifies noise; which RCA rejects noise best depends on the input circuit, not the cable. [RARE] Sources: CMA 2022 @50:01 · DIYMA 2021-09-17 · DIYMA 2021-11-26

In plain English: There are two common signal-cable designs. "Twisted pair" is two wires twisted together; "coax" is one centre wire inside a surrounding shield, like a typical RCA cable. He says twisted pair suits differential inputs, and coax suits only single-ended inputs. To find out which input an amp has, use an ohmmeter between the RCA jack's outer shield and the amp's power-ground terminal. Near 0 Ω means single-ended, so keep any LOC (line output converter) you're using. Hundreds to thousands of ohms means differential, so you can feed it directly. Using the wrong cable for the input type, either way round, amplifies noise. And which RCA cable rejects noise best depends on the amp's input circuit, not on the cable brand.

4.10 #6 · Transformer ground-loop isolators "aren't a band-aid"; for alternator whine try a 4700 µF cap at the alternator first.

Put a polarized 4700 µF cap on the alternator output post, negative to chassis; if that fails, use a real transformer isolator such as the ~$20 PAC SNI-1. Sources: DIYMA 2013-03-19 · DIYMA 2008-08-19

In plain English: Isolators are sometimes dismissed as a band-aid, a quick patch over a bad install. He says transformer isolators "aren't a band-aid", meaning they are a legitimate fix. For alternator whine (a whine that follows engine speed, from the car's charging alternator), he suggests trying a capacitor first. A capacitor stores electrical charge and can smooth ripples in a supply. Use a polarized one (it has a marked + and − side) of 4700 µF (microfarads, its size). Connect it to the alternator's output post, with the negative side to the chassis. If that doesn't work, use a real transformer isolator, such as the PAC SNI-1, which costs about $20.

4.10 #7 · Amps with built-in DSP are inherently noisier.

Their preamp must attenuate any input to fit the A/D (~1–1.4 V RMS max) and stay very quiet doing it; know whether the gain control works before or after the A/D. If hiss falls with gain turned down and nothing connected, the amp is the source; on a tweeter-only amp, lower its gain and boost highs in the DSP. [RARE] Sources: DIYMA 2021-10-14

In plain English: An amp with a DSP built in has an A/D converter that can only accept about 1–1.4 V RMS. So whatever voltage comes in, a preamp stage first has to shrink it to fit, and it must stay very quiet while doing so. That, he says, makes these amps inherently noisier. Find out whether the amp's gain control works on the analog signal before the converter or on the digital level after it. His test: with nothing plugged in, if the hiss drops as you turn the gain down, the amp is making the hiss. If that amp only drives tweeters, turn its gain down and boost the treble in the DSP instead. He explains there is room for that boost because treble content in most music sits around −20 dB.

4.10 #8 · Small DSPs need an isolating DC/DC supply; leave the MS-8 mic unplugged; whack the box for intermittents.

Turn-on thumps and engine noise with miniDSP-class units call for a converter that isolates signal ground (JBL's MS-2 got one); the MS-8 mic is multiplexed to its inputs. Sources: DIYMA 2011-11-11 · DIYMA 2011-11-21

In plain English: These are three separate tips. First, small, inexpensive DSP boards (the miniDSP type) can make thumps at turn-on and pick up engine noise. He says they need a DC/DC converter, a small power supply that isolates the signal ground from the car's ground. JBL's MS-2 processor was given one for this reason. Second, on the JBL MS-8 (a specific car DSP that sets itself up using a calibration microphone), unplug the microphone when you're not calibrating. The mic is "multiplexed" (switched) into the same circuitry as the audio inputs, and he says leakage from it adds noise. Third, for noise that comes and goes, give the unit a firm whack. He says that helps reveal bad solder joints or loose connectors.

4.10 #9 · Odd noise sources: leads touching metal and passive crossovers near electronics.

A sub's tinsel lead touching its grounded basket makes a ground loop on bass notes (or sends the amp into protect). For whine in one door, check the speaker leads aren't touching metal, then move the passive crossover: at low current its coils pick up radiated noise. Sources: DIYMA 2011-10-16 · DIYMA 2022-01-02 · DIYMA 2021-08-26

In plain English: Two unusual causes of noise. First, a subwoofer's "tinsel leads" are the flexible braided wires that carry signal from its terminals to the moving voice coil. The metal frame (basket) is often grounded to the car through its mounting screws. If a tinsel lead touches that frame, it can create a ground loop that is heard on bass notes, or it can send the amp into "protect" mode (it shuts itself off). Second, for whine in just one door, check that the speaker wires aren't touching metal. Then move the passive crossover (the filter box of coils and capacitors) away from the car's electronics. He says that when little current is flowing, its coils pick up noise radiated through the air.

4.10 #10 · Turn-on pop: find which box makes it, and know the IR2092 fix.

Unplug the amp's inputs: if it still pops, the amp is the source (more than a few mV DC at its outputs is unacceptable; fix a no-DC pop with delayed speaker relays); if not, delay the amp's turn-on. Amps built on the IR2092 class-D driver IC have an unavoidable on/off offset: put a resistor (~10 Ω) across the tweeter protection cap, and keep that cap even in active systems in case an amp fails. [RARE] Sources: DIYMA 2011-08-20 · DIYMA 2022-03-16 · DIYMA 2021-07-06

In plain English: A turn-on pop is a thump from the speakers when the system switches on or off. First find which box makes it: unplug the amp's input cables and switch on. If it still pops, the amp is the source. In that case, measure the DC voltage at its speaker outputs: he says more than a few mV (thousandths of a volt) is unacceptable. If there's no DC and it still pops, fit relays (electrically operated switches) that connect the speakers a moment later. If it doesn't pop with the inputs unplugged, the pop comes from earlier gear, so delay the amp's turn-on instead. Amps built around the IR2092 chip (a common class-D driver part) have a small, unavoidable voltage jump at on and off. Many tweeters have a protection capacitor wired in line to block bass. His fix is to put a resistor of about 10 Ω across that capacitor. And he says to keep that capacitor even in an active system (where a DSP does the crossovers), in case an amp fails.

The full series

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