Access Official Technical PDF

Share Summary:LinkedInEmail

Free-space optical communication usually sends information by modulating light intensity and detecting the received signal. But when distance, channel loss or background illumination reduces the received light, the signal can become difficult to distinguish from noise. A time-correlation approach offers another option: rather than relying only on received intensity, it encodes data in the timing relationships between detected photons. A demonstration using a gallium nitride micro-LED and a single-photon avalanche diode (SPAD) showed data transmission at low received power while tolerating both steady and modulated background light.

Why use timing to detect a weak optical signal?

Conventional optical wireless communication (OWC) systems directly detect changes in optical intensity. At very low light levels, background photons and detector noise can obscure those changes. Some sensitive communication methods use more complex equipment, such as balanced photodetectors, local oscillators or optical phase-locked loops. Photon-counting receivers provide a different route to sensitivity, but many such systems depend on precise timing alignment between transmitter and receiver.

Time-correlation encoding changes what the receiver looks for. The transmitter sends short optical pulses in a repeated pattern, and information is represented by the presence or absence of a particular time relationship between pulses. A SPAD records individual detection events, including events caused by the signal, ambient light and detector dark counts. The receiver then evaluates the time separation between events to identify a correlation associated with the transmitted pattern.

How free-space optical communication uses pulse correlations

In the demonstrated on-off keying (OOK) scheme, a binary “1” is represented by repeated pairs of short pulses separated by a fixed interval. A “0” is sent without those pulses. When the receiver calculates an autocorrelation of the photon detection events, the repeated pulse separation produces a peak at the corresponding delay. If the expected peak is absent, the receiver interprets the symbol as a “0.”

Random background and dark-count events can also produce correlations, but they are less likely to repeatedly create a peak at the selected delay. The receiver builds a histogram over multiple repetitions of the pulse pattern and applies a threshold to decide whether the signal peak is present. Repetition helps distinguish the signal-related correlation counts from random counts, although it also limits the achievable data rate.

The pulse interval must also account for the SPAD’s dead time: after registering a photon, the detector needs time to recover before it can register another. In the reported single-detector setup, the pulse separation was chosen to exceed the detector dead time. A SPAD array could help address this constraint, but the demonstration considered a single SPAD.

Demonstrated performance and noise tolerance

The experimental transmitter used a 405 nm GaN micro-LED, with 5 ns optical pulses, and the receiver used a silicon SPAD. In the reported measurements, the system achieved a data rate of 100 kb/s at 8.25 pW of received optical power, with a bit error ratio (BER) below 10-3. The reported operating point corresponded to 27 detected photons per bit. At a lower received power of 2 pW, a 10 kb/s data rate was achievable at the same BER target.

These results illustrate the central trade-off: reducing received power requires more pulse-pattern repetitions to make the correlation peak distinguishable, which lowers the data rate. Performance depends on the signal detection rate, the number of repetitions and the background-generated correlation counts. For this reason, conventional signal-to-noise ratio alone does not fully describe how readily the receiver can separate signal and noise. The overlap between their Poissonian count distributions provides a more relevant measure of distinguishability.

The tests also examined background illumination. With high constant background, the system could still operate with signal power below the background power, although more signal power was needed as background counts increased. Modulated background had limited additional effect at modulation rates well above the link’s data rate. When background modulation was close to the link data rate, decoding became more difficult because the noise correlations varied between bit periods.

Engineering trade-offs and implementation

The method uses relatively straightforward optical components: a pulsed LED source, a single-photon detector and electronics to process detection events. In the initial characterization, an oscilloscope and computer handled data collection and offline decoding. A separate real-time implementation used FPGAs for transmission, clock recovery and decoding. That setup demonstrated a 20 kb/s link, with clock information included in the transmitted frames and therefore reducing the share of the link available for payload data.

The approach is not a drop-in replacement for every optical link. Its modest data rates reflect the need to collect enough repetitions to identify correlations, along with the SPAD dead-time limit. The reported implementation used OOK and did not include forward error correction. The authors identify pulse-position modulation and error-correction coding as possible ways to improve photon efficiency, while a SPAD array could help address detector dead time. These are potential development paths, not results established by the reported single-SPAD tests.

Potential applications and practical implications

Low received power, compact semiconductor components and tolerance to background light make this architecture relevant to links where size, power consumption or optical losses are important design constraints. A satellite simulator demonstration used separate test-bed systems to power the transmitter and receiver, showing operation in a representative nanosatellite development environment. The experiment establishes a laboratory and simulator proof of concept; it does not by itself establish performance for an operational satellite link.

For engineers assessing the technique, the key design question is whether the application can exchange higher data rate for lower received optical power. Link planning must account for pulse timing, detector dead time, photon detection probability, background count rate and the processing needed to recover symbol and frame timing. Optical filtering can help reduce background counts, while the timing-based detection method can provide an additional way to distinguish the intended signal from uncorrelated illumination.

Conclusion

Time-correlation encoding shifts photon-counting communication from measuring intensity changes alone to identifying repeated timing relationships in detected light. The demonstrated micro-LED and SPAD link transmitted data at pico-watt received power and operated under substantial background illumination, while making clear the associated trade-off between sensitivity and data rate. For industrial and other technical decision-makers, the approach is most relevant where low-light operation and modest data rates matter more than high throughput, and where the detector, timing and background conditions can be engineered as part of the link.

Access Official Technical PDF

About The Author