Science Spin January 2009

Saving the world, one watt at a time

By Sean Duke

One of the biggest issues facing the world today is the issue of energy. How do we meet our ever growing energy demands as oil and gas supplies dwindle? There is a lot of talk about alternative energy, such as wind and wave power. But, there is another way of looking at the problem. If we could become more efficient in our use of energy, then our energy demands would not be as big - achieving more with less.

This brings us into the world of electronics. In this consumer world, we all have lots of portable electronics in our possession, such as laptops, and mobile phones. At any time we might have two phones charging, and a laptop in waiting mode in the corner. Think of every household in every developed nation doing something similar, and we are talking about massive energy usage. If the energy requirements for portable devices could be reduced, it would represent huge energy savings worldwide.

The use of power within portable devices is highly inefficient at present, with 50 per cent loses of energy as it flows from the battery to the various components. Tyndall National Institute researchers, lead by Cian O'Mathuna, Head of Microsystems, have developed a way of increasing the devices' energy efficiencies by 30 to 35 per cent.

Power

One of the great issues with electronics is the requirement to deliver power to the various electronic components. The delivery of power requires the regulation of voltage from the battery to the integrated circuits. In mobile phones, for example, it is increasingly the case that there are different voltages for different circuits, as the phones become more complex and offer more features. The RF (radio frequency) chips, needed for wireless communication between phones, will have a different voltage from the microprocessor chips, or from the phone displays, for instance.

The current method of controlling power delivery from the battery to the circuits within an electronic device is through the use of what's called 'linear regulators'. These are very cheap, which is one reason they are widely used for this task, but they are highly inefficient, with about 50 per cent efficiency - which means that 50 per cent of the power is lost by these regulators. That represents significant losses.

There is a better way of delivering power, and this involves the use of a higher efficiency power converter called a 'switch mode power supply'. The use of this method will reduce losses and extend the life of a battery significantly, as less power will be drawn from it over any given period of time. The typical switch-mode power supply would have power efficiency of up to 80 or 90 per cent. This means that just 10 or 20 per cent of the power is being lost - a big improvement on linear regulators.

So far so good, but there is a catch with switch mode power supplies - they require more circuitry and more components, and this means they can be bulky. In the context of a world where components are being constantly miniaturised, bulky is not good. One of the bulkiest of all the components are the inductors and capacitors - used for filtering power and energy storage. These must fit around the switch mode power supplies. The inductors are wire-wound components, and quite large, for example.

Challenge

The challenge that Cian O'Mathuna, and his team set themselves, starting about 10 years ago, was to miniaturise the inductors and capacitors, to allow switch mode power to be used in smaller electronic components. It was a huge challenge, and many felt it couldn't be done.

The inductors and capacitors are so-called 'passive components' and they come with associated magnetics. For the frequencies at which power needed to be delivered it was going to be necessary to have a magnetic material around the coil of the inductor. This would allow for a 'smaller footprint, while also containing the electro-magnetic fields that result from having a high current travelling through that inductor.

What the team at Tyndall have managed to do, said Cian O'Mathuna, is to miniaturise the inductor so that it has a similar 'footprint', or area, to the silicon control circuitry, or the silicon switches. This was a big advance as it meant that the inductor could be assembled on top of the silicon. The power source could be integrated into the chip.

Cian said that Tyndall researchers, in the last year, have managed to demonstrate power efficiencies up around 80 to 85 per cent, with a roadmap to achieve efficiencies up to 90 to 95 per cent. That represents massive power savings as compared to linear regulators, and at the higher level, it means just 5 per cent of power is being lost.

The name of the game is to demonstrate high power efficiency, and a small footprint, and Tyndall has managed to achieve that through the expertise of a talented multi-disciplinary team. This more efficient way of delivering power - delivering it locally at source on the chip to exactly where it's required - is naturally of interest to companies like Intel that are focussed on microchip processing. If power can be provided to their chips more cheaply and effectively they are certainly interested.

Processing

The processing of energy is a big problem for Intel and others that are interested in making things smaller and more powerful from a processing point of view. Think of a laptop that heats up after a few hours of usage. That is due to the heat being lost from the processing going on in the microchips. The generation of chips that can process more powerfully must go hand in hand with an ability to remove heat efficiently from the processing process, or otherwise laptops and mobile phones will become red hot.

A way must be found to dissipate energy from increasingly powerful processing activities, or a way must be found to reduce the energy inputs into the process. In a mobile phone there is an RF chip for the wireless communications, a processor for processing the data and a graphics chip. Each has a local power requirement.

Cian O'Mathuna states that work at Tyndall, which provides for locally embedded switch mode power supplies - this is what's called 'power on a chip' - are far more energy efficient than the 'traditional' linear regulators, have a small footprint and can be manufactured, in conjunction with industry partners, in a cost-effective manner.

