Transcript ppt
SiLab presentation on Reliable Computing Combinational Logic Soft Error Analysis and Protection Ali Ahmadi May 2008 Outline • Introduction • soft error analysis of combinational logic • Techniques for Protection of combinational logic to soft error Why Combinational Logic? • increase in the amount of combinational logic per chip, increase in frequency of operation and a decrease in the logic gate dimensions, the contribution of combinational logic soft error rate to the MTTF is increasing. • Several studies have been and are being conducted to design which are more robust against these soft errors [1]. • According to [3, 4], the logic soft error rate (logic SER) will equal the unprotected memory soft error rate in 2010. • little attention has been paid to increase the robustness of combinational logic elements. Soft error analysis • Soft errors arise from the interaction of alpha particles and cosmic neutrons with silicon. The electron-hole pairs generated by this interaction induce current pulses at transistor junctions, which can result in a logical fault. • standard cells aren't equally sensitive, and all transistors within a standard cell don’t contribute equally to the logic SER of a gate. • in combinational logic, a particle strike causes a transient error, manifesting itself as a glitch propagating to the primary outputs or the next level of flip-flops Soft error analysis n-transistor directly connected to the output, has the highest probability of inducing a transient. 1- higher mobility of electron 2- location of transistor The failure rate values in the table are rounded values based on critical charge simulations using models calibrated with data from alpha and neutron SER measurements on memories. [1] Circuit level protection of SER • Transient probability depends heavily on the input combination, which influences the amount of sensitive area and the total drive strength of the transistors driving the output node. • Use gate multiplication method [1] How to Handle Soft Error • Error detection and retry – using concurrent error detection (CED) – If an error is detected, the system recovers through rollback and retry thereby preventing a failure. • Error masking (TMR) – Real-time systems Is Soft error latched? If a soft error occurs at an internal node of a logic circuit, there are three factors that determine whether it will be latched and result in a error: 1) The rate at which an SE of sufficient strength to propagate to a latch occurs at a node 2) The probability that there exists a functionally sensitized path from the node to a latch 3) The probability that the SE is captured in latch [3] Techniques for protection • A well known circuit level design approach is triple mode redundancy (TMR). - More than 200% overhead - Sensitivity of voter itself to soft error • Partial error masking Cluster sharing reduction Dominant value reduction Partial error masking ( Cluster sharing reduction) • soft error susceptibility of certain nodes in the logic circuit can be orders of magnitude higher than that of the other nodes in the design. • nodes with low observability and controllability be clustered together. • The clusters are removed from two out of the three copies of the TMR design. [2] Partial error masking ( Cluster sharing reduction) Highlighted gates clustered. [2] Partial error masking (Dominant value reduction) • Differentiates between the logic 0 and logic 1 soft error susceptibility of a primary output. The idea is to identify such outputs and replace triplication by duplication in such instances. Partial error masking (Dominant value reduction) original failure rate is the sum of the logic 0 and logic 1 failure rates for a single copy of the circuit. [2] Partial error masking Combine Cluster sharing reduction and Dominant value reduction [2] Simulation results of Partial error masking [2] New technique for Soft error Protection • Failuer Rate of a Gate – Type of gate – Input vector of gate • contributions of all gates in the circuit are affected by input vectors • Disadvantageous of last techniques – Didn’t consider combination of input vectors Input stimuli methods • Statistical Input Vector (SIV): uses only one (statistical) input vector • All Possible Input Vectors (AIV) • Random Input Vectors (RIV) - This method uses a certain number of random input vectors. Circuit with 50% SER reduction by duplicating only 3 gates [1] References [1] A. K. Nieuwland, S. Jasarevic, G. Jerin ” Combinational logic soft error analysis and protection”, 12th IEEE International On-Line Testing Symposium (IOLTS'06) [2] K. Mohanram, N. A. Touba” Partial error masking to reduce soft error failure rate in logic”, 18th IEEE International Symposium on Defect and Fault Tolerance in VLSI Systems (DFT’03). [3] W. Wang, H. Gong, ”Edge triggered pulse latch design with delayed latching edge for radiation hardened application”, IEEE Trans. on Nuclear Science, vol 51, no. 6, pp. 3626-3630, Dec. 2004. [4] R. Baumann, “The impact of technology scaling on soft error rate performance and limits to the efficacy of error correction”, IEDM , pp. 329-332, 2002.