Dougan Ref. K — the design
Nothing on the Ref. K is styled. The egg behind the midrange, the cast waveguide, the alignment of the bass, even the three-degree lean — each one is the survivor of a long, patient search: some thirty thousand finite-element iterations, reviewed by AI over 48 hours and questioned by humans. The shapes you see are what the mathematics left standing.
See the method ↓ ← Back to the Ref. K01The method
Acoustic design used to be a craftsman's guess refined by a prototype or three. Cast stone does not do a prototype or three — so the prototypes became mathematics. A model proposes a shape; finite-element analysis asks how it vibrates, resonates and radiates; the answer feeds the next proposal. Round after round, around the clock.
AI sits inside that loop as the tireless junior engineer: it reviewed some thirty thousand iterations in 48 hours, discarded the losers and brought the humans only the shapes worth arguing about. The computer never got tired, and we never got sentimental.
02Modal analysis
Put a driver in a box and the air inside becomes a set of organ pipes: every pair of parallel walls rings at its own family of frequencies, and each of those standing waves pushes back on the cone exactly where the music is most exposed — voices, strings, the midrange. A modal analysis maps every one of those waves, and how hard each couples back to the driver.
The egg is what survived that analysis: no parallel surfaces anywhere, a flat front for the driver, and a narrowing tip that herds the back wave into damping wool, where it quietly dies. Eight litres, cast into eighteen kilograms of stone. The shape is not a flourish. It is the printout.
03Directivity
A dome tweeter on a flat baffle sprays treble wide at some frequencies and beams it at others, and the room hears that inconsistency even when the listening chair does not. The waveguide fixes the handover: the baffle itself continues in cast stone out to a 180 × 160 mm mouth, loading the dome so that its dispersion matches the midrange exactly where the crossover passes one to the other.
The edges came out of the same study. The roundover grows from nearly nothing at the foot of the baffle to wide and soft around the tweeter, where a sharp edge would be audible as diffraction. Almost impossible to machine in MDF. In a cast material — free.
04The alignment
Just under 60 litres, tuned to 30 Hz through two 65 mm front ports: the modelled response reaches about 31 Hz with a lift of barely a decibel. It is a deliberately conservative alignment — it keeps the speaker honest near a wall, and it never asks the ports to raise their voice.
And because a model is a model, the ports are cast long and trimmed to length on each finished, measured cabinet. The tuning is set on the speaker you take home, not on the one in the computer.
05Time
The tweeter's dome and the midrange's cone do not sit at the same depth in the cabinet, so stood bolt upright, their sound would arrive at slightly different moments. The error is measured in microseconds — which is exactly the scale your ear uses to decide where a voice is standing in the room.
We modelled the arrivals for a seated listener and tilted the entire face back three degrees; the tweeter's recessed waveguide takes care of the remainder. It photographs as styling. It is timing.
06Then the humans
Every drawing then went through independent review on its way to the shop floor — more than 150 findings raised, argued and closed before the first production pair. Most were small. That is the point: at this price, small is where the sound lives.
We argue on paper because paper is the last cheap place to be wrong. Before a first pair can even be attempted — the machine programming, the mould-making, the materials bought and cured — the investment in a speaker that does not yet exist is extraordinary. Nothing here is run up on the fly. The Ref. K is one integrated acoustic system — stone, cabinet, chambers, crossover — and it has to work together flawlessly the first time.
We treat the modelling as a design aid, and we are careful to call it nothing grander. The final voicing is done the old way, with microphones, in a room, by ear.
Every part checked against every neighbour — magnets, braces, panel edges — before anything was cut.
Each glue joint reviewed for area, grain and clamping: a box this heavy does not get second chances.
Down to the depth of a screw — literally. Finding and fixing boring problems is what review is for.
Cast long, then trimmed to length once the finished cabinet has been measured — the one place we let reality overrule the model.
“The computer proposed thirty thousand shapes. The last word went to a pair of ears.”
Request an auditionOne hundred pairs a year, each pair numbered. Auditions by appointment — John Dougan, +7 903 220 0175