One product, built without compromise.
Analog Realism keeps a deliberately small catalogue. Each release is engineered to be useful for years, not seasons. Today: Sphinx 101.
Sphinx 101
Master bus processor. Component-accurate analog modeling with TrueRail Technology. Three main circuits - SLL, Nevy, Amok - with fifteen analog modeling mechanisms tuned to the harmonic and dynamic signatures of the modeled console classes. Added with known hardware circuits - Pultey, Nevy, SLL, Amok and Maney - for EQ, Filter and all Dynamic Modules.
Fifteen mechanisms. Always active.
-
01
Summing amplifier finite bandwidth
Real amplifiers aren't perfect. Our modeled summing amp rolls off at the frequency extremes, adding warmth that no EQ curve can replicate - because it's not EQ, it's physics.
-
02
Per-component manufacturing tolerance
No two real capacitors are exactly 100nF. Every component in Sphinx has randomized tolerance within real specs (±1% resistors, ±5% caps, ±10% transistor gain). Your left and right channels process through slightly different circuits - natural stereo depth impossible with mathematically perfect components.
-
03
Thermal drift
Three independent slow oscillations modulate circuit parameters over time. The sound breathes - never quite static, just like hardware that's been powered on for an hour. Per-component value modulation is small but audibly active across the chain.
-
04
Power supply rail sag
When the compressor clamps hard, it draws current from the shared supply. The rail voltage dips, affecting every other stage's headroom and saturation point. This is the "glue" that makes analog bus compressors feel cohesive. Every module pulls current AND reads the rail back to adjust its own operating point - a two-way loop, just like real hardware.
-
05
Cross-channel crosstalk
Real hardware shares a chassis, a power supply, a circuit board. Signal leaks between L and R - frequency-dependent, stronger in the lows. Sphinx models this coupling, creating a "wide but cohesive" stereo image that mono-summed processing can't achieve.
-
06
Transformer core hysteresis
The transformers use Jiles-Atherton magnetic model - the same math used in electrical engineering to model real cores. They remember their recent magnetization history, producing asymmetric, program-dependent saturation that no static waveshaper can replicate. Each core's harmonic balance is tuned to match published electrical measurements of the modeled unit.
-
07
Harmonic chain accumulation
Each stage adds its own tiny harmonic signature. By the time audio passes through drive stage, transformer, compressor, EQ, and output transformer, these harmonics have accumulated and interacted in ways unique to this specific chain. Measured: H2 through H7 all present with circuit-dependent ratios.
-
08
Class-A crossover nonlinearity
The drive stage models the slight crossover distortion of real amplifier topologies. SLL (BJT) produces clean odd-order harmonics. Amok (tube) produces rich even-order harmonics with H2/H3 ratio exceeding 5:1. This is the "warmth" and "presence" that defines each circuit's character.
-
09
Crosstalk frequency shaping
The L/R coupling isn't flat - it's stronger at certain frequencies, consistent with how real PCB-trace coupling behaves. This creates frequency-dependent stereo interaction that contributes to the three-dimensional imaging analog consoles are known for.
-
10
Compressor program dependence
The compressor's behavior changes based on what it's been doing. A Vari-Mu tube compressor working hard has a different gain reduction curve than one that's been idling. Beat 8 of a drum loop produces measurably different compression than beat 1. Measured: up to 82% program-dependent variation.
-
11
Transformer memory
The core's saturation curve depends on recent signal history. A loud bass note changes the magnetic operating point, affecting how the transformer handles the next transient. This "memory" creates the living, breathing quality that separates real transformers from static saturation curves.
-
12
Inter-module phase interaction
Each module introduces frequency-dependent phase shifts. These interact across the chain, creating subtle constructive and destructive interference at module boundaries. This is what gives real analog chains their characteristic "depth" - the sense of front-to-back dimension that digital processing rarely achieves.
-
13
Programme-responsive noise floor
Device noise rises gently with signal level and drops during quiet passages. The console's background isn't static - it breathes with your programme the way a real chassis full of active components does, shifting on fades and reverb tails.
-
14
Tube transient sag
A loud hit drives the tube grid positive for a moment, charging the upstream coupling capacitor and pulling the bias negative. Gain dips briefly and recovers over about 100 ms - the classic breathing quality tube stages have after a hard transient.
-
15
Inter-transformer coupling
The input and output transformers sit close together on the virtual chassis. A fraction of the input core's magnetic state reaches the output core through fringing flux - the subtle iron-to-iron interaction that happens in any real hardware with two transformers sharing a frame.
Two skins. Same controls, different temperature.
-
01 - Amok skin / full panel
-
02 - Nevy skin / full panel
-
03 - Signal analysis / see the verified Plugin Doctor measurements
Install Sphinx 101.
Universal installers.
One installer for both Apple Silicon and Intel Macs, plus the Windows build. Recommended for everyone.
Single installers.
Native builds for one specific architecture. Use these only if you need to force a native install or troubleshoot the universal package.
Specifications.
AU on macOS 11 (Big Sur) or later - two separate single-architecture builds: Apple Silicon (arm64) and Intel (x86_64). The Intel build also runs on Apple Silicon via Rosetta 2; the Apple Silicon build does not run on Intel Macs. VST3 on Windows 10 or later (x64). 4 GB RAM, approximately 50 MB disk.
Sample rates 44.1-192 kHz (verified identical character). Buffer 64-2048 (bit-identical output). Internal precision 64-bit. Oversampling 2× / 4× / 8× / 16× applied globally - one minimum-phase polyphase IIR half-band wraps the entire signal chain, every module runs at the elevated rate. The minimum-phase design eliminates pre-ringing; phase rotates monotonically with frequency, matching analog hardware. Noise floor −110 dBFS; full-chain measurement floor sits in the −70 to −74 dB region from the component noise model. Per-element alias measurements documented in the shipped HF Alias Regression test report; full-chain Plugin-path alias measurements in the shipped Plugin-Path Alias Regression test report.
−6 dBFS = 0 VU. Bypass transparency −68.28 dB residue (TrueRail chassis character). Polarity non-inverting at all frequencies. Sample-accurate DAW latency compensation.
79 parameters, all automation-compatible. 15 TrueRail mechanisms always active. CPU: 3-10 % single-core at 2× on Apple Silicon. Authorization one-time, 2 seats per serial.