The Enhanced Linear Regulator

Linear regulation, past its own limits

An Enhanced Linear Regulator is an inductorless, switching-free DC voltage regulator that co-packages a linear regulator (LDO) with patented photovoltaic-output optocouplers. The optocouplers recycle power normally dissipated in a conventional LDO, so the device keeps the low noise, small footprint, and simple bill of materials of a linear regulator while reaching efficiency approaching that of a switching regulator.

The problem

The voltage-regulator trade-off

Every electronic system needs DC voltage regulation, and for decades every regulator forced the same compromise between three things designers want at once.

System size

Small, or switching

Switching regulators need inductors, EMI filtering, and careful layout. Linear regulators are tiny and simple — an IC and a few passives.

Energy efficiency

Efficient, or linear

A conventional LDO burns the entire input-output voltage difference as heat. Its efficiency is capped at VOUT/VIN — the larger the step-down, the worse it gets.

Power integrity

Quiet, or efficient

Switching converters spray ripple and EMI across the band where RF front-ends, data converters, and precision sensors are trying to work. Linear regulators are quiet by construction.

Achieving all three simultaneously is what blocks power designers in noise-sensitive, SWaP-constrained systems. The usual workaround — a switching pre-regulator followed by an LDO — costs board area, components, and design time, and the switching noise still has to be filtered out.

How it works

Photons recycle the wasted energy

Inside an ELR, a silicon LDO is co-packaged with Polaris's proprietary photovoltaic-output optocouplers. The LDO regulates exactly as it always has — same feedback loop, same clean output. But instead of burning the dropout energy as heat, the ELR routes current through the optocoupler's LEDs, converts it to light, and converts the light back to electrical power with on-chip photovoltaic cells — power that is fed forward to the output.

Because nothing switches, there is no ripple, no EMI, and no inductor. The same mechanism also stacks voltage above the input rail, which is how Polaris built the world's first efficient switching-free boost regulator.

The terms, defined once

  • Switching-free — no part of the regulator switches; regulation is entirely linear, so the output carries no switching ripple or EMI.
  • PV-output optocoupler — an optocoupler whose output side is a photovoltaic cell that delivers usable electrical power, rather than a phototransistor that merely signals.
  • Polaris Monolithic Optocoupler (MLOC) — a single monolithic GaAs chip integrating both the LED and the photovoltaic cell. The breakthrough section below looks inside it.
  • Optocoupler-enhanced mode — operation with the optocouplers active. Every Polaris device can also run as a conventional LDO by shorting two pins ("LDO mode").
Block diagram of an Enhanced Linear Regulator: an LDO chip with pass transistor and feedback circuit, co-packaged with optocoupler chips whose LEDs and photovoltaic cells route recycled power to the output.
The buck ELR topology — the LDO keeps its feedback loop; the optocouplers recycle the dropout energy.
The breakthrough

The optocoupler, reinvented as a power device

Optocouplers have relayed signals for half a century — a discrete LED facing a discrete photodetector across a gap. Polaris re-engineered the optocoupler to move power, then patented both the chip and the regulator circuits built around it.

In a conventional optocoupler the emitter and receiver are separate dies, assembled and aligned in a package and coupled across an optical gap — built to pass a signal, not to deliver power. The MLOC eliminates the gap: photovoltaic cell, semiconductor blocking diodes, and LED are grown together as a single GaAs crystal stack, with an integrated mirror that folds the LED's upward emission back down into the PV cell. The blocking diodes pass light but block current, isolating the PV cell from the LED at the tens-of-volts level a power-management device needs — this is a power-transfer chip, not a kilovolt safety isolator. No die-to-die interface, no alignment loss: the dominant loss paths of a discrete pair are simply gone.

