Design electronics for cold environments
Engineers at Agilent Technologies found a need for a digital multimeter (DMM) technicians and electricians could use in cold environments. "I went to college in Canada where temperatures often got down to below -20°C," said Boon Juan Tan, R&D manager in Agilent's Basic Instruments Division. "At that temperature people would put a DMM inside their jacket to prevent the LCD from freezing."
To create the new U1273AX handheld DMM, which operates between -40° and 55°C, designers at Agilent took advantages of earlier DMM designs. The new DMM could not only operate in colder environments, the ergonomic design let warmly dressed users easily operate and handle accessories. "The main problem was that an LCD would freeze or become slow to respond at about -20°C," said Tan. "So we changed to an OLED display that operates down to -40°C and offers a wider viewing angle."
"We also chose a lithium battery that would operate down to
-40°C, and designed a case that protects against water and dust, and meets the IP54 standard," continued Tan. "The characteristics of interconnects and switches change with temperature, so we simulated them and designed for cold-weather use. We also created a knob and pushbuttons large enough to operate with gloves on."
Although we don't yet use DMMs on the surface of Mars or the Moon, some engineers routinely design electronics to operate in these and other cold environments. Martian surface temperatures can drop to -113°C, or 160 Kelvin (K), at a pole in winter and down to -150°C (123K) on the Moon in darkness. By comparison, liquid nitrogen has a temperature of 77K and surfaces in space can get as cold as 4K. Design of electronic equipment that will operate under those conditions requires much care.
Dr. Randall Kirschman, an independent consultant, explained that engineers shouldn't get frightened by designing circuits that must work at low temperatures. "There are hundreds of low-temp circuits operating in the field, some to as low as a few degrees above absolute zero (-273°C)," he said.
"The main challenge becomes finding components for low-temp use," said Kirschman. "There exists no large market for them, so most companies don't specify components for low temperatures. Designers typically purchase semiconductors and test them at low temperature to find the ones that work. But when they later purchase more they might discover the manufacturer has changed its process. The normal-temperature behavior remains the same, but the low-temperature behavior has changed. The new components don't do what the engineers need."
Kirschman continued, "As you decrease temperature, the performance of a component can experience a gradual or an abrupt change, or it could stop operating altogether. A MOSFET or CMOS device will often operate below liquid-nitrogen temperature (-196°C) and often, performance improves. On the other hand, a standard silicon bipolar transistor stops operating at about –150°C, or it has such low gain it becomes useless." This results from band-structure changes, not from “freeze out.” Freeze-out in silicon -- when dopants no longer ionize and the semiconductor becomes an insulator -- doesn't occur until about –230°C."
The interface, both electrical and thermal, between low temperature and room temperature requires attention. And engineers should remember that components have different environmental and internal temperatures. "Devices have a higher internal temperature because they dissipate power," explained Kirschman. "If you have thermally insulated the device, it can have a warm internal temperature and can work well even though you have a lower environmental temperature. A high-power device can heat up quite a bit inside its package."
Dr. Henning Leidecker, chief failure analyst for the Electronics Division at NASA's Goddard Space Flight Center (GSFC), noted that at low temperatures changes in the properties of materials also can cause problems in electronic systems.
"NASA has experienced problems with grease on connectors at least twice at the Kennedy Space Center," said Leidecker. "Usually the grease gets squeezed away and the contacts make good electrical connections. But when the pins cool to below 50K, the grease becomes solid, insulates the pins, and the pin-socket pair becomes 'open-circuit.' This problem cost NASA quite a bit of money before we tracked it down."
"We realized once we had the electronics in place we would never undo the connector," continued Leidecker. "So we removed the socket body, mated the individual pins, soldered them in place, and filled everything with foam."
In 1981, NASA started to build the Cosmic Background Explorer (COBE) satellite. The signal from a detector passed through a fiber-optic cable from the cold part of the satellite at a few Kelvin to a warmer section that contained the other electronics. "When we tested the satellite during development, this signal died," said Leidecker. "We discovered that in a cold environment the index of refraction for the fiber core and its cladding changed markedly with decreasing temperature. They swapped their relationship, so light injected into the core immediately went into the cladding which absorbed the light. That took a while to figure out."
The Swift Gamma Ray Burst Explorer used 100,000 commercial off-the-shelf (COTS) components. "Because the satellite's gamma-ray detector has 16,000 channels, we could tolerate the loss of data from several thousand without degrading our measurements," noted Leidecker. "We chose a particular type of COTS op amp because of its exceptional specs, and we built electronics for 200 channels with COTS parts from a US distributor. "The test examples of this op amp worked fine, but the batches used for production did not: some experienced high-amplitude low-frequency oscillation, or 'motor boating,' at -40°C, some at -20°C, and others at -10°C. This test illustrated the batch-to-batch variability in part lots. Not surprisingly, we couldn't replace the original op amp with any other type. Fortunately, we found substitute parts for circuits that used this component." Sometimes you must design circuits to tolerate this type of performance difference or adjust it to account for such differences in cold environments.
Some designs for low-temperature electronics might require an ASIC to perform special operations or signal conditioning. "But design models do not exist for 'cold electronics,'" explained Dr. Akin Akturk, vice president of CoolCAD Electronics. "Models assume room-temperature operation, and many designers think those models work the same way at low temperatures, which they don't. The SPICE models don't go below about -50°C [223K]."
"We started to help NASA with ASIC designs that would work at temperatures between 20 and 40K," said Dr. Neil Goldsman, president at CoolCAD. "We used semiconductor physics to create models of how electrons move in circuits and components and developed new design tools so engineers know they will get accurate results."
"Before we worked with NASA their people focused on exotic semiconductor processes because at the time general wisdom held that carrier freeze would cause problems at low temperatures," continued Goldsman. "But mainstream devices, such as CMOS FETs, didn't freeze out, so NASA could use standard semiconductor processes to reduce ASIC costs. It turned out the current CMOS process works well down to 4K."
"I'll give you an example of what can happen when you go and use a component at low temperature," said Akturk. "Many sensor circuits need a current amplifier. Say the sensor produces 0.1 mA and you get 50 mA out--a gain of 500. But at low temperatures designers saw the gain increase from 500 to 5000. They had room-temperature circuit models and didn't know where the increase in gain came from."
"When people use SPICE models, they assume everything will operate at the same temperature," said Goldsman. "But you could have a sensor and signal-conditioning components at 20K and other parts of the circuit at room temperature [293K]. You must simulate the parts of a circuit at their operating temperature."
"When you design an ASIC, you must keep your models up to date,too," emphasized Goldsman. "We have seen fabs change a process, but use the same name given the old process. That will cause confusion."
Time to move on
After writing for ECN since 2003, the time has come for new challenges. I have enjoyed researching and writing about interesting topics and hope you found my articles and columns useful and worth reading. See you down the road. --Jon
For information about the Kelvin temperature scale and its origin, visit: http://en.wikipedia.org/wiki/Kelvin.
Agilent Technologies U1273AX Handheld Digital Multimeter: http://www.home.agilent.com/en/pd-2180564-pn-U1273AX/.
CoolCAD Electronics: http://www.coolcadelectronics.com/.
Extreme Temperature Electronics tutorials: http://www.extremetemperatureelectronics.com/tutorial1.html.
For other resources, please visit: http://www.extremetemperatureelectronics.com/resources.html.