Logo   | Search
About
Project Feasibility Study
Hardware Design &
Fabrication
Program Development & Debugging
Product Characterization & Correlation


 
   
 
 
Enter
 
 
Keeping Cool
 

Making electronic devices more powerful while still keeping them cool is a long running problem. Steve Rogerson looks at some of the latest techniques.

The problems of thermal management in electronics are not new. For decades, designers have been trying to squeeze more processing power into tighter spaces and have thus been given the headache of removing the resultant generated heat one way or another.

The history of the problem dates back to the late 1970s and early 1980s when it was the main reason for the switch first from bipolar to MOS and then from MOS to CMOS.

“CMOS gave us almost relatively static power consumption, which is why almost all electronics started to use it,” recalled Andreas Wild, executive director of the Freescale Semiconductor Crolles research centre. The idea behind CMOS was to use two transistors instead of one to do the same job, which meant that power was only needed when the transistor was actually doing something.

"And that is still with us today,” said Wild. “The most efficient way to build digital circuits out of transistors relies on CMOS.”

Though there are no major improvements expected on that technology, there are still innovations being made in the design itself, such as rearranging the transistors in a structure that generates less heat and appropriate positioning of the circuits that switch more frequently, and design tool exist that will weight such circuits accordingly.

Design tools also allow for state assignment. State machines have a finite number of states, something common to all controllers. When an input comes in, they change state, but the way they move from one state to another has power implications and can therefore also influence heat dissipation.

Clever circuit design can block switching altogether in parts of the circuit thus meaning that the voltage to those areas can be switched off when they are inactive. Power is only applied when it is needed.

Also, over the years, the voltage of integrated circuits has been steadily falling, helping reduce power consumption. But this has followed shrinking geometries, and that has brought its own power dissipation problems. With 90 and 65nm circuits, silicon dioxide has been used as the insulator. But at 45nm and below there can be significant leakage currents that cause power dissipation not related to the useful operation of the circuit. This has been mostly solved with high K dielectrics using hafnium oxide.

Designers are also looking at different transistor architectures to help control the leakage through the channels themselves, such as putting channels on each side, the so-called dual-gate approach. This adds process complexity but does reduce leakage current.

“This is in the research labs at the moment, but will be used eventually,” said Wild. “It could be four to five years before we see it in real products.”

Another problem is parasitic capacitance on the connecting layers.
Surrounding the transistors in a vacuum is obviously not practical but using a porous material can help significantly. These are known as low K dielectrics.
“We are in the third generation of low K dielectrics,” said Wild, “and we keep pushing this lower to reduce the current during the active switching of the circuit.”

Power dissipation
How this all relates to final applications is patchy. For laptops and desktop PCs, Wild believes that there is no reason for a power dissipation today of more than 30W.

“But some say 100W should be acceptable,” he said, “but that is more than the user should have to tolerate.”
But at the forefront of power dissipation technology are portable products. Though the power consumption of these products has reduced dramatically over the years, they are becoming more sophisticated with more processing power, meaning designers have to continue fighting the battle. And even with the best technologies, there still will be heat that needs removing. One new technology being used is a heat pipe.

A heat pipe is a tube full of liquid. At one end, the liquid is closely thermally coupled to the chip and at the other end to a cool area, such as the outside air. The liquid evaporates at the hot end and condenses at the cool end.
“This is very efficient,” said Wild, “and they can be built today using MEMS technology.”

Where space is severely limited, another method is to use the chassis or body of the product itself to remove the heat.

“The heat management requirements in these are very much on the increase,” said Eoin O’Riordan, business unit manager at Chomerics. “They don’t want to design in fans to remove the heat because they are too noisy.”

This has led to a growth in the use of gap filler materials, basically elastomeric sheets filled with a ceramic material.

This can be aluminium oxide or other hybrids that provide a thermally conductive but not electrically conductive layer. The sheets are typically between 0.5 and 10mm thick, depending on how much room there is in the chassis.

These sheets can be moulded to the design of the equipment, an advantage over most heatsinks. According to O’Riordan, “the gap filler material forms a thermal path to remove the heat. It will fit, say, between the component and the frame of the TV.” This type of material can also be used in harsher environments, such as automotive and telecoms, though here a thermally conductive paste – it looks a bit like toothpaste – is also a common method for removing heat.

Informed design
Designers of circuits are becoming more aware of heat management.

“Designers are becoming much more informed when it comes to thermal management,” said O’Riordan. “Once, we were brought in at the last minute if they had a heat problem. Now they look at it early and spread hotspots across the board. It is much higher on the priorities. With power densities going up, it has become an issue that people need to consider.”

This also needs to be coupled with an understanding of the environment in which the product will be used. The design will be influenced by the temperature or pressure outside the enclosure.

“The thermal design engineer should delve into these details early on and do all preliminary and subsequent modelling with these possible effects in mind,” said Jo Proulx, an engineer with Flomerics. “By including exterior environment factors from the beginning, modelling and empirical testing need only be done once.”

 



 
Copyright © ChipTest, All Rights Reserved | Disclaimer
Designed & Developed by Cherry
Home