Low-voltage Low-power Log-domain Integrated Circuit Technology For Synthesis Of One Dimension Continuous Wavelet Transform

The Wavelet Transform is a new style mathematic analysis tool. It is a new theory system developed from the studies of Y. Meyer, S. Mailat and I. Daubechies in 1980s. The Wavelet Analysis is a kind of a self-adapted time-frequency localized method, which tune the time-frequency window automatically and has a strong flexibility. Its applications relate to a lot of fields, such as the communication signals treatment, the radar images analysis. For adapting the real-time requirements of signal treatment in project, the study of using hardware to realize wavelet transform are developed rapidly.The realization of wavelet transform includes two sides: the Discrete Wavelet Transform and the Continuous Wavelet Transform. So far as achievements in study, it is successful in realizing the CWT using digital circuits, and a lot of applied chips had appeared. But the developments of study in using analog circuits to realize the CWT are slowly and only few system architectures and applied chips had been given. There have a lot of aspects need to be improved, such as system integration, analysis precision and power dissipation.In this paper we study the realization of the CWT using analog circuits, and by synthesizing log-domain circuit design strategy we present a new system architecture. First based on existed methods of the realization of CWT, we present a method using log-domain circuits to improve system capability, that is to say, by using log domain technology to design the function modules in existed architectures we realize the purpose of improvement of system performance. According to the simulation results, a lot of aspects of system performance, such as the tunability of core frequency, the bandwidth of signal treatments, have been improved. Second according to the expression of CWT in both time and frequency domains, by analyzing three kinds of expressions of wavelet functions which have similar construction, that is Marr wavelet, Morlet wavelet and DOG wavelet, we present new CWT systems with share units array, both time and frequency domains. For realizing the functions in frequency domain, we present a design strategy of approaching functions on piecewise frequency; for realizing the Gaussian function in time domain, we present a method ofmutual conversion between characteristic curves in time and frequency domains. According to the system simulation, the implementing current in system is at the level of microampere, demonstrating that we have realized the object of circuits on low supply voltage with less power dissipation; according to the analysis of eight-channel approaching architecture, we realize the signal analysis with a wide frequency bandwidth, from several hundreds KHz to several hundreds MHz.