HELIOS-8
Eight voices of analog synthesis in 42 HP. Two oscillators per voice, a discrete waveshaper between them, and per-voice control voltage over everything that matters.
Two oscillators, one disagreement.
Each of the eight voices carries two analog oscillators with independent pitch and FM depth. They are not meant to agree. The detune control sweeps a real analog spread, not a cached copy, so a HELIOS-8 pad is never the same waveform twice. The discrete waveshaper sits between them and folds the sum before the mixer ever sees it.
The shaper is the reason for the module. Wavefolding a summed pair of analog oscillators produces overtones that respond to playing dynamics: hit the voices harder and the fold point moves, so chords breathe. A wavetable cannot do this, because a wavetable is a recording of a moment and not a living argument between two circuits.
Sub and noise live in the mixer, both normalled to the filter input, both breakable with a patch cable. Per-voice CV is exposed for pitch, FM depth, shaper fold, filter cutoff, and envelope amount. If a control is on the panel, there is a jack for it. No exceptions.
- Polyphony
- 8 voices, full analog path
- Oscillators
- 2 per voice, discrete transistor core
- Waveshaper
- Discrete foldback, temperature compensated
- Filter
- 2-pole multimode, resonant to self-oscillation
- Envelope
- Analog ADSR per voice, velocity to amount
- Voicing modes
- Poly 8, Poly 4 stack, Unison 8, Mono legato
- Tuning tolerance
- ±0.02% across full CV range, measured
- Drift character
- Configurable: locked, gentle, wide
Read the path. Then patch it.
This is the actual topology, simplified to one voice. Heavy strokes are normalled connections you can break with a cable.
The sheet.
Measured on the production line, per batch, and printed on the calibration certificate inside your box. The numbers below are the trailing-12-month worst case across all batches, not the marketing best case.
| Format | Eurorack, 42 HP, 26 mm deep including jacks |
|---|---|
| Power | +12 V 340 mA / -12 V 310 mA / 0 mA 5 V measured, all voices active |
| Polyphony | 8 voices, voice allocation automatic or fixed |
| Oscillator core | Discrete transistor, temperature compensated, 0.1 Hz to 22 kHz |
| Pitch tracking | <0.04% error, 0 to 10 V, measured at 20 points |
| Waveshaper | Discrete foldback, 2 to 14 folds, CV over fold depth |
| Filter | 2-pole multimode, LP / BP / HP, self-oscillates from 40% resonance |
| Envelope | Analog ADSR, 0.5 ms to 22 s per stage, retrigger modes 3 |
| Signal to noise | 102 dB(A) at full output IHF-A weighted |
| Crosstalk | <-98 dB between voices measured at 1 kHz |
| Latency | 0.12 ms gate to audio analog, no processing delay |
| Depth switches | Poly 8 / Poly 4 stack / Unison 8 / Mono legato, front panel |
| Firmware | 3.2.0, CERN-OHL-S, signed community builds accepted |
| Warranty | 5 years parts and labor, transferable with the serial |
“The first polysynth where I understood what every voice was doing, because the panel told me before the manual could. The detune control alone justifies the module.”RAVEN M. / MODULARSYNTH WEEKLY / REVIEW 2026-04
“We tracked it against our reference poly for a month. Tuning moved 0.017% over a full studio day with the lights on. That is not drift, that is a spec.”BENCH TEST / OSC/26 QC REPORT / BATCH 2026-07
“Folding two analog oscillators instead of one is the trick nobody told you about. Chords get a physical presence that software cannot fake and hardware rarely tries.”PRODUCER INTERVIEW / SIGNAL & ROOM / PODCAST 214
Three faceplates, one circuit.
SELECTED: STANDARD / GRAPHITE ON AMBER
All variants share the identical board, firmware, and calibration procedure. The faceplate finish is the only difference, and it is honestly the least interesting thing about the module. Prices are identical too.
PROCESSING / 18 HP
TESSERA
Granular texture under the pads. The pairing that started two records we know of.
MODULATION / 14 HP
MERIDIAN
Four envelopes whose drift is musical. Built to sit next to HELIOS-8's CV inputs.
