Deep Dive
Why Andrew Jones Started Measuring Instead of Listening
Andrew Jones grew up wanting to build speakers but faced a fundamental problem: how would he know if his designs were good? As a teenager, he lacked both the trained ear and the high-end listening rooms needed to evaluate his work. He realized early that his listening skills were underdeveloped and that guessing wouldn't cut it. So instead of relying on subjective judgment, he pursued measurement. By university, studying physics, he was already thinking about how to accurately quantify speaker performance using tools like delay spectrometry and other emerging techniques. This foundation set the trajectory for his entire career — measurement as the foundation, listening as the interpreter. He's never wavered from that approach, even as technology made measurements more accessible to the masses.
The Danger of Cheap Measurement Tools
Andrew points out a critical paradox: fifty years ago, the industry moved away from simple sine-sweep room measurements to more sophisticated techniques like delay spectrometry. Yet today, with inexpensive computer-based tools available, people are reverting to those outdated methods, believing they now understand how to measure. Erin agrees completely — he's gotten measurements wrong countless times himself. The real problem is that without deep understanding of measurement theory, cheap equipment lets you get the wrong answer quickly and confidently. Andrew notes that KEF was using £100,000 HP analyzers in the late 1970s for both design and production testing, ensuring consistency. Modern gear democratized access but not knowledge. Erin emphasizes this by spending 15+ hours per speaker using multiple ground-plane and quasi-anacoic techniques specifically because he doesn't trust a single measurement method. The Clipple analyzer solves many problems, but only if you understand what you're looking at.
Anacoic vs In-Room: Why Flat Measurements Don't Mean Flat Sound
Erin raises a misconception he encounters constantly: the belief that a flat anacoic measurement will produce a flat in-room response. It doesn't work that way. A monopole speaker naturally radiates omnidirectionally at low frequencies but becomes increasingly directional as frequency rises. This means treble energy concentrates toward the listener while low frequencies spread throughout the room. The result is that flat anacoic speakers naturally tilt down in rooms, losing high-frequency energy — which is actually desirable. Force a speaker to measure flat in-room, and you've overcompensated, creating a bright, fatiguing sound that hurts your ears. Andrew adds another layer: what seems like boosted treble to one listener might sound like scooped midrange to another. Everything is relative. Both perceptions are objectively correct; they're just different subjective interpretations of the same frequency curve. This relativity is why designers can't simply chase numbers.
The Eureka Project: Modeling Room Reflections in Real-Time
In the late 1980s and early 1990s, KEF collaborated with Bang & Olufsen and the Technical University of Copenhagen on the Eureka research project — a groundbreaking effort to understand how speakers interact with rooms. Andrew hand-built 32 concentric two-way drivers and suspended them on fishing floats around a listening chair inside Europe's largest anechoic chamber. Each driver was fed by current-driven amplifiers and connected to a discrete 32-channel DSP processor built in-house. The team created a computer model that could simulate any room shape and size, calculating reflections grouped by three-degree angles and modeling frequency absorption characteristics of any surface. This allowed real-time listening experiments: turn on the floor, turn off the walls, isolate the ceiling bounce, adjust distances and absorption. The system revealed what contributed most to perceived sound quality. Floyd Toole later built on this work at the National Research Council of Canada, eventually establishing the scientific foundation for preferred speaker directivity and power response characteristics.
Why Directivity Matters More Than You Think
Daniel Queen's research paper at London University identified speaker directivity as the next frontier in design — a thesis that drove KEF to develop Uni-Q, the coincident driver concept. The insight was simple: we don't listen to speakers in anechoic chambers; we listen in rooms full of reflections. A pulsating sphere (the ideal omnidirectional source) maximizes room contribution, which actually hurts sound quality in typical listening rooms. Through Floyd Toole's research, designers learned that the ceiling reflection is the most problematic because it's consistent for both speakers and arrives with enough delay to interfere. If drivers are coincident in space and time, the off-axis response becomes much more consistent, reducing ceiling bounce interference. However, directional speakers have a trade-off: they act like horns, concentrating sound toward the listener, which reduces room reflections (improving detail) but also makes imaging feel like headphones if directivity is too aggressive. The challenge is balancing directivity across frequencies — a task made harder because wavelength physics forces different directivity at different frequencies.
