The True Genetics of the Behringer DeepMind
Or: How a Forgotten Italian "Calculator" Gave Birth to the Modern "People's Analog" Synth
Most musicians and bedroom producers religiously believe in a beautiful marketing fairy tale: “The Behringer DeepMind is a modern clone of the legendary Japanese Roland Juno-106.” And this notion — painfully obvious in its stupidity — travels from review to review, hardening into an absolute truth. People look at the abundance of faders, see the familiar subtractive synthesis layout, and parrot a “Brand of Origin” mantra that has absolutely nothing to do with reality.
The truth is, ergonomics can be incredibly deceptive. The control panel layout of most classic polyphonic synthesizers is more or less identical. So to truly look under the hood of the DeepMind, you have to arm yourself with a mountain of technical documentation — both current and historical. And then the veil of myth drops: the DeepMind has ab-so-lute-ly NOTH-ING in common with the Roland Juno!
In fact, inside the body of the Behringer DeepMind, beneath that Italian temperament, beats a cold, calculated American engineering heart!
This conclusion, reached after a detailed investigation, was preceded by a chance discovery on the internet (yeah… I’m writing “StarLord” on purpose ;)): an ultra-rare keyboard synthesizer from 1986 called the Elka EK-22, which surprised me, puzzled me, and left me desperately trying to remember: “What modern machine does this thing remind me of?!”
[Forgive me, brothers, but the absurd copyright chaos surrounding photos has finally done me in, so there will be no images here or below. If you’re curious — seek and you shall find on the web.]
However, I didn’t have to rack my brain for long, because around the same time, on July 14, 2023, shortly before my discovery, I recorded a video titled “What makes the Behringer DeepMind unique and why I will never sell it”
{ https://rutube.ru/video/31fd3f389738338c8beeb2bbf79b7859/ }
[I have a feeling that after reading this, you too — if you own a DeepMind — will say to yourself: “I’ll never sell it!” And if you don’t own one, you’ll definitely want to buy a DeepMind.]
…so anyway. At that point I was deeply immersed in the topic. I didn’t even need to double-check the specifications of the two synthesizers — one from 1986, the other from 2016. They were like twins! Struck by the realization of who actually defined the architecture of one of the biggest true-analog bestsellers of our time, I yelled “Eureka!” — because I had never believed the DeepMind had anything in common with Roland synths. If anything, I was more willing to believe the DeepMind was a ground-up design by Behringer’s own engineers. Their creation was just too drastically different from 99% of the analog synth hardware of the past and present.
Today I know that the fate of the Elka EK-22 was tragic. Sometimes it feels like this amazing instrument carries a curse that haunts anyone who dares touch its history. That curse even touched me…
I immediately shared my amazing discovery with a friend in a well-known messenger app. We used to discuss all things electronic music on those digital, unstable-as-water pages, but on July 28, 2023, when I opened our chat, I found absolute emptiness and purity!
“Sasha, what did you do?! Did you delete our entire chat history?”
“Yeah. My phone was lagging, so I cleared everything out.”
“For God’s sake!.. You should have deleted it on your end! Why did you wipe it from mine???”
Anyway, I won’t reproduce the rest of my screaming and mourning over the loss of this secret knowledge, which I had only recently acquired and — lo and behold! — completely forgot!..
Shortly after, Sasha vanished into the sunset, and I don’t even know if he’s alive. And I never did remember what that synthesizer was — the one that inspired Uli Behringer, and the one we all owe for the existence of the gorgeous DeepMind.
Part 1. Anatomy of a Deception: The Juno Fairy Tale vs. Silicon Reality
To understand why the DeepMind is technically unrelated to Roland, we must examine their fundamental engineering differences.
As you can see, the similarity between the DeepMind and the Juno begins and ends with the physical look of the sliders. The inner voice, the envelope mathematics, and the synthesis logic of the DeepMind were born from a completely different technological branch — the Curtis CEM3396 chips, used in American Oberheim synths, but finding their most radical expression in Italy.
[I’ve isolated a detailed breakdown of these chips into an appendix at the very end of this text. If you’re curious — step right up.]
