We have obtained a method for deriving an analytic trade-off between energy efficiency, complexity and signal quality, for the complete DSP chain. We have applied this method to study the digital signal processor (DSP) of an RF Pulse Width Modulation (RF-PWM) transmitter and we identified points for improvement. We derived new algorithms for improving the performance of DSP blocks. We have shown performance guarantees in terms of their accuracy and cost. We implemented them, in software and hardware, and carried out tests illustrating their performance. The algorithms are scalable and come in variants aiming at satisfying different performance constraints. We provide a methodology for designing DSP blocks using a modular approach. We use templates to represent the different modules. We have formed a library of modules for constructing telecommunication system DSP blocks. The library contains functions based on our new algorithms as well as additional functions, a number of which are based on more than one algorithm. All the library functions are implemented in fixed precision using only simple arithmetic operations, register shifts and bit operations. We incorporate them in the programs by instantiating the templates using specific values for their parameters. The modular approach makes the design of DSP blocks easy to maintain, customize and allows transparent updates and modifications. We used our algorithms to design the DSP block of the RF-PWM transmitter and implemented it in an FPGA. We have derived a design flow for communication system DSP blocks. Finally, our new algorithms have an impact in quantum computing. Indeed, it is not difficult to convert the newly derived algorithms to quantum algorithms and circuits.
We have obtained the following results:
• A method for the determination of an analytic trade-off between energy efficiency, complexity and signal quality, for the complete DSP chain
• Determination that algorithms computing inverse trigonometric play a key role in the DSP of the RF-PWM transmitter and that existing algorithms were not adequate. New algorithms would be necessary
• Design algorithms, derive their accuracy, and cost analytically for the selected RF-PWM DSP block
• Derive new scalable and efficient algorithms for inverse trigonometric functions that are critical for the RF-PWM DSP block
• Obtain digital libraries (in Matlab and Verilog) that include modules for building DSP blocks as well as the entire DSP block for the RF-PWM transmitter
• Implementation of the DSP bock of the RF-PWM transmitter on an FPGA
• Obtain a seamless design flow for DSP blocks
• Show a sequence of successively refined steps combined with a modular approach, where modules rely only on input and output register sizes for their combination. This leads to a seamless design flow for DSP blocks and their hardware implementation.
The work in SAPHIR has resulted in two papers, an MS Thesis, 6 presentations at seminars/workshops and 5 further workshops to communicate the Idea behind Saphir and the MSCA Fellowship. The researcher had been working in the US until the commencement of this project. The project has provided him with valuable knowledge enhancing his future career potential and to reintegrate in Europe. IFAT has extended the researcher’s contract beyond the end of this project. IFAT is a leading semiconductor company. The results of this project will be used in the ongoing and future development of its portfolio of transmitters and wireless infrastructure and further projects.