The payoff is LED-to-PV power-conversion efficiency above 52% — more than 25 times the percent-or-two a conventional optocoupler can move — with an unamplified current transfer ratio above 72% per LED/PV pair. That margin turns "recycle the dropout energy as light" from a laboratory curiosity into a regulator typically 1.2× to more than 2× as efficient as a conventional LDO, with the largest gains on large step-downs, and it is what makes efficient switching-free boost possible at all. The chips are built on the same mature, high-volume III-V production lines as commodity high-brightness LEDs — no exotic processes, no precision optical assembly.

Cross-section diagram of the Polaris Monolithic Optocoupler: a single chip stack with a mirror on top, an LED region beneath it, a blocking-diode region below that, and a photovoltaic region at the bottom. Arrows show light from the LED reaching the PV region directly and by reflecting off the top mirror.
Notional cross-section of the MLOC: light from the LED reaches the photovoltaic cell directly or via the top mirror; the blocking-diode region passes light but blocks current.
>52%
LED-to-PV power-conversion efficiency
>72%
Current transfer ratio per LED/PV pair
>25×
Power transfer vs. a conventional optocoupler

The optocoupler is the enabling chip — but a chip alone is not a regulator. Three granted U.S. patents cover the invention at every layer: the monolithic device, the buck topology that recycles dropout energy, and the founding patent on photon-coupled regulation itself.

The monolithic optocoupler

The device patent: LED, blocking diodes, and photovoltaic cell grown as one monolithic GaAs stack, with an integrated mirror that redirects emission into the PV cell — the structure that makes optocoupler power transfer efficient.

The efficient buck

The step-down topology: the LED string shares the input-to-output voltage drop with the regulator's pass device, and the photovoltaic section feeds the recovered current forward to the output — the circuit at work in every buck ELR.

The founding patent

The 2018 priority filing that started it all: a linear regulator that moves power on photons, covering efficient step-down of large voltage drops and — by stacking photovoltaic voltage above the input rail — DC up-conversion with no switching, the basis of every switching-free boost ELR.

Measured, not promised

What it means in practice

Every number below is published — in the white paper, an application note, or a product document.

Efficiency

60–85% buck conversion

Buck ELRs operate at 1.2× to >2× conventional-LDO efficiency depending on the voltage step; the BT29 boosts at >65% peak with zero switching.

Noise

<1 µV RMS configurations

The BK30 series reaches sub-microvolt RMS output noise (10 Hz–100 kHz) with PSRR above 120 dB near 1 kHz — linear-regulator power integrity at far higher efficiency.

Radiation

Radiation-tolerant: 200 krad(Si)

All four families are rated radiation-tolerant to 200 krad(Si), and the BK19 is built on a QML-V radiation-hardness-assured LDO core. Behind the rating is a full characterization campaign: cobalt-60 TID with published output-voltage shifts under 1.5%, destructive single-event-effects testing, and proton, electron, and neutron irradiation, plus MIL-STD-883-style life and thermal-cycle screening — reports available to customers.

Thermal camera comparison of two regulator boards converting 10 V to 3.3 V at 180 mA: the conventional linear regulator glows at 82.2 °C while the Polaris regulator runs at 48.7 °C.
Recycled energy you can see: 10 V → 3.3 V at 180 mA. Conventional linear regulator: 82.2 °C. Polaris ELR: 48.7 °C.

Validated on orbit

Polaris ELRs launched to the International Space Station in October 2025 aboard JAXA's HTV-X1 on the Aegis Aerospace MISSE-21 platform. The flight-heritage experiment has been operating continuously since January 8, 2026 — thousands of hours logged and counting — with devices performing as expected against a ground-control twin.

Built on trusted foundations

Proven LDO cores, enhanced

Polaris doesn't ask you to bet on an unproven regulator core. Each ELR family is built around an industry-standard LDO that design teams already qualify and trust — the QML-V radiation-hardness-assured RH1965 (BK19), the high-current MIC29152 (BK29, BT29), and the ultra-low-noise LT3045 (BK30). The optocouplers enhance the core's efficiency; its regulation behavior, protection features, and noise pedigree carry straight through.

