Table of Contents
Motivation
In the Andromeda III rocket, there are many challenges and extremes in terms of its power design. From the low power used for the avionics package to the high current draw of the main engine solenoids, many considerations have to be made in order to create a system that can power the wide range of components on board. This is particularly true during the launch procedure, where the rocket must seamlessly switch its power source from the ground system to the onboard battery supply.
Attempts have been made in past to create a circuit that can gracefully switch and closely monitor these different power supplies but success has been limited by the constants of problem. First of all, any significant drop out in power while switching would prove disastrous for the flight computer as much of its launch-ready capabality relies upon volatile memory. The other challenge is the large amount of current such a switching device would have to pass in order to properly energize the main engine solenoids, which are responsible for controlling ignition of the engine. Earlier boards used electromechanical relays controlled by a small integrated circuit but such boards ultimately failed to cope with the large amount of heat generated by high current loads. (honestly idk how to make this a suggestion other than putting it here but i think this paragraph is unnecessary because it’s just talking about how hard it is to create this thing without giving any useful information relating to how it was actually made)
During the launch, an umbilical cord from the ground system powers the rocket. Its use also conserves power while the rocket is on the launchpad. The umbilical cord is then released from the rocket and, ideally, it should switch seamlessly from the ground system to the battery. It should also switch seamlessly from the primary battery to a number of backup batteries in case of a system failure. The ideal behaviour of the rocket is summarized below.
Circuit Design
Several mechanisms to facilitate switching and voltage sensing were considered while planning the rocket. Single Pole Double Throw (SPDT) relays were familiar, but quite slow, with the fastest available models clocking in at around 100 microseconds. Considering the high current draw and tight voltage tolerances, even a 100-microsecond drop out would require ten of thousands worth of capacitance to compensate, which was far too much added weight. Relays also used a great amount of current to keep circuits closed, a function more useful in other parts of the rocket. Devices like MOSFETs were found to be a superior option as they require very minimal current and can switch rapidly.
The next proposed design was a diode ORing circuit. As the name suggests, a diode ORing circuit functions much like a logical OR gate, but instead will OR two power sources of similar voltage level into a single power source from the perspective of the load. The diode’s cutoff voltages could be selected such that one power source has priority over the other; in this case, the ground system receives higher priority. The diode ORing looked promising, but concerns surrounding the tolerances of high power diodes and the lack of versatility with a passive circuit meant this solution was far from perfect.
After consulting a professor at the University of Calgary who specializes in sensors, biophoniotics, and VLSI, the use of a power multiplexer was considered. The multiplexer outputs a single power source selected from a multitude of available power sources based on a select signal.
Texas instruments makes a line of priority power muxes dubbed the “TPS” series and after doing a bit of research, the TPS2121 fit all the requirements.
Although the TPS2121 functioned as 2 to 1 multiplexer for power, a single IC was not enough to accommodate all possible power sources for the rocket. Instead, two TPS2121s were strung together, effectively expanding the number of ports.
The last step before building any kind of test circuit was determining the required component values. TI’s power muxes are programmed for priority based on the voltage certain pins will receive, which can be calculated via formulas available on the datasheet and then implemented using simple resistor dividers. To aid in the design process of the circuit, a calculator was created in Excel and several scenarios were run before settling on the final values.
Evaluation of TPS2121
Actually using the TPS2121 once it arrived proved to be extremely difficult, as the package itself was extremely small and had metal pads that were impossible to solder using any kind of conventional iron. Instead, a “breakout” board was designed to act as a converter between the minuscule QFN form factor and the breadboard-friendly 0.1″ pitch header pins. Extra vias were placed in pads that carried power in order to increase heat dissipation and trace width.
In order to get a feel for programming a device of this complexity, the circuit found in the “typical applications” section of the datasheet was recreated.
The switching speed was slow, at around 85 microseconds but the switching voltage was recorded at 7.94V versus 7.6V, which is within the tolerance of the resistors being used. After experimenting with the slew rate pin, the switching time was greatly reduced and shortened to 5 microseconds.
Next, the values of the components determined in our calculator were substituted into the circuit and a simulated load was added. Despite the use of 5% tolerance resistors, the measurement results were much closer to the desired result than expected.
Evaluation of Dual TPS2121
(Coming soon!)
PCB Design and Assembly
(Coming soon!)
Final Test
(Coming soon!)