Contents7 sections
The Xylem DSP amplifier series is now shipping in three configurations. All three share one design decision that separates them from most architectural amplification: the digital signal processor is not locked. The integrator who walked the room sets every filter, every routing decision and every protection threshold, rather than working inside a curve chosen by a manufacturer who has never been there.
This is the long version — what is inside the amplifier, what PsySculpt actually does, how the tuning is arrived at, and how the whole thing is driven from XSCACE Studio once it is on the wall.
Three models, one core
The series launched simultaneously. Every model uses the same DSP core, the same champagne aluminium chassis and the same programming environment, so an integrator who configures one does not have to learn another.
| Model | Channels | Power | Typical use |
|---|---|---|---|
| Xylem 2 | 2 | 2 × 200W — 400W total | Stereo zones, or a bi-amplified single speaker |
| Xylem 3 | 3 | 1 × 200W + 2 × 100W — 400W total | LCR for home cinema and immersive audio, or 2.1 with a dedicated bass channel |
| Xylem 4 | 4 | 4 × 100W — 400W total | Multi-zone from one unit, or a four-speaker distributed array |
All three are Class D in a 1U rack chassis, 482.9 mm wide and 324.9 mm deep. The power figures above are per channel at 8Ω; the 400W total is the sum, not a per-channel rating.
What "programmable" actually means here
Fixed DSP — where the manufacturer sets the curve and the integrator works within it — is a reasonable answer for simple systems in neutral rooms. Most luxury residential work is not that. Hard surfaces, irregular geometry, speakers placed where the architecture allows rather than where the acoustics would prefer, arrays with timing differences between drivers: a manufacturer-set average cannot address any of it without compromise.
The Xylem exposes its Analog Devices ADAU1701 in full, through SigmaStudio — Analog Devices’ graphical signal-flow environment. What that opens up:
- Parametric EQ — up to 28 biquad filters per channel
- Crossovers — Butterworth, Linkwitz-Riley or Bessel, any order, any frequency
- A routing matrix — any input to any combination of outputs
- Channel delay — sample-accurate alignment for time-coherent arrays
- Dynamics — per-channel limiter and compressor with programmable thresholds
- Preset storage — up to 8 programs, recallable by IP command or hardware GPIO
SigmaStudio requires no code. Blocks are dragged onto a canvas, connected, and compiled to the unit over USB. Anyone who has used it on another platform is productive immediately.
PsySculpt: correcting for the ear, not the microphone
PsySculpt is the tuning architecture that runs inside the Xylem, and it is the part most often misdescribed. It differs from conventional room correction in what it optimises for: the human auditory system rather than a flat measurement.
Flat measurement and flat perception are not the same thing. The ear is markedly more sensitive to resonance between 2 kHz and 5 kHz — the presence region, where consonants, overtones and spatial cues live — and measurably less troubled by brief standing-wave energy below 80 Hz when it is transient rather than sustained. A system that applies one flat EQ target across the whole spectrum is treating the ear as a measurement instrument.
So PsySculpt corrects tightly where the ear is discriminating, and more gently below 80 Hz, where masking and reduced sensitivity mean aggressive EQ tends to make a system sound thin rather than accurate. It corrects toward perceptual flatness.
The low-extension side-chain, and what it is not
There is a separate function that gets confused with bass boost, and it is worth being precise about. The low-extension side-chain synthesises harmonics of fundamentals the cabinet cannot physically reproduce, so the ear infers a fundamental that is not actually being played. Nothing below the driver’s range is amplified, because there is nothing down there to amplify — a 12 mm cabinet does not suddenly move air at 40 Hz.
That is why the control reads "extend down to X Hz" rather than naming a filter: it sets how deep the psychoacoustic effect reaches, with a target as low as 40 Hz, and it rides along with system presets rather than being dialled in separately on every job.
How the tuning is arrived at
The honest answer to "how is it made" has two halves, and only one of them is publishable.