PROCESSING / 16 HP
FLUX CORE
Stereo space and foldback. Sends the voices into a room and lets them argue with it.
Build slots for batch 2027-01 open November 3.
Each batch is 120 units, serial-numbered, and calibrated as a group so the voices within a unit and between units stay within family. Request a slot now; we confirm within two business days.
R-081
The eight-voice pad
Poly 8, drift gentle, VCO B detuned to 9 o'clock, shaper fold at 11, filter low. MERIDIAN channel 1 on cutoff, channel 2 on fold depth, both slow. Full recipe on the recipes page.
R-084
Certified bass
Mono legato, drift locked, sub full, VCO A only, through KILN at its printed knee. The one patch where locked tuning beats drift, because the argument belongs to the note and the filter.
R-089
Feedback choir
Mono out into FLUX CORE, then return 2 patched into send 1, which the routing matrix assumes you will never do. Musical below roughly 70% feedback. A different instrument above it.
Every HELIOS-8 ships with a one-page certificate printed from the batch sheet, the same sheet you can read in full on the technology page. The numbers below are the fields on that certificate and what each one is promising.
- Tuning error, per voice: the deviation of each oscillator's calibration from nominal at 261.63 Hz, measured after the thermal soak. Batch 2026-10 worst case: 0.018%.
- Tracking, 20 Hz to 20 kHz: how faithfully pitch CV maps to output frequency across the whole range. Under 0.04% means a note you tuned stays a note you tuned.
- S/noise, full-scale: 102 dB(A) unweighted, measured on the batch rig with the filter at unity and the shaper off.
- Crosstalk: below 98 dB between any two voices. In practice: nothing you will hear at any gain stage you own.
- Latency, CV in to audio out: 0.12 ms. If a patch ever feels late, the latency is not where the time went.
- Power: +12 V 340 mA, -12 V 310 mA. Size the supply for the peak, not the average.
CERTIFICATE SNAPSHOT / BATCH 2026-10
- Tuning, worst voice
- 0.018%
- Tracking, worst point
- 0.037%
- S/noise, A-weighted
- 102 dB
- Crosstalk, worst pair
- -99.4 dB
- Latency
- 0.12 ms
- Jack cycles, sample lot
- 10,400 verified
SIGNED BY THE BENCH THAT BUILT IT
Eight voices: real analog or voices in name only?
Eight full analog voices, each with its own VCO pair, shaper stage, and filter, all measured on the batch rig. The polyphony counter in the panel confirms it live; press eight keys and watch it say eight.
Will it fit my case?
42 HP wide, 26 mm deep behind the panel, +12V 340 mA and -12V 310 mA peak. The peak figure is on your certificate; size the supply for it, not the idle draw.
Why a shaper before the filter?
Because folding harmonics before the filter gives the filter something interesting to reject, and filtering after folding stays musical deeper into the range than filtering before it. The signal-path diagram on this page shows the order; the recipes page shows what the order buys.
What comes in the box?
The module, a power ribbon, mounting washers, the batch certificate signed by the bench, and a card with the warranty terms on one side and the bench email on the other. No software, no account, no registration step.
PAIRS WITH
MERIDIAN
Two LFOs on cutoff and fold turn HELIOS-8 from an instrument into a living thing. Phase-sync keeps the pair moving as one gesture instead of two arguments.
PAIRS WITH
OBSIDIAN MIX
The stereo outs want headroom and an honest drive stage after them. The batch sheet prints both: the knee per channel, and the headroom you own before it.
PAIRS WITH
TESSERA
Feed a HELIOS-8 stem into the buffer and scrub it against its own harmonic content. The grain module treats your analog voice like terrain, which is a compliment to both.
- Width
- 42 HP
- Depth
- 26 mm behind the panel
- Power
- +12V 340 mA / -12V 310 mA
- Price
- $1,290, identical at every dealer
- Included
- Ribbon, washers, signed batch certificate
- Firmware at ship
- 3.2.0, matched to the batch
- Warranty
- Five years, transferable
- Availability
- Batch 2026-10 open, 37 of 120 slots; next batch November 3