When the engineers at Behringer’s British division (the company Midas) designed the DeepMind, they adopted a “synth-on-a-chip” concept embodied in the Coolaudio V3396 integrated circuit — a 100% modern clone of the legendary vintage Curtis CEM3396 crystal. Inside this single silicon chip live two oscillators, a filter, and four analog amplifiers.
[Feeling where this is going?]
The exact same Curtis chips (or their close relatives) were utilized in the Oberheim Matrix-6/1000 rack module and the ultra-rare British Cheetah MS6. But only one full, five-octave keyboard monster fully unleashed the potential of this architecture in a 6-voice format. And its name is the Elka EK-22.
Part 2. The Italian “Calculator”: Elka EK-22
In 1986, the Italian company Elka was world-famous for its elite flagship synthesizer, the Elka Synthex (Jean-Michel Jarre’s weapon of choice). The EK-22 was envisioned as its younger, more affordable, and mass-market 6-voice sibling.
The Italian engineers designed a magnificent, dense analog engine using six Curtis CEM3396 chips, but made a fatal marketing blunder. They tried to mimic the frantic trend of the era — the “digital sheen” of the Yamaha DX7.
You could understand them: musicians in the mid-80s were mass-dumping analog knobs, demanding a “clean, sterile design.” So Elka buried its rich analog sound behind an interface that can only be described today as “calculator ergonomics.”
The EK-22 Panel Layout
• The Text Cheat Sheet. Nearly half of the synth’s front panel was occupied by a massive, silkscreened table listing parameters and their two-digit numbers. Without this “on-chassis manual,” programming a sound was physically impossible.
• Blind Editing. The panel featured just a single Data Entry slider. To alter a basic filter cutoff, a musician had to:
- Find the filter parameter number in the printed table.
- Input that number using rubber digital buttons.
- Move the solitary slider while staring at a microscopic two-line LCD screen displaying lifeless digits from 0 to 99.
It was blind, unmusical programming. A stunning analog engine was locked away in a digital crypt.
Part 3. Genetic Matching: From EK-22 to DeepMind
Thirty years after Elka’s release, in 2016, Uli Behringer engineered a reverse revolution. His team took that exact hidden, complex architecture of the Italian synth and pulled all its parameters out into the open, creating the DeepMind.
If you map their signal paths down to the smallest detail, the genetic relationship becomes undeniable.
1. DCO Architecture and Asymmetry
Both the Elka EK-22 and the baseline DeepMind 6 feature exactly 6 true analog voices (the larger DeepMind 12 simply doubles this voice board). Each voice utilizes 2 DCOs, distributed according to a strictly asymmetric Italian logic:
• DCO 1 (Primary): Features full functionality and generates classic waveforms (Saw / Pulse).
• DCO 2 (Auxiliary): Generates only a square wave (Square/Tone); it lacks a sawtooth. Its sole technical purpose in both synthesizers is Hard Sync with the first oscillator to tear out screaming, piercing lead sounds.
2. Independent PWM Modulation
In most vintage circuits, pulse width modulation (PWM) is strictly tied to a single shared LFO. But the Curtis CEM3396 chip allowed for more flexibility. In the Elka EK-22, each oscillator possessed its own dedicated digital PWM circuit. Behringer copied this feature completely: in the DeepMind, you can modulate the pulse width of the first oscillator independently of the LFO, creating incredibly complex phase shifts. This is exactly why a 6-voice instrument doesn’t sound small; it sounds like a massive, living wall of stereo sound.
3. A Mathematical Marvel: ADBDSR Envelopes
While 95% of synth on the market rely on the standard four-stage ADSR (Attack, Decay, Sustain, Release) scheme, the Elka EK-22 boasted rare, six-stage ADBDSR envelope generators:
Attack → Decay 1 → Breakpoint → Decay 2 → Sustain → Release
Behringer’s engineers brought this mathematical luxury over to the DeepMind down to the very last letter. This structure allows you to program incredibly detailed sonic motion. Upon a key press, the filter can violently snap open (Attack), back off slightly (Decay 1) to a specific intermediate brightness threshold (Breakpoint), instantly change its trajectory to smoothly descend even lower (Decay 2), and only then settle into a soft Sustain. For endless cosmic pads, this logic is irreplaceable.