The method is straightforward to describe. Perceptual sensitivity curves were mapped against measured room responses across more than forty residential and commercial installation geometries during the Xylem development programme. That data set is what determines where correction is applied tightly and where it is deliberately relaxed.
The second half is the co-design. Each XSCACE passive speaker has its own driver behaviour and enclosure resonance — the Acacia 6 runs 45–300 Hz at 84 dB sensitivity, the Acacia 10 runs 35–300 Hz at 88 dB — and the correction profile is calibrated per model, not per range. The amplifier knows which speaker it is driving, and applies model-specific correction on top of room-specific correction. The result is a system that is not over-processing to fight a limitation it should be respecting, and not under-correcting a problem that is physical and predictable.
The specific curves, coefficients and weighting that come out of that programme are not published, and this article is not going to pretend otherwise. What is published is the shape of the approach and the reasoning behind it, so an integrator can judge whether it suits the room in front of them.
One consequence worth stating plainly: the processing runs inside the amplifier. No external processor, no extra rack unit, no third-party DSP in the signal chain. The intelligence sits where the amplification does, and the amplifier is co-designed with the speaker. That integration is the precondition for the whole approach.
Driving it: XSCACE Studio
SigmaStudio is the deep programming layer. For day-to-day configuration and commissioning, the Xylem is driven from XSCACE Studio, which discovers every unit on the network with zero configuration — the amplifiers announce themselves, so there is nothing to pair and no address to type.
Live DSP control
The full signal chain opens up: per-channel gain and delay, four crossover modes with the curve rendered live, and six bands of parametric EQ per channel. Changes hit the DSP over the local network with no cloud round trip. The delay block is accurate to 3 µs on a commanded 5.000 ms.
X-Sense room correction
Guided calibration in minutes: sweep measurements from up to three listening positions, corrective EQ fitted against six sound profiles, and room diagnostics. A verification pass then re-measures with the EQ live and overlays before, target and after — so the correction is demonstrated rather than asserted.
An eleven-instrument calibration toolkit
X-Sense EQ, SPL Meter, Delay, Levels, Polarity, Frequency Response, Sub Align, Transfer Function, RT60, RTA and Mic Cal. A single run yields distance, level, polarity and signal-to-noise for every channel, and any measurement can export the raw microphone capture as a 16-bit WAV — so a disputed result can be taken apart offline rather than argued about.
It refuses rather than guesses
A measurement tool that returns a confident wrong number is worse than one that returns nothing. RT60 follows ISO 3382-1 practice — Lundeby truncation, a minimum 12 dB fit span, an r² ≥ 0.90 quality gate — and where a room will not support a defensible reading, it says so and tells you which way to move: raise the level, or quieten the room.
Where the Xylem is the right answer — and where it is not
It is not the right amplifier for every job. It earns its place when:
- The room needs site-specific EQ — hard surfaces, low ceilings, awkward geometry
- Subwoofers need real crossover management and sub-satellite integration
- Several speakers need sample-accurate time alignment
- The brief is hospitality or commercial, where one program is deployed identically across many identical rooms
- The integrator specifies at a level where manufacturer-defined tuning is a limitation rather than a convenience
For a straightforward stereo zone in a furnished room with no acoustic problems to solve, a fixed-DSP amplifier will do the job and cost less to commission. Specifying a Xylem there buys capability that will never be used.
The preset library
A companion SigmaStudio preset library covers the XSCACE passive range — Bonsai, Cane, Camphor, Juniper, Ghost, Cedar and the Acacia subwoofers — each profile calibrated to that speaker’s driver and enclosure.
The presets ship as fully editable SigmaStudio project files, not locked binaries. They are a starting point rather than a ceiling: site-specific adjustment, room correction and crossover refinement are expected, and the files are structured so that work is efficient rather than laborious.
Specifications for each model are on the Xylem 2, Xylem 3 and Xylem 4 pages. XSCACE Studio is a free download, and the PsySculpt write-up goes further into the psychoacoustics. For pricing and availability in your territory, speak to your regional distributor.
