A Parallel Transmission Scheme Based on the Turbo Product Code for Through Silicon via Array in Three-Dimensional Integrated Circuits

Through silicon via (TSV) is the key technology for the vertical interconnect in three-dimensional integrated circuits (3-D ICs). With the help of TSVs, higher throughput in signal transmission can be attained. However, the tightly clustered TSVs in the TSV array suffer from crosstalk noise, a situation which results in transmission errors. This study investigates, the channel model of the TSV array, involving main factors affecting transmission performance, such as transmission loss, inter-channel interference, and crosstalk noise with different digital patterns. A parallel transmission scheme based on a turbo product code (TPC) parallel coding is proposed. In this scheme, the binary bits of information are reshaped into two dimensional blocks. Each block is parallel encoded using a TPC into a codeword block for parallel transmission through the TSV array channel. At the receiver, the sending information is reformed by a concurrent hard decision decoding algorithm of a TPC. This parallel transmission scheme achieves low bit-error-rate, high throughput, and a lower system overhead relative to that of its ground TSV shielded counterpart if the type of TPC is carefully selected. The simulation results confirm that this scheme reduces inter-symbol interference, minimizes structural defects in the TSV array, and improves transmission performance in 3-D ICs.

Hard Decision Decoding Performance Improved Using Turbo Product Codes

The performance of soft decision decoding, whose for which the design is complex, is superior to the performance of hard decision decoding. In this paper, we propose a turbo product code with a bit flip algorithm to improve the performance of hard decision decoding. The performance of hard decision decoding is improved with low complexity using multidimensional turbo product codes. The reliability of decoding in a communication system to detect and correct errors is discussed .Maximum a posterior probability (MAP) decoding is employed to improve the hard decision performance of turbo product codes with multiple dimensions. Our results include comparisons of multiple dimensions—2D, 3D, 4D, and 5D—and the number of iterations in soft and hard decision decoding.