Measurement Data Is Objective, But Interpretation Is Always Subjective
Andrew makes a crucial distinction: the measurement process itself can be standardized and objective — that's why the Clipple methodology is now incorporated into IEC and ANSI standards. But interpreting that data is inherently subjective. No speaker measures perfect. Far from it. Andrew wishes speakers were like amplifiers, which can measure flat, but speakers have one input and multiple directional outputs, each mattering for real-world performance. When designing, Andrew always listens first to establish what he hears, then makes measurements to correlate against those impressions. If he hears an anomaly, he looks for it in the curves, postulates a fix, remeasures, and listens again. This loop is susceptible to bias — he can convince himself the problem is solved when it might not be. But without listening, measurement alone tells you nothing about sound quality. Erin adds that correlating subjective impressions to objective data requires patience and multiple measurement techniques. He sometimes takes large speakers outdoors for ground-plane testing specifically to gain confidence before publishing data online.
Why You Can't Chase 'Live Sound' — And Shouldn't Try
Andrew spends considerable time on a reality most audiophiles ignore: we have no idea how the original performance sounded. Recordings are interpretations shaped by microphone choice, placement, EQ, compression, and mixing decisions. He cites pages of debate on Sweetwater forums about recording string bass — should you use a condenser or dynamic mic? Over the bridge or through the F-hole? Everyone arguing, meaning no one actually knows how the engineer captured the sound. If recording professionals can't agree on microphone technique, how can a listener know whether their speaker's presentation of that bass is accurate? The purest approach would be dummy-head binaural recording, the only method that captures what a listener actually heard at that moment. Anything else is a facsimile compromised from the start. Erin emphasizes this with a concrete example: a female vocal might have been recorded bright, then the engineer EQ'd down 5-6 dB at 5kHz because they thought her voice was too sibilant. A listener hears the EQ'd version and has no reference to know sibilance was reduced. Chasing live sound through a speaker means chasing a phantom.
The DIY Speaker Competition: Designing for Judged Preferences
Andrew judges an annual DIY speaker competition at Parts Express, evaluating about 60 different speakers across multiple judging panels. Each judge brings a pre-selected playlist determined months before the competition — a fact that troubles Andrew. Contestants essentially design their speakers knowing exactly which music will be played by which judge. That's backward. He'd prefer to walk in and surprise them with different material. But the current system reveals something valuable: contestants spend months optimizing their designs not for their own listening room, but for what they believe judges want to hear. It's a direct window into the subjective preferences driving real-world design. The experience reinforces Andrew's belief that ultimately, someone's got to buy the speaker or it's just another DIY project gathering dust. Designers have to choose what they represent through their work and hope enough people connect with it to make a business viable. This is why Andrew can't second-guess every listener's preference — he does what he thinks is right and trusts that his vision resonates with enough people to sustain his career.
The Amplifier Comparison: Why Speakers Are Harder to Design
Andrew's twin brother designs amplifiers, which operate on principles fundamentally different from speakers. An amplifier has one input and one output, making it straightforward to measure: flat response, low distortion, stable impedance. Andrew jokes that amplifier design must be easy because the data is clean. Speakers, by contrast, have one input and multiple directional outputs that all matter — on-axis and off-axis performance at every frequency. You can't just sum them together; you have to group them by reflection path (floor, ceiling, walls, late arrivals) and understand how each contributes to the listening experience. This is why measurement alone isn't enough. Amplifiers can be designed to performance criteria with minimal listening. Speakers demand listening because the measurement-to-sound correlation is rich enough to support it. Andrew wrestles with sibilance as an example: amplifier A might sound more sibilant than amplifier B, yet he can't see a measurement that explains it. With speakers, it's obvious — the measurement shows exactly what's causing the perceived sibilance. That's the difference between a device you can judge on spec sheets and an instrument you have to hear.