Design And Implementation Of Low-Power Digital Filter

In recent years, the rapid development of biomedical equipments, IoT(Internet of Things), wearable devices, mobile phone and other portable devices led to huge power dissipation, because of the more and more powerful processing capabilities and the faster and faster processing speed. As well as the chip power density has a rapid increase with the decrease of CMOS process dimension, the battery technology has not been a breakthrough. The development of low-power technology is imperative, because the endurance of battery has restricted the development of portable devices. Digital filter is widely used in DSP(Digital Signal Processing) because of its various advantages. Especially, because of the liner-phase, the FIR filter is the most popular one and widely used in medical devices such as the hearing-aid. Meanwhile, digital IC is designed based on the worst case, which rarely appear, so it is bound to cause a waste of resources. Therefore, the resource consumption can be reduced by sacrificing the accuracy of some results, which is called the approximate processing theory. The approximate processing was first applied to the computer systems, then by virtue of its excellent performance is widely applied in various fields. According to approximate processing, treading off between the power dissipation and the accuracy of output, this paper has designer a lowpower digital FIR filter.First, this paper introduces the development trend and difficulties of digital integrated circuit, and the value of low power technology. Meanwhile, this paper introduces the development of DSP, and the significance of the design of digital filter is analyzed. Secondly, this paper introduces the component parts, cause and calculation method of the power of digital IC, and some low-power technology starting from the level of design. Thereafter, this paper pays attention to the multiplier that is widely used in digital IC, and presents a CSD truncated multiplier with simple compensation circuit and minor error, which is very suitable for the digital filter. Then, this paper presents a method of applying the theory of approximate processing to digital filter, and this method is verified by the theoretical derivation. Finally, after comparing between IIR and FIR filter, a low-power digital FIR filter based on the theory of approximate processing is proposed, whose order and coefficients can change dynamically. It can reduce 70% power at a cost of the accuracy of products. And, the front-end design, back-end design and the simulation analysis of the proposed filter are completed based on 130 nm standard CMOS technology.