P14254: Underwater Thermoelectric Power Generation


Table of Contents

Power Generation (Mechanical) Subsystem


The system was left underwater multiple times during testing, with the longest time being 24 hours. No leaks were detected once the gasket issue was resolved.

Power Generated

The table below shows the testing results for the generation subsystem.
Date Heat In (W) Electricity Out (W) Conversion Percent
Mar 18 500 14.4 2.88
Mar 18 630 20.17 3.20
May 07 500 13.8 2.76
May 07 630 20.3 3.22
May 07 630 20.0 3.22

The system was disassembled and reassembled between the March 18th and May 7th tests, so the repeatability of the results shows a robust design. The conversion efficiency above 3% is better than many systems in the literature.


Fujitsu pressure film was used to test our 90 psi clamping pressure requirement. The first clamping test showed that we were not getting a very uniform clamping pressure, likely due to burrs on the heat spreader. The results of this test can be found here.

The second clamping test proved more successful. The clamping pressure was much more uniform, however the average pressure of 130 psi was higher than the design pressure of 90 psi. This is not a concern, however, since the thermoelectrics can handle this pressure as well. The results of this test can be found here.

ZETA Converter

The raw data of all the Zeta testing can be found here (XLSX).


The GaNFET Zeta converter was soldered to the TE Board along with all the other components and subsystems, yet only the Zeta converter was actively being tested. The input to the board was a DC power supply set to 12 V and a resistor was used as a load instead of a battery. The controller (ATTiny) was not used and a function generator creating a 5 V Pulse Width Modulation (PWM) was used in its place. The AND Gate enable was pulled high and all the other outputs of the ATTiny were pulled low to avoid floating nodes.

The PWM was started at 50% duty cycle and the voltage at the output was 4 V but the expected voltage was 12 V. This was a cause of the power supply reaching its current that was set to 100 mA. A second test at 20% duty cycle was preformed ant the voltage at the output was 2.8 V and the expected voltage was 3 V. This was only a 6.7% percent error. The efficiency was not calculated however. A third test at 30% duty cycle put the power supply at its current limit again, so the current limit was increased to 200 mA at that point clicking noises were heard and the test was turned off. The 20% duty cycle test was repeated and the results were that the supply was in constant current which signified that a GanFet was blown.

The cause for these results was that the driver for the GanFETs were not able to switch the dual GanFETs in a desired manner which led to the high side GaNFET shorting to ground and being blown.

The raw data of this test can be found here.


The "new" Zeta design was with a three P-Channel Power MOSFETs in parallel, the 8.2 uH couple inductor, a 13 ohm load, a 94 Ohm resistor in the driver circuit, and the input was a simulated thermoelectric; which consisted of a DC power supply set to 19.7 V with a 4.6 Ohm power resistor in series with the supply.

After doing some tests with a function generator as our PWM the results proved promising but the function generators duty cycle only ranged from 20%-80%. Due to this the function generator was switched for the ATTiny which had the ability to output a full range of PWM duty cycles. The results were that at mid-range duty cycles the expected output voltage was reached within 1-20%. Within this same range of duty cycles the efficiency of the converter peaked at 47.19% (which was at 47.5% duty cycle). The maximum power point was also reached at within the mid-range duty cycles, 53.7% duty cycle specifically, with a power input of 20.26 Watts and the efficiency at 45.11%.

We estimate that the diode dissipated 1.4 W, the gate drive circuit 1 W, and there was 7 W dissipated through the transistors due to switching losses. The switching losses were the result of an imperfect gate signal on the parallel PMOS's.

The figure on the left shows that the maximum output voltage of the system was around 11 V, which is too low to charge our six cell lithium ion battery pack. As a result, we were unable to test whether the system could charge the battery.

The raw data at which maximum efficiency was achieved can be found here.

Power Analysis of Zeta