4. Filter Character (VCF) and the High-Pass Stage (HPF)
The low-pass filters of both instruments share a truly European, velvety, and “chocolate” Curtis character with a smooth bottom end, distinct from aggressively resonating Japanese designs. Cranking the resonance to its maximum sends the filter into pure self-oscillation (a clean sine wave), allowing you to play it across keys like a third oscillator.
At the same time, a high-pass filter (HPF) found its way into the DeepMind’s signal path. This is a completely analog stage that works both ways. Slid upward from the center position, it gently rolls off low-end rumble to sit the sound cleanly in a mix. But if you activate the LF Boost function, it mutates into an analog bass booster, generating a massive hump around 55 Hz and compensating for any loss of low-end density inherited from vintage circuits.
Part 4. Cosmic Chorus: The Great Digital Hybridization
A chorus effect is what transforms a dry analog signal into infinite space. And here, two eras collided in a beautiful engineering paradox.
• In 1986, the Elka EK-22 featured a fully analog chorus built on BBD (Bucket Brigade Device) delay lines. The signal physically moved through capacitors like water passing down a chain of buckets. It was noisy and slightly blurred the sound, but it created an unmistakable, warm stereo image. It was hard-wired directly into the master output — either on or off.
• In 2016, Behringer used a clever shortcut. Putting old, noisy BBD chips into a mass-market instrument wasn’t practical. By that time, Behringer’s parent company (Music Tribe) had already swallowed legendary brands like Midas (mixing consoles) and TC Electronic (the gods of spatial processing).
Midas engineers took the pristine digital algorithms of iconic TC Electronic analog pedals (including highly accurate digital emulations of those exact BBD chips) and integrated a powerful 4-slot effects processor deep inside the DeepMind. The signal inside the synthesizer remains 100% analog but passes through 24-bit converters right at the output to be processed by the DSP. This alone would blow any electronic musician’s mind, but Behringer clearly loved the DeepMind with a special passion, allowing the FX engine parameters to be controlled via the modulation matrix.
Now, the rate or depth of that “Italian” chorus inside the DeepMind can be modulated by those exact 6-stage ADBDSR envelopes or keyboard aftertouch.
Part 5. A Double “Calculator”: Elka EK-22
All this staggering math of six-stage ADBDSR envelopes, independent PWM modulation, and hard-synced asymmetric DCOs in the Elka EK-22 would be physically impossible if the voice allocation of the analog signal path wasn’t governed by a single silicon “brain.” Here we run straight into the irony of fate that tied the musical analog of yesteryear tightly to the computing world.
While Yamaha loaded its DX7 with proprietary Japanese chips and wasn’t above using a purely Western Motorola 6803 (formally, of course — the Yamaha HD6303XP, an enhanced CMOS version of the Motorola 6803 from Hitachi), Elka’s engineers stuck to a proven 8-bit classic. The central processing unit of the EK-22 became the legendary Zilog Z80 microprocessor, created by the brilliant Federico Faggin right after his dramatic departure from Intel.
The Z80 was the true gray eminence of the 1980s music industry. It didn’t make sounds itself, but its architecture managed the auto-tuning of temperamental American Curtis chips in the Prophet-5, Memorymoog, Oberheim Matrix, and, of course, our Elka EK-22. The Italians literally forced an 8-bit CPU to calculate bizarre sonic trajectories in real-time, making the synthesizer a “calculator” twice over — both in its horrific front-panel ergonomics and its silicon gut.
The Finale: Tragedy and Triumph. A Paradox of Two Eras
The trajectories of the Elka EK-22 and the Behringer DeepMind look like mirror reflections separated by thirty years of silence.
The Elka EK-22 became the embodiment of the tragedy of “late analog.” It arrived at the wrong time and in the wrong place. And worst of all — in the wrong form. In 1986, the world was dreaming of digital, and a noble, difficult-to-control analog instrument found itself unwanted. It left no deep footprint in global pop music and birthed no iconic stadium anthems. You won’t find “Top 10 Cult EK-22 Riffs” videos on YouTube. Its technical aesthetics were secondary to the root cause — the Curtis chips — and the nightmare of its “calculator” interface buried the model completely. The instrument became a beautiful but dead-end branch of evolution, forgotten by everyone except a handful of collectors.
