Modern speaker judgment favors what’s easy to compare, not what’s easy to live with.
Peter Comeau has spent nearly 50 years in hi-fi. He’s the Director of Acoustic Design at IAG, overseeing Wharfedale, Mission, and Quad. And he’s arguing that the on-axis frequency response graph, the measurement audiophiles use to judge speakers, fails to predict how they actually sound.
The objectivist community at Audio Science Review is not taking it well.
Measurements Work for Designers, Not Buyers
Comeau isn’t saying measurements are useless. He’s saying they’re useful to the wrong people.
That distinction matters. A designer uses measurements to diagnose problems during development. When something sounds wrong, measurements help pinpoint why. But consumers and reviewers use those same graphs to judge finished products. They treat the measurement as a verdict when it was only ever meant to be a diagnostic tool.
The fundamental limitation is the test signal itself. Most standard speaker measurements use controlled test signals (often log sine sweeps/chirps) taken at a fixed mic position (commonly 1m on-axis), which don’t resemble music’s constantly changing content.
Comeau isn’t hedging, though.
He claims he’s deliberately being blunt because he thinks buyers and reviewers have started treating measurements as the main event, not a supporting tool.
In his view, the industry has shifted from using graphs to diagnose design problems to using them as a shortcut for judging whether a speaker is “good” or “bad.”
But if the standard graph misses so much, what exactly is it missing?
The Inaudible Resonance You Can Hear
Metal dome tweeters often sound “metallic.” And a basic on-axis frequency response plot doesn’t always make it obvious why, especially when the main resonance sits at 26kHz, above human hearing.
In other words, an ultrasonic resonance can leave a fingerprint in the audible band. Yet, Comeau argues a single on-axis curve still isn’t a reliable proxy for how that behavior will be perceived in real listening.
He makes a similar point about the baffle step, where energy begins to wrap around the cabinet and feed the room as frequency drops.
“The ear can discriminate between this and you can hear this junction between the sound coming directly at you and what is bleeding round into the rest of the room,” Comeau explained.
“It’s almost like another crossover.”
His design process reflects these hidden variables. Comeau typically goes through 15 to 174 crossover iterations per speaker. The Mission 770 sat at the extreme end, requiring 2.5 years of COVID-era development and all 174 iterations. Each version met the measurement criteria. Each sounded different.
When a speaker’s issues aren’t obvious in a simple response curve, you may still ‘hear’ them as subtle strain because your brain is constantly correcting. Comeau describes it as the “world’s best graphic equalizer” running constantly in your head, which corrects imperfections in real time.
Comeau describes this compensation as “a supercomputer working flat out and getting very hot.”
That processing cost, Comeau argues, is what separates a speaker that sounds “fine” from one that lets you relax into the music.
Critics Say Comeau’s View Is Marketing, Not Engineering
The objectivist community at Audio Science Review had a simpler explanation for Comeau’s heresy. He’s selling something.
The forum champions a measurement-first philosophy, and members treated Comeau’s claims as marketing rather than insight. If he admitted measurements worked, the pitch for IAG speakers would need to change.
Beyond the ad hominem attacks, some members engaged in the logic directly. The designer versus analyst distinction, critic Phorize argued, “pre-supposes that the designer and analyst aren’t measuring for the same reason, which is not a good assumption.”
Reviewers and designers both want the same thing from measurements. Why would those measurements fail for one group and not the other?
But the steady-tone criticism found similar resistance.
Neither side has produced blind test data to settle the question. But one DIY builder found the blind spot anyway.
One Builder Found What the Graph Missed
On Pink Fish Media, DIY designer Fatmarley reported building a speaker with a slight peak at the crossover frequency that showed up off-axis. The on-axis measurement didn’t show it. The speaker sounded poor.
He adjusted the crossover, kept the on-axis response identical, and fixed the off-axis problem. The speaker sounded better.
Fatmarley wasn’t arguing philosophy. He was reporting a design blind spot that affected real listening. The standard graph missed it completely.
What both sides seem to agree on, if nothing else: there’s more to how a speaker sounds than any single graph can capture. Reviewers keep publishing those graphs anyway.
The arguments rage on. And somewhere in the gap between what gets measured and what gets heard, 174 different-sounding crossovers wait to prove everyone right.
So called complex musical waveforms are simply lots of sinewaves cf Fourier transform LOL. I love it when people claim that music is somehow different….. definitely marketing
The problem with both ASR (an organisation I admire tremendously) and the purely subjectivist camp is their somewhat rigid thinking and unjustified certainty. The science is far from settled, there are unkown unknowns, and when biological systems and pschoacoustics are taken into account the picture is insanely complex. Yet there is no doubt that measurements can identify design flaws and suggest remedies. Yes, single tones are limited, but there are tools to get round that, and off axis measurements now form part of most analyses.
Its all good, its all useful, and the sooner we abandon the ridiculous tribalism and sneering at one another the better.
True. As an acoustician and former recording engineer, this is right on. The room effects outweigh the impact of a pristine design, because of the composite you listen to. But the design has to be a benchmark. Further, tones are a test, and a chirp with TDS or other means isolates the driver and the crossover of course. But the crossover in only an octave or two – midrange voice etc. The full fidelity response with the room is the transfer function.
Love your authentic approach to journalism, telling both sides of the issue!
Many measurement first devotees do tend to look first and foremost at the FR on axis measurement but the Klippel provides so many other measurements that round out the picture. So naysayers complaining about an on axis FR only focus are missing the fact that people like Amir at ASR and Erin at Erin’s Audio look at far more than that.
If anyone claims on axis me1surements alone ate good enough, that person doesn’t know what he is talking about. What Comeau is doing here is a bit of marketing yes but mostly just gate keeping, guarding his own position. I design my own speakers, designing a speaker is mainly just a big game of compromises, that’s it, that’s all he said here.
Measuring is a base requirement for speakers. After that accuracy of output to input, then last word is personal preference, which can and is all over the place! Some folks have NEVER heard a Unamplified sound!!!
Well most “audiophiles” are idiots. Especially those from ASR.
To me a good set of measurements doesn’t prove a speaker can sound good it proves it is properly engineered. I won’t bother to listen to speakers that measure poorly, but good measurements and correct priority can help you create a great one
I would imagine differences will show up in the transient response while simultaneously applying multiple frequencies. For instance measuring a white noise volume step & measuring the time dependent spectrum.