A chance conversation about solar panels set Microsemi chip designer Shawn Fahrenbruch on a voyage of discovery, resulting in an active diode with a forward drop of only 50mV.
It has only two terminals, there is no separate power supply.
"Seven years ago I was talking to a professor from Stanford and he said one of problems of solar was susceptibility to single-point failure," Fahrenbruch told Electronics Weekly.
That single point failure is the bypass diode that prevents one cell in a string being reverse-biased when it is shaded and the rest of the cells are not.
Without the diode, the reverse-based cell restricts string output current at best, or burns out at worst.
These diodes lead a hard life, sealed into the junction box bolted to the back of a hot panel.
According to Fahrenbruch, as panel output currents are increased to many amps, the diodes are less and less likely to make their 30 year specification life.
And once the diode goes, the panel goes next.
The reason that this is not a well known problem, he claims, is that few solar installations have continuous monitoring, so many broken panels are still in-place.
Fahrenbruch's self-imposed brief was to design an active diode from silicon that replaces the Shottkys used at the moment, but with a far lower voltage drop to keep temperatures low, and engineer it from scratch for life in a solar panel junction box.
"The whole device is designed for 30 years," he said.
Inside is a mosfet connected across the terminals to provide an open or short circuit, plus a control circuit and a switched-capacitor boost converter to power the whole chip from energy harvested from its own 50mV forward voltage drop.
"This may be the first production-worthy energy harvesting solution on the planet which is a self-enclosed solution," said Fahrenbruch.
The actual design remains under-wraps, but it is known to have more than one charge pump - having at least one to get it going from a handful of millivolts and a more powerful one to run the chip after start-up.
Internal feedback maintains the 50mV drop from tens of milliamps up to 20A.
Room temperature reverse leakage is typically 20µA at 22V, increasing to 200µA at 150°C, said Fahrenbruch, compared to 1-10mA at 85°C from a solar Schottky, which rises to 100-500mA at 150°C - sometimes leading to thermal run-away if a hot Schottky is suddenly reverse-biased.
Fast hard breakdown at 25V protects against induced surges from lightning.
"A solar panel is just a magnetic antenna for lightning. You might have 100-1,000A in a diode," said Fahrenbruch, "which is no problem if it is forward biased. In reverse, a Schottky gets destroyed at 150mJ. Active turn-on at 25V reverse gives our part 1.5J survival.
In normal operation, turn on time is 200-500µs.
"It is not something you are going to use in a dc-dc converter," said Fahrenbruch.
The well-established rules-of-thumb that ensure the reliability of consumer chips did not apply at the temperatures and lifetimes expected in and from solar panels.
"This kind of design is a lost art," said Fahrenbruch. "I didn't know, and when I asked what a lifetime of 10 years really meant, no one else knew."
So he went to engineers who had designed chips before the days of ready-made design flows.
"I had to go back to the basics on this. I had to go back to the old guys, the 70 year-old physicists, and talk to about reliability," he said. "Probably the most interesting thing about the design was the calculation on every single transistor, contact and via."
MicroSemi engineers calculated that 175°C testing for 3,000 hours would equal 30 years in a typical junction box that peaks at 85°C and averages 55-60°C.
"We are already over that," said Fahrenbruch. "We have had a good sized sample in the oven with 20A running through them in 140°C airflow, so the junctions are 170-175°C. They have been going to 4,000h so far."
Others samples have are up to done 4,500 un-powered cycles from -40 to +165°C, and a third set has been rapidly cycling from 20A at 140°C, to 0A at -40°C several months.
"There is not a single failure anywhere yet," claims Fahrenbruch, whereas a percentage of Schottkys failed in similar circumstances.
The result of this work is a surface-mount chip under 2mm tall: the 7.5x11.5mm surface-mount LX2400 which has already been adopted by more than one solar company.
Microsemi does not have the solar 'lossless diode' market to itself.
ST Microelectronics' SPV1001 is loosely similar, but runs itself from an on-IC capacitor which is charged by briefly turning the mosfet off periodically.
Each module has many solar cells, protected in blocks by bypass diodes. In an installation, many modules in series will be producing perhaps 600V.