Frequency-Temporal Enhanced Recovery for Ultra-Low-Resolution ACO-OFDM OWC Links

We propose and experimentally demonstrate a novel frequency–temporal enhanced recovery (FTER) framework to mitigate quantization distortion in ultra-low-resolution (3-5-bit) asymmetrically clipped optical OFDM (ACO-OFDM) optical wireless communication (OWC) links. The FTER scheme combines frequency-domain feature extraction with a temporal convolutional network (TCN)-based denoiser to effectively recover signals suffering from severe quantization noise. In our 1-m OWC experiment with a 680 nm vertical-cavity surface-emitting laser (VCSEL) transmitter, FTER first extracts a rich set of primary and higher-order spectral features from the quantized received signal and employs a random forest to select the most informative features. These features are then fed into a dilated TCN, which learns to suppress structured quantization artifacts and restore the signal. Experimental results confirm that the proposed method enables reliable 32-QAM and 64-QAM transmissions at baud rates up to 0.625 GBaud and 0.75 GBaud, respectively, achieving bit-error rates (BERs) below the hard-decision forward error correction (HD-FEC) threshold. Specifically, under 4-bit quantization with 64-QAM at 0.75 GBaud, the proposed FTER scheme improves the BER from 7.2 × 10^–3 to 3.7 × 10^–3. These findings demonstrate the feasibility of deploying compact, low-cost OWC links.