The end of last year saw three leading electronics companies come together in London, England, to lead a debate on the delivery of low power, with particular reference to a multimedia device. 3DLabs, Synopsys and Wolfson were invited to discuss the power issue, giving insight on the issue from an analogue, digital and EDA perspective. Topics included the consumer needs, battery evolutions, delivering multimedia and the shift to multi-core. The following is a summary of the discussion:
Technology has advanced further than anyone at the table could have anticipated. Indeed, as Wolfson’s CTO, Peter Frith stated: “The CD player is the first of the digital consumer products to appear in world market, and I remember when it first came out making the wrong prediction that all CD players might get down to … £100.
“My second epiphany moment was for the GSM phone. The first prototype was on 19-inch racks in the back of a van. I remember strongly thinking ‘they’re never going to make that into handset’. Between then and now I did actually realize that anything that somebody could think of was probably going to be possible in a portable device.”
Battery life
The consumer product is now ubiquitous in multibillion volumes. Everybody has become used to the concept that any multimedia functionality can be picked up and carried around. Making the battery last longer has become vital. But the battery evolution has not kept pace with that of the electronics industry.
We started with nickel cadmium batteries, which could recharge at least 20 times. And then nickel metal hydride gave about 40 recharges. Next, lithium came along and it then got interesting, but nothing has changed in nearly 10 years.
It seems nobody has come up with any new battery. There are rumours of energy storage technologies, such as carbon nanotubes, that are significantly more capable than lithium, but no evidence of such technologies from large companies. Fuel cells are repeatedly talked about, but fuel cells on their own are never a solution. It can only work in parallel with batteries.
Ultimately, the drive for battery life also depends on the device’s cost. Does the iPod last long enough? Does the cell phone last long enough? Will you pay much more to get 10 times the battery life?
Furthermore every generation of device needs to use a physically small battery e.g. the battery in the latest iPod is tiny.
The ability of a battery to store power has doubled but the power consumption of the products that they’re powering has gone down by at least 10 fold. So the guys doing electronics have done phenomenal job.
The awkward part of portable power is size. You can get a bigger battery but it’s become deeply set in the psyche to have a very thin device. If it doesn’t fit in the jean pocket it doesn’t fit the culture. So the consumer now dictates what he wants. ‘I don’t want something fat.’ The goal is a credit card sized piece of technology. Everyone’s used to carrying those around there’s no issue with that.
The difficulty in predicting trends
Fortunately from the industry’s perspective consumers are happy to buy overlapping function products. They’re still happy a phone with a camera and an mp3 player and buy a camera and an mp3 player.
This has probably arisen from there being huge discrepancies between ideals for individual component. Device vs. screen size for example - market research is showing the best screen size is bigger than most of the products. But, you see the kids will accept a reduced screen resolution, or a reduced size because it’s cool.
However the killer applications are often impossible to predict. SMS texting was a gimmick which was simple to put into handsets but no one believed it would be used. Whereas everybody thought video-phoning would be a must have. The Chinese New Year saw approximately 2 billion texts sent. Who could have originally predicted that would happen?
We used to be practical and ask ‘what do we really need?’ Now we ask ‘can it be done?’ and ‘is there a demand?’
The importance in being multimedia
Whatever the case, multimedia is a must. Even if they are rarely used and even if people still buy additional devices, a phone will not sell without having video, camera and mp3 playback.
Therefore, the industry needs to hit the lowest power that can actually be used to convert a form of stored digital information into something that you can perceive as a human being. People need to listen to the files, look at them. Smell is probably out of the question but the more senses activated, the better.
In terms of the sound, the actual energy you need to energize a form of membrane to make sound move and energize your eardrum is tiny; sub microwatts. So, the physical objective at the end of the day is: how low should you be aiming to get it?
Typically, the transducer is a moving coil transducer. A wound wire coil with a magnet and the movement burns power. About 0.1 miliwatts would be burned on average when used to generate the audio sound at the levels usually listened to. The efficiency is poor. But it’s good enough and the battery lasts long enough, so, there’s no pressure at the moment to adopt the much more efficient methods. But the 0.1 miliwatts typical playback converts to 30-miliwatts of electrical power to playback mp3 spread between the digital and analogue domains.
Using the 80:20 rule you focus on the major power drains and on a device, the backlight’s efficiency is just terrible. The analogue and digital power demands are approximately equal but the I/Os all consume power. Any time chips communicate power is consumed.
Small geometries etc has enabled audio chips to run on sub milliwatt power, but half to one miliwatt is being burned from sending clocks and data and its worse still for video data.
Examining the mobile space, power research is focusing on very low voltage swings with differential high-speed serial links, very few signal wires and going differential to remove noise. Voltage swing is now as low as 0.3.
Minimising power consumption between chips
One chip is not possible as you move away from fixed functions. The result would be a huge, compromised chip.
Ultimately, the information transferred between the parts of the system corresponds to certain number of digital toggles. We have to reduce those toggles and the swing of each of the toggles as much as possible.
In the future, the digital products will consist of five components:
Memory
RF
Display, high voltage but remotely located from the system
Digital, as small geometry as possible
Large geometry, high voltage real word chip - analogue
For analogue, signal to noise ratio is approximately 96dBs, to deliver CD quality audio. Theoretically at least, the lowest power way to achieve that signal to noise ratio is to increase the voltage and the signal swings. To halve the noise, you have to quadruple the curve. Doubling the signal swing, quarters the current thus halving the power: P=VI.
In particular, the needs of digital and analogue are extremely different and this makes the SoC environment untenable in reality.
Ultimately, the trend will be for multichip devices rather than SoC, the designers need to examine the most cost effective routes and decide when a functionality makes commercial sense to include.
