Subject Spotlight: Signal Processing

Welcome to a series covering the variety of electrical engineering subjects available at Melbourne Uni. This series hopes to give those who have not done the subjects a more overall outlook of what the subject entails, tips and more interesting information that is only known by those who have done the subject.

Disclaimer: These articles are written from past experiences and may not reflect what the subject is currently like or will be like in the future. The opinions expressed are purely the authors’ and not representative of the Melbourne Uni Electrical Engineers Club or the University of Melbourne. In no way is anything here presented as fact. Do not come complaining to us if after reading something you think the subject is easy and then you fail or if something crucial has changed in the subject.

Year this subject was taken: 2016

Signal Processing deals with applying transformations onto signals, mostly in discrete time, digitally. In this subject, we learn to design digital filters and implement them to fit certain required specifications.

It is a good idea to do this subject alongside Communication Systems as the two subjects are closely intertwined, even though Communication Systems is wholly focused on analog communications. They share common topics although never at the same time through the semester.

It should be quite obvious that Signal Processing is the prerequisite to Advanced Signal Processing, from the handbook Advanced Signal Processing is focused more on probability and estimation, which is not explored in Signal Processing.

Topics:

  • Sampling of Analog Signals (Revision of Signals and Systems but also much more)
  • Anti-aliasing filters, Butterworth filters, Oversampling
  • DTFT (Mostly revision from Signals and Systems)
  • LTI Systems (Mostly revision from Signals and Systems)
  • z-Transforms (Revision of Signals and Systems but also more in depth with Region of Convergence)
  • Digital Filters, types, properties
  • Analog Filters,  types, properties
  • Design of IIR filters, bilinear transform
  • Design of FIR filters, using windows, using Parks-McClellan method
  • Discrete Fourier Transform (Revision of Signals and Systems but then much more), circular shift, circular convolution, properties
  • DFT based filtering, Overlap-Save, Overlap-Add methods
  • Fast Fourier Transform
  • Converting sampling rates, up-sampling, down-sampling, multi-stage filters
  • Polyphase decomposition of filters
  • Energy Spectrum

Lectures:

The lecturer for Signal Processing is Prof Erik Weyer, lectures include going through slides, as well as written examples and MATLAB examples.

Also Erik will sometimes do a multiple choice question where everyone gets a few minutes to discuss the question then vote by a show of hands what the answer should be. Erik emphasises that everyone should participate, whether that means guessing if you have no clue, so that he can gauge how well the class is understanding overall.

Sometimes there will be another lecture in the week depending on circumstances, the lecturer will send emails about it. This extra lecture time is not on the timetable on Student Portal, but the lecture at this time is usually revision of Signals and Systems or going through exam questions. (Details on this may change as in this particular year 2016 Erik planned to be away for a week and so almost every other week had an extra lecture to make up)

Prior Knowledge:

This subject, like Communication Systems, is a continuation of Signals and Systems.

Signals and Systems is an overall requirement for this subject, as all the basics of different signals, fourier transforms, systems properties are taught in that subject (see the topics section above).

It is a good idea to revise how to write code in C as the workshops will use this extensively.

Workload:

An Analog Devices BF533 DSP board

The workshops consist of 3 workshops with the DSP Board and 1 MATLAB workshop over multiple weeks.

Most workshops have a Section A that has to be completed and shown to the demonstrator before you can get a DSP board and do the rest.

Workshops usually consist of doing mathematical equations, proofs or MATLAB realisations of filters, then replicating them physically on the DSP board.

Workshop groups are not allocated and are groups of 3. The workshop report is shared and is handed in at the assignment boxes in the EE building.

The mid-semester test is 45 minutes and 2 double-sided A4 sheets of paper are allowed. It is a multiple choice test of around 10 questions but there is deducted marks for wrong answers. The final exam is open book.

Tips:

With the lecture slides there is also a student expectations document for each topic use it as a checklist for studying.

Even though lectures and tutorials are recorded, don't assume the recording works, and sometimes what is written on the projector is not recorded.

In this subject, you really have to prepare for the workshops beforehand. There is limited time in the workshops for using the DSP board so all other work should be done before the start of the workshop.

There is only one demonstrator in the workshop so wait times for them can be very long. The first section has to be shown to the demonstrator before you are allowed to get a DSP board, and the filters on the DSP board have to be shown to the demonstrator to be checked off so plan ahead.

The DSP boards require C code to implement the filters, so sometimes you will spend the entire workshop playing spot the missing ; or trying to find a small mistake somewhere. Demonstrators can sometimes not spot the small error in C code either.

The workshop reports do not provide much guidance in how to code the filters or use the software. So brush up on your C skills. There is a debugging tool that will cycle through the code and update variable values so you can spot any errors.

Although highly unlikely, there can be a situation where the specifications in the MATLAB workshop for the filters will be impossible to achieve.

In the exam for this subject, a lot of the 'easier' questions are easy to understand, but difficult to get full marks because the solution to get full marks is deceptively long and it is very easy to make small mistakes. Working out should be double checked, and methods have to be justified.


BY Dennis Nguyen

Melbourne University Electrical Engineering Club

2017

Leave a Reply