Oskars Ozolins

The rapid growth of AI-driven applications is pushing short-reach optical interconnects toward higher per-lane data rates, putting strain on modulator and electronic driver technologies. In this work, we investigate thin-film lithium niobate (TFLN) modulators under direct and enhanced signal-to-noise ratio (SNR) driving conditions using a multi-channel arbitrary waveform generator (AWG) and balun-based signal combining. We demonstrate up to 425 Gb/s (net 400 Gb/s) PAM4, 450 Gb/s (net 423 Gb/s) PAM6, and 474 Gb/s (net 446 Gb/s) PAM8 transmission over 500 m single-mode fiber (SMF) without optical amplification, all below the 6.25% HD-FEC threshold. Furthermore, we demonstrate 224 Gb/s OOK/PAM4 FEC-free operation over 6 km of SMF.

The 1-µm wavelength band is investigated for high-speed, short-reach intensity-modulation/direct-detection (IM/DD) transmission using a 1064 nm lithium-niobate Mach–Zehnder modulator (MZM). On–off keying (OOK), four-, six-, and eight-level pulse-amplitude modulation (PAM4, PAM6, and PAM8) are evaluated after 100 m of standard single-mode fiber (SMF) transmission and compared with optical back-to-back (ob2b) operation. Net rates of 178 Gb/s for OOK, 200 Gb/s for PAM4, and 211 Gb/s for PAM6 and PAM8 are achieved below the 6.25% overhead hard-decision forward-error-correction (HD-FEC) threshold. These results support the potential of 1064 nm externally modulated links for future low-latency short-reach interconnects and radio-access-oriented optical transport systems.

Burst-error-dominated channels are commonly encountered in photonics-assisted terahertz (THz) wireless communication systems due to high-speed optoelectronic device impairments and receiver-induced error propagation. Traditional defenses, including block codes and L-interleaving, offer negligible protection against such severe temporal correlations, whereas conventional turbo product codes (TPCs) suffer from component vulnerability. In this paper, by introducing a diagonalized codeword mapping, a burst-error-resilient TPC scheme is proposed, spatially dispersing consecutive errors without increasing computational complexity. Both simulation and experimental results demonstrate that the proposed diagonal TPC significantly truncates burst-error exceeding three symbols, offering a hardware-efficient architectural solution for future ultra-high-speed THz networks.

Integrated sensing and communications (ISAC) is a key enabling technology for 6G networks, as it unifies radar sensing and data transmission within a shared infrastructure. Analog radio-over-fiber (ARoF) has emerged as a crucial technology for extending radar coverage and enabling distributed radar sensing. By incorporating pulse position modulation (PPM), it is possible to simultaneously measure the target’s range and transmit data, thereby increasing the energy efficiency of such systems. In this research, we present a novel ARoF-based ISAC-capable pulse radar system that features a custom-built PPM modulator employing a digital-to-time converter, and a high-accuracy time tagger as the demodulator, allowing for pulse time of arrival measurements with an root mean square accuracy up to 1.5 ps. The experimental results show a 5 Mbit/s data transmission along with a bit error ratio less than 5.86 × 10⁻⁷ at a 95% confidence level, using a PPM slot duration of 750 ps in an optical system featuring more than 40 km of single-mode optical fiber and a simultaneous range measurement with error less than 0.86 cm. In addition to ISAC exploration, the impact of fiber-induced chromatic dispersion on the PPM signal is studied, and the system is validated in an outdoor environment with a moving target.

Photonic convolution is a cornerstone for broadband artificial intelligence hardware, yet existing implementations often suffer from low computational density, requiring numerous wavelength channels. Here, we propose a resonator-free convolutional architecture in the synthetic frequency dimension, marking a paradigm shift from isolated spectral transmission to coupled spectral computation. By programming the driving waveforms of an IQ modulator and synchronizing symbol durations with time delays, we achieve arbitrary, high-precision, and reconfigurable convolutional kernels without iterative optimization. This approach increases the multiply-accumulate (MAC) operations per symbol per wavelength from 1 to N² for an N × N kernel, enhancing computational density while reducing laser source requirements. In a proof-of-concept experiment, we demonstrate Gaussian kernel-based feature extraction of a barcode image with a root mean square error of 0.073, validating the feasibility and accuracy of our architecture. This work establishes a new physical paradigm for high-density photonic computing hardware.

Digital subcarrier multiplexing (DSCM) enables flexible spectrum allocation in photonics-assisted terahertz (THz) communication systems, offering a promising solution for next-generation multi-user wireless networks. However, transmitter-induced IQ impairments cause symmetric subcarrier crosstalk, degrading the demodulation performance. To address this issue, we propose a pilot-tone-based transmitter impairment monitoring algorithm for THz-DSCM systems. We experimentally validate the algorithm using a proof-of-concept 350 GHz photonic DSCM system with four subcarriers. The results demonstrate robust performance, with mean estimation errors within ±0.16 dB for amplitude imbalance, ±0.7° for phase imbalance, and ±0.19 ps for IQ skew.

The mid-infrared (MIR) free-space optical (FSO) transmission systems based on quantum cascade laser (QCL) have attracted substantial attention. Nevertheless, a major limitation persists at the receiver side, which remains dominated by mercury cadmium telluride (MCT) detectors due to the lack of competitive alternatives. This prevalent reliance is particularly problematic in the mid-wave infrared (MWIR) band, where MCT detectors are fundamentally constrained by limited modulation bandwidth and cooling requirements. This study explores the potential of MWIR FSO transmission that leverages an MWIR uni-traveling carrier photodiode (UTC-PD) to attain high-speed communication. We present an experimental demonstration of an MWIR FSO link at 4.65 μm, combining a directly-modulated (DM) QCL with an uncooled UTC-PD detector based on InAs/InAsSb type-II superlattices. A transmission data rate of up to 14 Gb/s over a 31-m optical path length (OPL) is successfully achieved, demonstrating superior performance compared to a benchmark commercial MCT. Our findings suggest that the UTC-PD-based MIR FSO receiver offers a promising solution towards the next generation of optical wireless communication networks.

Neural networks (NNs) have emerged as an effective equalization approach for high-speed intensity modulation and direct detection (IM/DD) optical links, where chromatic dispersion, limited bandwidth, and device-induced nonlinearities degrade signal quality. However, the heavy reliance on multiplications in conventional NNs leads to high computational complexity, limiting hardware deployment. Addition-based convolutional NN (AdderCNN) was originally developed for image classification, where it reduces complexity by replacing multiplications with addition-based operations. In this work, we propose a multiplier-free AdderCNN equalizer for up to 200 Gbaud on-off keying IM/DD links using a 26 GHz O-band ring resonator modulator. After 500 m single-mode fiber transmission, AdderCNN achieves bit error rate (BER) below the 7% overhead (OH) hard-decision forward error correction (HD-FEC) threshold of 3.8 ×10⁻³, outperforming conventional equalizers. Compared to classical CNN, AdderCNN eliminates all real multiplications (RM), reduces bit operations (BOP) by 87.5% and the number of adders and shifters (NABS) by 93.9% per equalized symbol under 32-bit full-precision with comparable performance. With 7-bit quantization, it further reduces both BOP and NABS by 71.7% per equalized symbol, while keeping BER below the 7% OH HD-FEC threshold. Therefore, AdderCNN has the potential to become a hardware-efficient NN equalization solution for next-generation optical interconnects.