Hybrid automatic-repeat-request (ARQ) is a flexible and efficient technique for data transmissions. In hybrid ARQ, subpacket schemes are more attractive for systems with burst errors than complete packet schemes. Although subpacket schemes were proposed in ARQ systems, optimum subpacket transmission is more effective to maximize throughput in a dynamic channel. Since convolutional codes have properties of burst errors in decoding, the optimum subpacket can be applied to convolutional codes. This paper investigates the performance of subpacket transmission for convolutionally coded systems. An efficient method is proposed to estimate the optimum number of subpackets, and adaptive subpacket schemes, i.e., schemes that enable a system to employ different optimum numbers of subpackets under various conditions, are suggested to achieve the maximum throughput of the system. Numerical and simulation results show that the adaptive subpacket scheme is very effective for the convolutionally coded hybrid ARQ system, and it can provide higher throughput, smaller delay, and lower dropping rate than complete packet schemes. Moreover, the adaptive subpacket scheme can be flexibly
used with packet combining techniques to further improve the system throughput.