When most people think about leading edge chip design, their minds turn to consumer devices, with things such as smartphones, or other devices that are designed to consume entertainment. But hold on: That's not always the case.
It is true that the ubiquitous smartphone is often the biggest and most complex design in the industry, using the latest fabrication technologies and garnering the most attention. These devices are attempting to maximize the levels of integration so that costs can be reduced, product size and weight reduced, interesting form factors explored, battery life extended — the list goes on.
Most certainly, integrated analog will be in play here. There are also technology issues pressuring this trend such as reducing costly off-chip communications which slow a product down and raise the power consumption levels. But it is not always on this “leading” edge that some of the problems have to be tackled. There is actually a much bigger wave that is happening that will require that some of these concepts are taken far beyond the demands of cellphones and add a whole new set of challenges as well.
One example of the kind of challenge I am talking about has been dubbed the Internet of Things (IoT). This term does not really capture the real magnitude and scale of the concept and it can range from the cloud and big-data issues all the way down to remote sensors, or the way in which communicating devices can act more intelligently than they do today. It is those sensors that I am talking about when I talk about some real challenges.
Consider battery life — it is quite possible that these devices will not have a battery or even access to a source of power. They must scavenge power from the environment in which they operate. This means that designs must not just be low-power, but ultra-low power where every picoamp is accounted for. Again, we can see how analog drives solutions in general and integrated analog drives solutions in particular.
Another dimension is that the devices must be very low cost. Think about the introduction of RFID tagging. It wasn't until the cost of these devices came down to single digit cents per device that it appears as if adoption has started to explode. Even my local library now uses this technology in all of their books and it makes checkout a breeze. Some people are talking about the 1 cent devices and at that point I doubt there is little from stopping every store from putting them on all of their products. That will bring about all kinds of benefits for the consumer and for the store. This kind of aggressive price reduction takes just as much cutting-edge technology as the creation of a cellphone.
Yet a third area where significant changes are likely to happen is security. We treat it with a cursory glance today and just hope and trust that people will not try and steal our data, our personal information, or attack our devices. Can the Smart Grid leave so much to trust? I don't think so. Just think of stuxnet — a worm designed to corrupt data that would lead to an industrial device going out of control.
So what is my point? I think there are many challenges that we, as an industry, are going to have to face. In the past we have been sloppy and that has almost been ingrained into our thinking. It doesn't matter if we over-engineer something a little as long as it gets us to market faster, or who cares if we have a little more computational power than we need, or if the software has been optimized as well as it could be. We are now going to have to find ways to tighten up on many aspects of the design flow.
Can we afford the extra accuracy; can we justify the safety margins? Sometimes the answer will be yes, other times an emphatic no. I wonder how much tools will have to change as well as the mentality of using them. How do we understand the operation of a component or a circuit more than just accepting the current limits? What about the analog/digital tradeoffs? Does the needle begin to shift back towards favoring analog, but perhaps only if analog can tighten its belt as well. I think that this is going to be one of the major drivers of design methodologies and EDA tools in the near future, but I would be interested in hearing your views as well.