“If it measures flat, it will sound peaky.”
Decades of acoustic research say flat frequency response equals good sound. Michael Børresen says that research is wrong.
Børresen is the CTO of Børresen Acoustics and co-founder of Raidho, a Danish designer whose speakers range from $5,500 to roughly $500,000 a pair. In a recent interview with Next Level HiFi, he argued that the industry’s obsession with flat measurements has made modern speakers sound harsh, fatiguing, and fundamentally unnatural.
“If it measures flat, it will sound peaky,” Børresen told the channel.
This claim directly contradicts Floyd Toole’s landmark research at Harman International.
So what’s going on?
Where the Extra 30% Goes
Børresen’s argument hinges on what happens at crossover points, where multiple drivers share responsibility for the same frequencies. When a woofer and tweeter both reproduce the same note, their sound waves interact. If the waves arrive out of phase, they partially cancel each other.
This is mainstream acoustics. At a 90-degree phase shift, signals sum to roughly 29% less energy than they would in perfect alignment. Speaker designers have understood this for decades.
His interpretation takes it further.
The compensation required to achieve a flat measurement, in his view, creates an aggressive, forward sound that listeners experience as harsh.
His solution? Intentionally dip the crossover region.
The BBC’s 2.5 kHz Dip
Børresen isn’t the first designer to question flat tuning. The BBC’s legendary studio monitors, developed in the 1970s, featured a notable dip around 2.5 kHz, what engineers call the “presence region.”
Research by H.D. Harwood at the BBC suggested that flat output made orchestras sound “extremely close,” while a slight dip provided a more natural listening perspective.
But the BBC dip has multiple competing explanations. Some engineers attribute it to off-axis compensation or masking early plastic cone defects, not psychoacoustic preference. As one Harbeth forum discussion notes, there’s “much myth, folklore and misunderstanding about this subject.”
Børresen also takes aim at driver materials.
This technical concern is real, as soft cones do exhibit breakup within their operating range. But the engineering picture is more nuanced than Børresen suggests. Paper composites with felt or wool often achieve the best balance of stiffness, mass, and damping. Metal cones have fewer breakups but more dramatic ones when they occur.
He cites the Yamaha NS10 as his counterexample, a flat speaker notorious for listening fatigue.
There’s one problem with that claim.
What the NS10 Actually Measured
Research by Philip Newell at the Institute of Acoustics shows the NS10 doesn’t actually have flat frequency response. It features an inverted-V characteristic that peaks in the midrange. The opposite of what Børresen described.
“While its frequency response wasn’t particularly flat,” Newell wrote, “in terms of time-domain and distortion performance it was outstanding.”
The NS10’s harsh reputation stems from elevated treble, not crossover flatness.
More fundamentally, Børresen’s central thesis, that flat-measuring speakers sound worse, runs against the weight of controlled listening research. Toole’s four decades of blind tests at the National Research Council and Harman consistently showed listeners prefer speakers that measure flat.
Not every designer agrees with the measurement-first approach. But even those who value listening over graphs don’t dismiss flat response entirely.
Jones trained at KEF, developing measurement techniques. He acknowledges that no speaker achieves perfect flatness. But he doesn’t argue that flat inherently sounds bad.
Half a Million Dollars of Confidence
Børresen remains unconvinced by the research. For him, measurements miss what matters most.
“Flat is flat says nothing about the quality what comes out,” he said. “It just says something about the SPL level.”
His focus is on timing and transient response, how quickly a driver starts and stops, which he considers “so much more important than frequency response.” These qualities are harder to capture in a frequency graph, and Børresen argues they’re what separates good speakers from great ones.
It’s worth noting the business context. Børresen designs speakers that compete in a market where competitors like KEF and Purifi publish detailed measurements and charge considerably less.
Forum discussions on Audio Science Review, a measurement-focused community, have questioned the value proposition.
Whether Børresen’s approach represents genuine insight or luxury-market positioning, his speakers continue to find buyers willing to bet $500,000 that he’s right.
