Quantization errors in digital motor control systems

When implementing a motor control drive scheme digitally, the quantization errors always exist in the system. Two major sources of quantization errors are analog-to-digital (A-D) process and numerical calculation in the fixed-point computing device. Typically, the effects of quantization errors due to A-D conversion contribute less than one produced by numerical calculation. This paper studies the quantization errors in a sensorless direct vector control system of induction motor system using a 32-bit fixed-point digital signal processor (DSP) from Texas Instruments (TMS320x28xx series). The investigation of quantization errors produced by numerical computation in such DSP is focused. Both simulation and experiment are carried out within DSP itself in three kinds of data formats; 16-bit fixed-point, 32-bit fixed-point, and floating-point. By comparing the results between floating-point and fixed-point implementation on one machine, numerical issues related to quantization errors can be verified and resolved. As a result, the system performance and behavior can be affected by quantization errors in 16-bit word length while it is not significant in the 32-bit word length.