The High–Performance Spaceflight Computing efforts from NASA will bring us wonderful things, but the processors need power.
Powering High-Performance Spaceflight Computing
By: Paul L. Schimel
NASA has started a wonderful initiative to improve computing capability in space. It is called High–Performance Spaceflight Computing (HPSC). From the silicon design side of things, this means that there is a big push to get the tiniest features into a radiation–hardened space processor. The new processes will reduce operating voltages, increase speeds and reduce the energy required to complete the mission by drawing less power. There are brilliant radiation–hardened Integrated-Circuit designers working on this feverishly. But alas the processor mustconsume power to do work. We do not get to fix this in code. The work on a power solution to drive these HPSC processor solutions will be as difficult as the design of the processor itself.
A bit of earthly history
High–performance computing comes with power delivery challenges that took decades to unravel. When we developed high–performance processors for gaming, simulation, and financial computing, these processors all came with heavy power requirements. They had very tight tolerances on the voltage rails, extremely fast dynamic loading, and extreme operation including jumping from low–power sleep mode into a highest–performance mode in a couple of power conversion cycles.
It was quickly realized that a high–efficiency, switching power solution was needed beyond a single–phase buck regulator to make these things work. The current demands showed us that a multiphase solution was needed. This gave rise to multiphase buck architectures which in turn started the low–voltage metal-oxide-semiconductor field-effect transistor (MOSFET) revolutionto deliver power switches that could operate at these extremes efficiently. This drove a plethora of innovations from discrete MOSFETs, inductor, and capacitor improvements to MOSFET pairs with integral drivers, to monolithic MOSFET structures that were perfectly tuned for optimal efficiencies. Technological progress eventually crept into wide bandgap power switches for thelowest losses, highest efficiencies, and highest speeds. The operating modes showed us that we needed phase fold back operation to keep the converter efficient for low–power sleep mode and then a means to turn on all phases fast and deliver massive amounts of current at very high speeds. This then showed us the need to vary the speed of the feedback loops of these converters to maintain optimal stability over these wildly varying load conditions. This was not analog control territory. Digital control had to be implemented and it was done in several wonderful ways. The industry learned a lot.
This history is of paramount importance for HPSC power solutions. These lessons must be applied. The missions and the processor will not wait years for a power solution. We do not have time to reinvent this wheel. The design challenges must migrate to surviving heavy ion, gamma, proton, and neutron radiation events in space.
HPSC will deliver wonderful contributions to controls, data gathering, and processing. These contributions will enable deeper space exploration, untold discoveries, and improved understanding and modeling of the world around us. But the processor cannot do it without a power solution.
About the Author
Paul L. Schimel is a principal power electronics engineer in the Aerospace and Defense group at Microchip Technology. He has over 27 years of theoretical and hands-on experience in power electronics, spanning military, aerospace, automotive and industrial markets. Paul’s work regularly includes module design, DC–DC converter design, device-level analysis, root cause analysis, failure analysis, EMI mitigation, PCB layout, control loop compensation, inverter design, transformer design, rotating machine design, bench-level measurement and validation techniques and system-level analysis/comprehension. He has designed DC–DC converters from milliwatts to megavolt-amps, inverters to 5,000 HP. He is a licensed professional engineer in several states and holds two FCC licenses (First Class Radiotelephone and extra class amateur). In addition, Paul holds four patents on power electronics matters with two pending.