The Behringer DeepMind emerged in the era of postmodernism, long after the age of legendary bands and monumental analog discoveries had passed. Today, music has largely been devalued — just as classic film photography was once devalued by digital. It happened first under the weight of DAWs and endless plugins, and now AI neural networks generate unique musical streams on the fly to match a listener’s personal taste in a single click. In this new reality, a complex, professional analog machine shouldn’t have stood a chance.
Yet, the DeepMind achieved a miracle of a different order. It became the analog equivalent of a vintage film camera like the Soviet “Zenit.” The Zenit didn’t capture the world’s absolute photographic masterpieces and wasn’t utilized by top National Geographic photographers. Over its lifespan, it did something greater — it became the storyteller for millions of ordinary families. It took a complex, elitist art form and made it accessible to everyone.
Thanks to its unprecedented affordability, the DeepMind brought a genuine slice of musical magic to ordinary people and hobbyists. It unlocked the door to the cosmos of “grown-up” analog synthesis for hundreds of thousands of bedroom producers, indie musicians, and electronic sound enthusiasts who could never dream of budgets for vintage juggernauts. The DeepMind essentially handed the keys of a personal spaceship to everyday creators, letting them fly toward the stars simply by turning a wheel… scratch that — simply by moving physical sliders.
Messenger chats get wiped out due to sheer stupidity, and marketing fairy tales keep blasting from every speaker. But cold circuitry guards the truth: great engineering ideas never die. Every single time someone dials up another deep, hovering pad on a DeepMind, the velvety, proud voice of the unjustly forgotten Italian Elka EK-22 silently comes alive in the air.
APPENDIX 1. A Bit of Silicon Anatomy: Curtis Chips vs. Roland IR3109
To finally strip away the myths surrounding the architecture of the Behringer DeepMind and the Elka EK-22, we must descend to the level of semiconductor crystals. Understanding how and why these components were engineered closes many debates about their sound.
The Birth of the “Synth-on-a-Chip” Concept
In the first half of the 1980s, manufacturers of analog polyphonic synthesizers hit a brutal crisis. Musicians demanded polyphony, but classic discrete circuitry was insanely expensive and fragile. Every single one of the 6 or 8 voices required a massive, separate board with dozens of transistors, capacitors, and resistors. These instruments cost as much as a car, went out of tune from a slight breeze, and required regular calibration by a technician wielding an oscilloscope.
Then the undeniably brilliant American engineer Doug Curtis and his California-based company, Curtis Electromusic Specialties (CEM), changed everything. They started shrinking entire functional synthesizer modules into monolithic integrated circuits.
[Just like the pragmatic architect Ted Hoff of Intel, who proposed implementing a universal processor on a single chip instead of — almost — a dozen large-scale ICs required by the Busicom project, so that all calculator algorithms could then be implemented in software.
We’re having a lot of calculator parallels today, aren’t we?
And that’s nothing!.. We haven’t even mentioned the legendary Italian Federico Faggin yet!.. Anyway…]
At first, these were individual chips: the CEM3340 (oscillator) or CEM3320 (filter) — the building blocks of the legendary Sequential Prophet-5 and Oberheim OB-8. But by the mid-80s, Curtis pushed further, creating the “Synth-on-a-Chip” family — the CEM3396 / CEM3394 series.
Architecture of the CEM3396: Inside the Silicon Labyrinth
The CEM3396 chip is a fully self-contained analog monosynth shrunk down to the size of two human fingertips. To build a 6-voice instrument (like the Elka EK-22 or DeepMind 6), a factory technician simply needed to mount and solder six of these chips in parallel on the mainboard.
Inside a single CEM3396 casing physically live:
• Two waveform generators (DCO 1 and DCO 2).
• An analog four-pole low-pass filter (VCF) with voltage-controlled resonance.
• Four independent voltage-controlled amplifiers (VCAs) that route signals within the chip and balance the mix.
How It Works Technically
• Where is the “digital”? The oscillator cores inside the chip are absolutely, 100% analog. They shape waves through the physical charge and discharge of capacitors. However, the frequency of this cycle (the pitch of the note) is controlled not by unpredictable analog voltage, but by ultra-precise digital pulses from the synth’s CPU (Master Clock). The processor registers a key press and shoots microsecond digital “clicks” of the right frequency into the Curtis chip, forcing the analog core to fluctuate electrical voltage (current) with absolute accuracy. This is the origin of the acronym DCO (Digitally Controlled Oscillator).
