Comparison of Power Converter Models with Losses for Hardware-in-the-Loop Using Different Numerical Formats

Nowadays, the Hardware-In-the-Loop (HIL) technique is widely used to test different power electronic converters. These real-time simulations require processing large data at high speed, which makes this application very suitable for FPGAs (Field Programmable Gate Array) as they are capable of parallel processing. This paper provides an analytical discussion on three HIL models for a full-bridge converter. The three models use different possible numerical formats, namely float and fixed-point, the latter with and without optimizing the width of signals to the embedded DSP (Digital Signal Processors) blocks of the FPGA. The optimized fixed-point model (OFPM) uses three and two times fewer DSP blocks or LUTs (Look Up Tables), and the maximum achievable clock frequency is also up to 35 % and 25 % higher than the float model and non-optimized fixed-point model (nOFPM), respectively. Furthermore, the models’ accuracy is proportional to the clock frequency, thus the OFPM is also the most accurate model. Finally, the paper shows the differences in the simulation when the models include or not losses, proving that not including losses leads to high errors, especially during transients.