How LOFIC Enables True HDR in Astra's CMOS Sensors

Dynamic range is one of the most misunderstood specifications in night vision. It is not simply about how dark an environment a device can operate in—it is about how well a sensor handles a scene that contains both very dark and very bright regions simultaneously. This is precisely the challenge that night vision operators face: a field that is mostly dark, with a vehicle headlight, campfire, or IR illuminator somewhere in frame.

Astra's large-format sensors address this with a specific pixel-level circuit technique called LOFIC—Lateral Overflow Integration Capacitor (not present in the X10, which uses a different HDR approach). Understanding LOFIC requires first understanding why conventional CMOS sensors struggle with HDR at all.

The Fundamental Tradeoff in CMOS Imaging

Every CMOS pixel collects photons and converts them to charge. That charge accumulates in a storage region called the photodiode well. The well has a maximum capacity—the full-well capacity (FWC), measured in electrons (e⁻). When the well fills, it saturates: additional photons produce no additional signal. The pixel "clips" to white.

The tension is this:

• High sensitivity requires low noise floor. To detect a single photon, read noise must be extremely low—ideally below 1 electron RMS. This demands careful, slow pixel readout circuitry.
• High dynamic range requires large full-well capacity. To handle bright light without saturation, the well must hold a lot of charge—often requiring larger pixel geometry or reduced sensitivity.

These two goals conflict at the pixel architecture level. Optimizing for one typically degrades the other.

For a night vision sensor, this tradeoff is particularly painful. You need extreme sensitivity (moonless nights, 0.00005 Lux environments) and the ability to handle sudden bright sources—IR illuminators, vehicle lights, muzzle flash—in the same operational context, often in the same scene.

Traditional HDR Approaches and Their Problems

Most consumer camera sensors use one of two approaches to extend dynamic range:

Multi-Exposure HDR

Capture two or more frames at different exposure times, then merge them. The long exposure captures shadow detail; the short exposure captures highlight detail without saturation. The combined image has greater apparent dynamic range.

Problem for night vision: Any movement between frames creates ghosting artifacts—a person walking appears as a blur or double image. In a surveillance or tactical context, subjects are almost always moving. Multi-exposure HDR produces unreliable, artifact-prone imagery in exactly the scenarios that matter most.

Dual-Conversion Gain

Switch the pixel's charge conversion factor between a high-gain mode (sensitive, low FWC) and low-gain mode (less sensitive, high FWC) on alternate frames or rows. Merge the two readouts.

Problem: Still requires two readout operations per pixel, adding latency and complexity. And the interleaved sampling creates fixed-pattern noise and row artifacts that degrade image quality.

LOFIC: Simultaneous Dual-Range Capture in a Single Pixel

Lateral Overflow Integration Capacitor (LOFIC) solves the problem at the pixel architecture level, without any multi-frame merging or temporal artifacts.

The Basic Architecture

A LOFIC pixel has two charge storage regions:

1. The photodiode: a small, high-sensitivity region that accumulates charge with extremely low noise. It has a limited FWC—optimized for low-light sensitivity, not high-light capacity.
2. The lateral overflow capacitor: a secondary capacitor, physically adjacent to the photodiode, connected by a carefully biased overflow barrier. It has a much larger charge capacity—typically 10× to 100× that of the photodiode.

During integration (exposure), the photodiode fills normally. When it reaches its overflow threshold, excess charge spills laterally into the LOFIC capacitor rather than into neighboring pixels (which would cause blooming) or being lost (which would clip highlights). The capacitor continues accumulating charge from the overflow throughout the rest of the exposure.

The Readout

At the end of the frame, both regions are read out independently in a single readout cycle:

• The photodiode provides a high-sensitivity, low-noise signal for dark regions of the scene.
• The LOFIC capacitor provides a high-capacity signal for pixels that received bright illumination.

The ISP then combines the two signals using a pixel-level decision: for pixels where the photodiode did not saturate, use the photodiode signal (maximum sensitivity). For pixels where the photodiode overflowed into the LOFIC, use the combined signal to reconstruct the full photon count.

The result is a single frame—with no temporal artifacts, no multi-exposure merging, and no motion ghosting—that contains both the faintest detectable signal from the photodiode and the full dynamic range of bright sources captured by the LOFIC capacitor.