• The Secret to the “Fatness”: Internal Analog Switches. Because Curtis crammed everything into a single package, the distances between components on the silicon crystal are measured in microns. Signals within the chip are re-routed using built-in analog switches (electronic gates) driven by a digital interface. They physically reroute analog current from the oscillator to the filter or instantly engage the Hard Sync circuit. In discrete designs, this would require meters of circuit board traces and dozens of operational amplifiers.
What Makes the Curtis Sound Unique? (The Pros and Cons of Chip Physics)
Cramming an entire synth into a single IC gave the Elka and DeepMind an incredible sonic footprint, but it introduced distinct physical limitations:
• Perfect Phase Consistency: Because both oscillators of a voice share the same crystal, they warm up identically and simultaneously. When you engage Hard Sync or cross-phase modulation, the phases never drift over time, resulting in a surgically precise, solid, punchy sound.
• Waveform Specifics: To save real estate on the crystal, Curtis deployed a clever trick: the first oscillator generates a classic sawtooth (Saw) and pulse (Pulse), while the second produces only a square wave (Square). It didn’t need to be full-featured; its job was to act as a modulator or add sub-harmonics. This is why the second oscillator on the DeepMind and Elka has stripped-back shapes — it’s a direct consequence of the Curtis crystal topology.
• Smooth “European” Filter: The Curtis filter in the CEM3396 relies on a cascade of transistor pairs designed to keep the sound round and singing. It exhibits phenomenal musicality in polyphonic pads and string textures. At high resonance, it doesn’t aggressively gut the low frequencies, preserving the body of your sound. This vintage geometry of American Curtis chips was beautifully unlocked by Italian engineers in the Elka EK-22, and inherited 30 years later by the DeepMind.
The Japanese Counterpart: The Roland IR3109 Enigma
To fully understand why the DeepMind is technically detached from the Roland Juno, we must look at the opposing camp. Behringer’s marketing team targets the Juno-106 simply because it’s a household name. However, the Japanese machine tells a completely different silicon story: it runs on the custom Japanese Roland IR3109 chip.
The Roland IR3109 is a custom analog integrated circuit engineered at the turn of the 70s and 80s specifically for Roland filters (powering the Jupiter-8, Juno-6, Juno-60, and early Juno-106).
Inside the IR3109 chip, a single silicon wafer combines four independent Operational Transconductance Amplifiers (OTAs).
How the Roland IR3109 Filter Worked
These four OTA stages inside the IC were connected sequentially, creating the famous 4-pole (24 dB/oct) low-pass filter. The quirk of the IR3109 architecture was that as you pushed the resonance (feedback), the filter screamed, spitting out a bright, aggressive, and sometimes plastic Japanese color. The physics of the circuit meant that high resonance completely castrated the low frequencies, rendering the sound thin but razor-sharp — perfect for 80s pop hits.
The Curtis CEM3396 chips (in the EK-22 and DeepMind) behave fundamentally differently: their transistor cascades smooth out the resonance peak, locking in a thick, meaty low end.
Modern Epilogue: How Behringer Resurrected Dead Silicon
When Behringer set out to revive this architecture, original Curtis chips were long out of production. Uli Behringer tasked his subsidiary, Coolaudio, which specializes in semiconductor reverse-engineering.
They took the original topological blueprints of the Curtis CEM3396 and remanufactured them on modern semiconductor assembly lines under the name Coolaudio V3396. The architecture remained untouched: the same internal logic, the same analog DCO cores, the same analog routing gates, and the same velvet filter.
The only difference between the modern clones and the original 80s chips is higher-grade silicon purification and modern lithographic processes. Because of this, the chips inside the DeepMind run with greater stability, require almost zero warm-up time, and exhibit a lower noise floor, delivering authentic analog history straight to our ears in its purest form.
What Would These Geniuses Have Played Without Simple Piano Tuners?
