Communication over low-SNR environments faces various challenges and data detection designing can be arduous. An aggregative sampling technique with TCP feedback is proposed to transmit in low-SNR channels. The proposed scheme overcomes intrinsic losses by having a physical-transport cross-layer interaction. Samples are aggregated to make a single bit decision. As the quantity of aggregated samples is increased, the bit error rate (BER) is reduced. The transport-layer loss information is fed back to the physical layer to dynamically control the amount of redundancy, therefore, reducing intrinsic loss. Results show that with an SNR of 0 dB the system is able to reach over 10 Mbps with a BER of near 10(-9). It is demonstrated that by implementing the proposed technique it is feasible to reduce the BER by a factor of 10(9) by reducing the effective throughput by a factor of 3. For SNR environments of over -10 dB a BER of 10(-8) is achieved. Performance improvement of 11 dB or more is obtained compared to the analyzed techniques.