For the difference in price, me as a non audiophile (aka someone who doesn’t need $400 diamond plated cables) I can accept a the difference.
An explanation that a flat response may cause more harshness or listener fatigue may be due how the frequency response is measured or realised at the design stage. It may be flat at a metre or so measurement, but at a distance the high frequency at the crossover point will be more beamed as the effective speaker size producing the crossover frequency is larger than any other frequency as it will be the woofer + tweeter + interspace size added. And that’s for a relatively short wavelength envisaged to be taken over by a small tweeter. So a dip may help, depending on the intended listening distance. Other remedies include using a lower crossover point, more smaller drivers at (3-way), closer spaced drivers, a large decoupled spacing, a single driver, or coaxial driver, or line driver. That’s to address this issue, whilst each solution has other issues. It’s all compromise.
Having had extensive conversations with Floyd Toole and Ian Paisley, I lean towards their analysis. The biggest issue in high end audio has been harshness on the amplification end, not speakers. Companies such as Krell, Levinson, and others have to me, become unlistenable. I find myself listening to Electrocompaniet, vintage hk, and vintage Yamaha. I could add Hafler. I’ve got a 9370 that is just freakish in it’s presentation. Amazing.
All these measurements are done with a single microphone at the normal listening height and distance. Does anyone use an artificial head plus ear canal to measure what the ear hears rather than free space.
There are a mixture of reasonable callouts and misused research creating uncertainty and confusion in this perspective.
First of all, selling a speaker for an enormous sum of money does not inherently represent superior design. In Toole’s own research, there’s a distinct focus on eliminating sighted biases, which include both aesthetically pleasing and larger sized models.
That aside, the claim that crossovers actually result in a boosted X-over region is both true in some literal senses, while also practically irrelevant for designers in the real world. As a designer myself, using a single on-axis measurement for design would be an immense wast of time and energy. Multiple off-axis measurements are needed to extrapolate in-room response, sound power, and imaging performance. When modeling and re-testing, any quirks of peaking behavior can be discovered and addressed.
In a situation where a designer is either only focusing on on-axis performance or does not understand the off-axis behavior of crossover types would this type of callout be meaningful. The comments on transients and breakup modes are legitimate and can be preferential in a design approach.
Andrew Jone’s comments are also valid but taken out of context. Remarkably flat linearity may be sacrificed for better overall in-room response, for example. All of this is accounted for when designing with a full picture of measurements.
A few years ago I was at Filtronique in Montreal where I auditioned 20 K speakers by Wilson, Audioquest, Magico and Raidho. I immedialtely noticed that the Raidhos had a cleaner sound, much like an electrostatic. I Bought the X3s. Only later I understood that what made them so clean was that the crossover frequencies were masked one pair of woofers crossing over were masked by the other pair. I now have Oephis Transcendence in another system and I am struggling with harsness. Dirac live removed a bump around the crossover frequencies and solved 80% of the problem. I think Borrensen is absolutely right. Flat is good, but not at the crossover frequencies. SPL is misleading at those frequencies.
So high fidelity has a problem: too much fidelity.
I believe that crossovers should be first order. This keeps the drivers all in phase with each other exactly like live natural sound. The step response graph will show this. The next important is timing, when all the drivers are aligned in time and phase the sound of the frequencies impacting your ears all arrive at the same time. For example when you listen to a snare rim shot , it incorporates the tweeter, midrange and woofer and the rim shot is so exact and real when the speaker is time and phase correct. I have measured the Borresons step response and the response shows all drivers are not in phase with each other which is supposed to be a straight triangular line without dips or peaks, this indicates the drivers are not in phase with each other. Vandersteen audio in the USA is one of the only speakers in the world that does this correctly. Give them a listen. You will be astonishingly surprised.
I’ve never heard any of his speakers sound even average, not the Raidho or his own brand name models.
When he buy some decent meaduring kit and starts producing spinoramas he might be worth listening to.
The brain does not like everything perfect. Too even frequency seems less natural and the faster the brain gets tired. There should be about 5% unevenness. Large distortions are also bad.