To fully grasp the scale of Federico Faggin’s microprocessor architecture’s expansion into the musical world, we need to look at the service manuals of the analog renaissance era. While Doug Curtis’s chips (CEM) and custom Roland assemblies (IR3109) handled the physical shaping of waveforms and filtering, the Zilog Z80 did all the dirty digital work. It scanned the keyboard, tracked fader movements, allocated voices in polyphonic mode, and hardware-set a flawlessly precise frequency for the DCO, completely eliminating any possibility of temperature drift.
But the “calculator” past that Faggin tried to escape even in the early stages of designing the first large-scale ICs caught up with his brainchild forty years later in the most cynical form.
While today’s bedroom producers twist physical faders on the modern Behringer DeepMind, completely unaware that under the hood of the software hides a digital copy of the Italian engine originally designed for the Z80, the original processor… continues to take advanced math exams in the pockets of American students!
Right now, Texas Instruments is mass-producing its ultra-current TI-84 Plus CE graphing calculators. And in their heart, stuffed to the brim with matrices, function graphs, and…
APPENDIX 2. PATENT REGISTRY AND BIBLIOGRAPHY
I. Integrated Circuit Patents (Curtis, Roland, Intel)
- Curtis, D. R. (Curtis Electromusic Specialties). Integrated Sound Generation Circuit for Musical Instruments — U.S. Patent No. 4,509,398. Issued: April 9, 1985. (The historic Doug Curtis patent on the architecture of the analog microprocessor series CEM3396 “Synth-on-a-Chip.”)
- Curtis, D. R. (Curtis Electromusic Specialties). Electronically Controllable Voltage Filter with Wide Dynamic Range — U.S. Patent No. 4,250,453. Issued: February 10, 1981. (The foundational principles of Curtis’s transistor cascades that ensured the “European,” velvety character of the analog VCF.)
- Yamashita, T., Kawamoto, K. (Roland Corporation). Voltage Controlled Filter Circuit — Japan Patent No. S56-112310 / U.S. Patent No. 4,344,043. Issued: August 10, 1982. (The official patent on the architecture of the four-channel OTA crystal Roland IR3109, which defined the piercing and “screaming” Japanese character of Juno/Jupiter filtering.)
- Hoff, M. E., Mazor, S., Faggin, F. (Intel Corp.). Memory System for a Multi-Chip Digital Computer — U.S. Patent No. 3,821,715. Issued: June 28, 1974. (The patent on the architecture of the MCS-4 / Intel 4004 family, which gave Busicom a chance to escape from the dozen large-scale ICs — which, however, the Japanese never took advantage of. For which they soon paid the price.)
II. Technical Documentation and Primary Sources (Datasheets & Manuals)
- Curtis Electromusic Specialties. CEM3396 Dual Waveform Synthesizer-on-a-Chip. Technical Data Sheet. — San Jose, CA, 1985.
- Coolaudio Semiconductor. V3396: Low-Power Dual Waveform Processor. Specification. — Music Tribe IP Ltd, 2016.
- Elka S.p.A. ELKA EK-22 / EM-22 Polyphonic Synthesizer. Service Manual & Schematic Diagrams. — Castelfidardo, Ancona, Italy, 1986.
- Behringer / Midas. DeepMind 12 / DeepMind 6 Polyphonic Analog Synthesizer. User Manual & Firmware Architecture Notes. — Willich, Germany / Manchester, UK, 2016.
- Roland Corporation. Roland Juno-106 Programmable Polyphonic Synthesizer. Service Notes. — Osaka, Japan, 1984.
III. Historical and Analytical Materials
- Curtis, D. The Evolution of Analog IC Design for Electronic Music // Journal of the Audio Engineering Society. — 1983. — Vol. 31, No. 4.
- Milan, D. Vintage Synthesizers: Groundbreaking Instruments and Psychedelic Devices. — Miller Freeman Books, 1993. — Section: “The Italian Connection: Elka, Bit and the Curtis Revolution.”
- Faggin, F. The History of the 4004 // IEEE Micro. — 1996. — Vol. 16, No. 6. — Pp. 10-20. (Federico’s memoirs on how the Busicom calculator puzzle was assembled and the world’s first commercial microprocessor IC was created.)
deepmind
behringer
ek-22
синтезатор
roland
