Ideas for self learning project

Question

I have knowledge in DSP and some basic knowledge of RF. I wish to do a self-learning project that will incorporate the following:

• signal processing implemented on an FPGA.
• Interfacing the FPGA to an A2D, D2A, or both.
• Some board design (for the FPGA, analog frontends, and interfaces)
• Might add some MCU if needed.

Other considerations:

1. Since I wish to avoid too much RF trouble, I wish to avoid high-speed design and high frequencies where possible.
2. This has to be something I can accomplish on my own in my free time (but I don't mind if it takes a year to complete).
3. Since this is aimed to be a hobby and not something commercial, hardware cost is a concern. Thus I can't use a costly and big FPGA.
4. I have access to lab equipment, including scopes, network analyzers, and other needed equipment.

Do you have suggestions? I never did any project of this magnitude, but I did take part in complex projects, but only at a limited scope. I have experience with implementing algorithms on MCUs, as well as digital communication simulations (using MATLAB).

I thought of LoRa since the standard seems the simplest. I would like to get other ideas in general and/ or other relevant standards. I also looked at Thread, but LoRa seems simpler at first sight.

Answer 1

Since RF is mentioned as signal source, consider going really low-frequency to broadcast time standards like WWVB at 60kHz. There's bound to be one in your region.

• The ADC is not costly, since sampling rates need not be crazy-fast.
• RF preamplifier is fairly easy, making analog front end simple.
• Loop antennas or tuned ferrite-rod antennas are adequate.

I've done this project on FPGA, receiving WWVB at 60 kHz. Both AM modulation, BPSK modulation were processed, decoding to digital bits. WWVB signal specification is well-documented.

After completing FPGA, the project was migrated to an 8-bit microcontroller (10 MIPS) with integrated 10-bit ADC. Its ADC was the limiting factor, sampling 80,000 s/s. Was surprised that a 10-bit ADC was adequate, but 12-bits would allow a wider dynamic range to accommodate the very large noise amplitudes at these frequencies. For this microcontroller project, filtering was limited to rectangular filters, since the microcontroller had no multiplier. Decoded bits were sent out its serial port (20 consecutive minutes shown below):

The microcontroller project was attempted to see what DSP algorithms were most appropriate to use on a simple RISC 8-bit machine, and to find which of its hardware resources limited performance.