Greater control of all the power within a device would mean far better management of its energy usage. The parts of the mobile phone that are not required at any point in time could be shut down, or put to sleep for a while, for example, thus saving energy.

This approach to energy would also be well suited to current developments in the microprocessor industry. Companies like Intel are looking for ways to make computers work faster, and to do this they are talking about developing 'multi-core' processors. That means that instead of having one processor on a microchip there will be several, all performing different tasks, or perhaps working in parallel.

Future

A chip of the future might have an array of tens of 'cores' - each core being a processor. This will make everything go fast, but it will be necessary too to control the power delivery to each individual core. The technology developed at Tyndall should enable this to happen, by having a system that processes power locally each core, making them more energy efficient. As well as saving energy, the reduction in heat losses is crucial with more 'cores' to prevent devices overheating.

For Intel, the big question is: if we move to a multi-core processor how can we deliver power effectively to each of the cores. The Tyndall researchers, with their ability to put magnetics directly onto silicon - enabling power supply on a chip - believe that they have the answers a way around this 'technology roadblock'.

The Tyndall researchers have shown that they can get the passive components, the inductors and capacitors, onto silicon and that one piece of silicon could have its own power supply. Going a step further, the vision is to take the discrete power supply on a chip, and make it part of each microprocessor, or 'core' - of which each chip will have many in the future. The power supply would sit on top of each microprocessor, embedded within it. So, each core then would have its own power supply. This is the challenge for Tyndall and others operating in the same research 'space' in future.

Industry

For micro-processor oriented companies like Intel and IBM - who are primarily interested in information processing - power is a pain. What they want, said Cian O'Mathuna is to have power delivered without taking up space, costing nothing and lasting forever. That might sound impossible, but with the power supply on a chip, or the magnetics on silicon idea, he said, the technology takes up very little space. Silicon its enabling the technology to last a long time if not forever. On cost, the technology is getting cheaper because of the batch processing nature of silicon. It is also possible to embed the magnetics with the silicon circuitry, thus avoiding extra cost in the assembly of the inductor or the capacitor next to the silicon in a single package.

At a recent conference about 'power on a chip' Tyndall was overwhelmed with interest - expecting perhaps 40 people, but getting 120. This included representatives from Intel, Qualcom and ON Semiconductor. Tyndall is having ongoing discussions with about four companies, one of which is evaluating the technology to see whether it can be industrialised. This interest from multi-nationals is good for Tyndall and also good for Ireland Inc., said Cian. Ideally, Tyndall would like industry partners to set up in Ireland, with industry suites available at its Cork HQ.

A big issue for mobile phone manufacturers, is how can we make the battery in the phone last longer so you don't have to charge it as much? Laptops too, are constrained by their batteries, which last typically about 3 hours before they need recharging.

"If there was better control over the different functions within these mobile, portable electronic systems we would be able to enhance the efficiency of the power delivery and, therefore, enhance the lifetime of the battery," said Cian O'Mathuna. "All these big companies are talking about how would they save megawatts, or how would they generate megawatts from wind farms If you even think of all the chargers we have now in the house for phones, Nintendos - nobody plugs those out. All the stuff that's on standby, it's a waste of energy."

This is where Tyndall comes in, saving the world, one watt at a time.

PANEL:

Switch mode

Switch mode power supply is a circuit for efficiently converting one voltage level to another. In a typical electronic product such as a mobile phone, the power source is a battery which may have a voltage of 3.6V. Many of the phone subsystems such a processor, camera, or screen, require different voltage levels for efficient operation. Hence power or voltage converters are required in order to change the battery voltage to the required sub-system voltage.

The switch mode power supply works by placing a switch followed by a filter or smoothing circuit, between the input DC voltage level and the output. The switch rapidly turns on and off, effectively connecting and disconnecting input from output and chopping the DC voltage into a square wave voltage. The rate at which the switch turns on and off can be up to several million times in a second, and is called the switching frequency. The switching period is therefore a fraction of a micro-second (one millionths of a second). Over a longer time period, the filter smoothes this chopped voltage into a DC voltage again, which is effectively the average of the chopped voltage.

The output DC voltage level is therefore controlled by the fraction of the switching period for which the switch is on or off. Thus with the switch on for the entire period the output voltage would be equal to the input, or if the switch is on half the period and off half the period then the output voltage would be half of the input voltage.

The filter or smoothing circuit consists of an inductor and a capacitor. These components store energy during the switch on period and supply it to the output during the switch off period. In this way, even though the switch periodically disconnects the input from the output, the electronics being supplied by the output still get a continuous and regulated energy supply.