This week marks the first underwater tests of the transmitter, with the entire transmission chain set up. This is an important achievement, since it shows that the completed half of the system is ready for usage in existing systems. In particular, the interface between the GODIL board, the acoustic transmission board, and the transducer was completely validated, and the process was streamlined to speed up future trials.

The tests, using binary FSK (21 and 25,5 kHz) at 1500 baud, were performed in the OceanSys test pool, an environment that is extremely fading due to multipath from the pool walls, floor, and water surface. The measured channel coherence time is on the order of 100 ms, which will certainly be a challenge for the receiver. The reception performance will be tested over the next week, using the existing Simulink model. The preliminary (purely visual) analysis of the data strongly suggests that a frequency hopping scheme for added diversity would be very useful in improving reception error rates.

After revisiting the previous analysis of the system parameter space, I noticed that by decreasing the data rate to half of what was used in the underwater tests (750 baud instead of 1500 baud) ten symbol frequencies became available for use. This opens up new possibilities for increased constellation size and frequency hopping or coded modulation (e.g. 2 jumps per symbol with 4 symbols), which will hopefully contribute to achieving target throughput with reduced error rates.

The receiver design was not forgotten, and MATLAB tests were performed on the Xilinx implementation of the FFT core, using a vendor-supplied bit-accurate model. For a 12-bit input width, the phase factor width (twiddles) can be reduced, with better implementation area results. I also investigated Automatic Gain Control (AGC) strategies, as well as alternative DFT calculation methods which may prove more efficient than computing a complete FFT (such as the Goertzel algorithm or the sliding DFT).

Since we have expended a significant amount of time on the underwater tests, the workplan will have to be slightly rearranged. However, we have also thought of some ways to reuse much of the FSK modem infrastructure in the DPSK modem, which would help reduce development time and keep the most significant deadlines intact.