Dr. Milind Gandhe, Vice President Semiconductor, Tata Elxsi speaks to Abhishek Mutha and Priya Ravindran from EFY.
Q. What trends have you observed in the semiconductor space over the past few years?
A. There are about two or three trends that have been observed. Moore’s law has existed longer than we thought it would. Production of 14nm chips is doing the rounds and we have people looking beyond that too. Interestingly, the interest in lower nodes is no longer as single-minded as it was ten years ago. There are customers happy with the 130nm technology and they are choosing not to migrate.
Q. Any trends observed in the application side of things?
A. Another trend that is more application-driven is that the amount of digital in silicon is decreasing. We are seeing more of analogue, or to put it rightly – a mixed signal influence. The end markets driving this change are the automobile industry and the Internet of Things.
Q. How has the shift been influenced by the consumer market?
A. The semiconductor industry has seen a clear shift in trends influencing it over time. Computing drove it in the 90s and early 2000s. Since then, till about a year or two ago, phones dominated this industry. The sheer volume of production of phones drove innovation in this space. This trend of phones is slowly going downhill, getting replaced by another fast-rising market, the cars. This is the big change that has happened in the last two to three years. About 10 per cent of the 34-billion dollar semiconductor industry is going into cars.
Q. Comparing to smartphones as a segment, how does the automotive segment stack in terms of revenue?
A. In 2014, let us say you had 82 million cars manufactured in the world. With roughly 34 and a half billion dollars’ worth of semiconductor content going into cars, we can say that, on an average, every car has 420 dollars’ worth of silicon. The luxury cars would easily account for about 1000 to 1500 dollars worth of silicon. So, we see that although the volumes are lower than the phones segment, the price points are much higher. Also, the fact that software content in cars is booming is a big influence.
Q. What factors are influencing the automotive segment becoming the biggest driver for the semiconductor industry?
A. The power train is getting electrified. Even with simple gasoline, there is a significant amount of electronics going into the power train, owing to mileage and emission requirements. There are also a lot of hybrid vehicles coming up, working on the principle of kinetic energy recovery. The second huge trend has been infotainment. The pace at which the infotainment industry is evolving is way faster than the auto industry, the former having a cycle of months and the latter targeting years. The third factor would be the intelligence in the car. Be it safety, navigation, cruise control, or pedestrian detection, more intelligence is being built into the car.
Q. How would you address the issue of functional safety verification?
A. The idea is to design a system such that, even if a transistor fails, the chip will flag it to you. Lock step processing is a rigorous way of testing for safety-critical applications like say, in a pace-maker. Here, you would make two copies of the original and compare the two at every check-point in its life cycle. An error is immediately noticed. Take, for example, the case of a car. A yellow engine light glows indicating that something in the engine is not working right.
Q. How challenging is mixed signal design at chip level?
A. Mixed-signal is not so much of a challenge. The challenge is fundamentally analogue. In digital you can live in a nice idealistic world where you can easily model everything in zeros and ones while in the analogue, unfortunately physics intervenes. Most digital designers I interact with say that they’d lose their job if they didn’t get the silicon right in the first attempt. Most analogue designers would be over the moon if they got their silicon right the first time.
Q. What role do you think SoCs and IoT will play in the future for integration of electronics?
A. For 20 to 30 years, the trend has been to move towards larger chips, with regard to the number of transistors, components, etc. IoT is going to completely buck the trend. The winners in IoT are going to be tiny chips. The other constant winner has been lower and lower power and IoT is going to take that to extreme levels. A good IoT application should not require you to change the battery for a year. We will see very small chips with very high battery life.