The technology is documented in a technical white paper developed with The Ohio State University and supported by Texas Instruments, whose team supplied the radiation-hardness-assured linear regulators behind the paper's rad-hard ELR demonstration. Our application notes carry the same proof-first approach into practice: complete, measured reference designs — from all-linear Li-ion charging to adaptive multi-source power — built from ELRs and readily available commercial parts.

Applications & markets

Where clean power is the bottleneck

Noise-sensitive DC power is the high-performance slice of the roughly $14B global DC-DC converter market — a $1.5B+ segment today, and the one where engineers accept a power-loss premium because nothing else meets the noise spec. The ELR enters where that pain is sharpest, and is expanding from there.

Deployed today

Space, defense & precision instrumentation

SmallSat and spacecraft power, RF and microwave front-ends, scientific and medical instrumentation — mission-critical systems where signal fidelity and SWaP are hard constraints, not preferences. The ELR is already flying on the ISS, with first commercial shipments in August 2025.

Why ELR wins: the noise spec without the efficiency penalty

Scaling next

High-speed data, automotive & 5G/6G

Data-center interconnects and retimed links, automotive LiDAR and radar, radio units, and industrial sensing: high-volume systems where SerDes jitter budgets, EMC-qualification risk, and board space leave no room for switching noise or a two-stage filter chain.

Why ELR wins: switching-class efficiency, zero EMC surprises

On the horizon

Quantum, photonics & chiplets

Quantum systems, free-space optical communications, and chiplet-based heterogeneous packages — where power increasingly must be delivered inside the package, with no room for an inductor and no tolerance for EMI. A monolithic optocoupler is a natural fit for chiplet-scale integration.

Why ELR wins: inductor-free power that integrates where switchers can't

Spotlight: the high-speed interconnect rail

Inside an optical module — or an active electrical cable doing the same retiming job in copper — the 3.3 V cage supply has to step down to the lower logic rails that feed clock-and-data recovery, retiming, and jitter attenuation — in millimeters of board space, against a hard thermal ceiling, for circuits whose jitter budgets are measured in femtoseconds. Today that means either burning the difference in an LDO or filtering a switcher out of the most noise-sensitive signal path in the system. An ELR does the same step-down at switching-class efficiency with no switching edge at all: less heat in the cage, no spurs near the eye, nothing to filter.

The multiplication is the point. An AI cluster runs on the order of six optical links per GPU — before counting the retimed copper (AECs) filling the shorter hops; a 72-GPU rack carries hundreds of them; a 100,000-GPU campus approaches six hundred thousand — and every link terminates in CDR silicon that needs quiet rails, at both ends. With data-center operators committing more than half a trillion dollars a year to new AI capacity, the cleanest rail in the rack is no longer a niche.

Notional scale of the CDR-rail opportunity in an AI data center One GPU drives roughly six high-speed links, optical or retimed copper; a 72-GPU rack carries hundreds; a 100,000-GPU campus approaches six hundred thousand — every link end with clock-and-data-recovery silicon that needs quiet power rails. 1 GPU × ~6 links per GPU (optical + retimed copper) — CDR at both ends × 72 GPUs one AI rack: hundreds of links × ~1,400 racks one 100,000-GPU campus: ~600,000 links Every dot: a CDR whose rails must stay quiet.
Notional scale: dot counts are illustrative; the arithmetic — roughly six optical links per GPU across the network tiers, before counting retimed copper — is in the text.
Where it goes next

Wherever power integrity limits performance

Our work in noise-sensitive power began in aerospace, defense, and precision instrumentation — where the ELR is flying today. The same physics, and the same proven-core philosophy, scale to the markets that come next.

Get started

Ready to design it in?

Request evaluation samples, detailed datasheets, and SPICE models — an engineer answers, not a sales funnel.