Dynamic Range: By the Numbers

Dynamic range is typically expressed in decibels (dB) or stops:

Dynamic range (dB) = 20 × log₁₀(FWC / read_noise)

A conventional high-sensitivity CMOS pixel optimized for low-light might have:

• FWC: ~5,000 e⁻
• Read noise: ~2 e⁻ RMS
• Dynamic range: 68 dB (~11 stops)

With LOFIC, the effective FWC is extended by the lateral capacitor's capacity. A LOFIC capacitor can add 50,000–500,000 e⁻ of additional capacity, extending the effective dynamic range to:

• Effective FWC (photodiode + LOFIC): ~500,000 e⁻
• Read noise: ~2 e⁻ RMS (unchanged—the low-light photodiode still reads at full sensitivity)
• Dynamic range: 108 dB (~18 stops)

This is roughly 10× greater dynamic range than a standard CMOS sensor optimized for low-light use—and it comes with no sensitivity penalty in the dark regions of the scene.

Why This Matters Specifically for Night Vision

Consider a practical scenario: you are observing a dark field at night. A vehicle approaches from 800 meters. Its headlights are orders of magnitude brighter than the ambient starlight illuminating the field.

On a standard CMOS sensor, you face a binary choice:

• Expose for the dark field → the headlights saturate, bloom across surrounding pixels, obscure the vehicle's profile
• Reduce gain to handle the headlights → the dark field drops below the noise floor, invisible

On a LOFIC sensor, both regions are captured simultaneously in a single frame. The dark field retains full 0.00005 Lux sensitivity. The headlights are captured in the LOFIC capacitor—bright but not clipped, not blooming—allowing you to see the vehicle clearly while maintaining awareness of the surrounding environment.

This is not a subtle improvement. In operational use, LOFIC-enabled HDR is the difference between being blinded by a light source and maintaining full situational awareness through it.

LOFIC vs. Other HDR Methods in the Field

Method	Motion Artifacts	Dynamic Range	Latency Impact	Noise in Shadows
Multi-exposure HDR	Yes — ghosting on motion	High	+1 frame minimum	Low
Dual-conversion gain	Minimal	Moderate	Small	Moderate (row noise)
Tone mapping (software only)	None	No gain — only compression	None	Amplified
LOFIC	None — single frame	Very high (~108dB)	None	Low — photodiode preserved

Integration with Astra's LOFIC-Equipped Sensors

The LOFIC architecture in Astra's large-format CMOS sensors is complemented by several additional design choices that work together on LOFIC-enabled models:

• 18μm pixel pitch: Large pixels provide a large photodiode area (high FWC for the primary well) and sufficient physical space on-pixel for the LOFIC capacitor without shrinking the photodiode. Smaller pixels cannot accommodate LOFIC without sacrificing photodiode area.
• BSI construction: Back-side illumination ensures the LOFIC capacitor and associated circuitry sit below the silicon without blocking incoming photons from reaching the photodiode.
• F1.05 objective lens: A very fast lens maximizes photon flux onto the sensor, letting the photodiode operate well above its noise floor even in extreme low light—preserving the signal-to-noise ratio that makes the LOFIC dual-range readout meaningful.
• Advanced ISP: The onboard image signal processor handles the LOFIC pixel merging, tone mapping, and noise reduction in real time to deliver the final image with sub-10ms latency at 100fps.

The Result: 0.00005 Lux, With Headlights in Frame

Astra's LOFIC-equipped sensors achieve a minimum illumination of 0.00005 Lux—the sensitivity of the photodiode, operating at the limit of passive starlight imaging. At the same time, they handle IR illuminators, vehicle lights, and other bright sources in the same scene without saturation artifacts.

This combination—extreme low-light sensitivity and high dynamic range—is precisely what LOFIC was designed to deliver. It is not achievable through any purely software approach, nor through conventional single-well CMOS pixels, nor through multi-exposure HDR methods that introduce motion artifacts.

For night vision specifically, where the environment is simultaneously dark and unpredictably lit, LOFIC is not an optional upgrade. It is a foundational requirement for dependable operational performance.