Last week I covered Day 1 of CDNLive India. Today it is the turn of verification and PCB/system. With a digression to India's unusual time zone.

Day 2

On Thursday, August 29, there were tracks for Advanced Verification Methodology, Performance/Smart Bug Hunting, Emulation/Prototyping, and PCB Design—basically verification and PCB.

As always, the day started with keynotes. The Technical Keynote was a double act by Oz Levia, giving an overview of what is happening in the world of verification, and Saugat Sen talking about Cadence's move into Electronic System Analysis. I covered the first tool in the space when we announced it in my post Bringing Clarity to System Analysis.

NXP

The guest keynote was Sanjay Gupta of NXP Semiconductors, where he is VP and India Country Manager talking about "Next-Generation Mobility Innovations". He started with well-known "great" predictions, such as Thomas Watson (of IBM) saying that he thought there was a market for maybe five computers. That was in 1943. By 1977, Ken Olsen, the founder of DEC, was saying he saw no reason anyone would ever have a computer at home. Relevant to Sanjay's space, since NXP is #1 in automotive semiconductors, was the President of Michigan Savings Bank advising against investing in Ford (in 1903):

The horse is here to stay but the automobile is just a fad.

He followed that up with a couple of photos I've not seen before. On was New York city in 1900, with horses everywhere and no cars (actually one). By 1913, on the same street, no horses anymore (not even one). It was that fast. As a Ford's Chuck Gray said at Ludwigsburg earlier in the year, "our founder obsoleted the horse, that was a pretty big deal." See my post Ford: Automotive OEM to Software Manufacturer?

Going forward, 90% of the innovation is going to be through electronics and software:

• Connectivity, allowing us to enjoy more the hour or so we spend in a vehicle each day
• Autonomy, saving lives
• Electrification, meeting emission mandates

The idea is not new. These are pictures from 1939 and 1950 showing the "Electronic Highway of the Future." Sanjay's question is whether it will happen in our lifetimes.

I'm going to assume you know about the levels of autonomy that he then ran through. If not, see my post Automotive Industry Basics. Most of R&D is happening at level 3, and most OEMs (remember, that means car companies) are focusing on level 3.

The key technologies to make everything safe and secure are:

• Radar: short-range to high-resolution long-range
• High-performance vision
• Secure V2X: where x is V for vehicle-to-vehicle communication or I for vehicle-to-infrastructure communication, all based on the IEEE 802.11p standard

The sensor package in a L3 vehicle today is 1-3 radars, 1-5 cameras, 0-1 lidar, 4-12 ultrasonic.

The pros and cons of each technology are in the table. Basically, the weaknesses are:

• Radar lacks resolution
• Vision doesn't work in inclement weather (fog, snow, heavy rain)
• Lidar doesn't work in inclement weather (and is currently too expensive for industry adoption)
• Ultrasonic can't classify objects nor does it have good resolution, it is just distance

All autonomous systems contain at least one CPU, typically an Arm processor, along with some of GPU (highly parallel but high power), FPGA (flexible and future-proof, but expensive), DSP (efficient but complex), and machine-learning specialized processor (MLSP, highly efficient, non-standard, tailored to the application). The compute requirements for safety go from supercomputers to do the driving to safety computers to watch out for errors, and to provide backup in the event of failures.

The result of all of this is that how we build cars is going to change. Today, cars are built with small ECUs all over the vehicle, connected with low-bandwidth networks. In the future, there will be a small number of domain computers and controllers that interface to the sensors and actuators.

One of the biggest challenges is the software. The number of lines of code in a car is much more than in anything else, with estimates of 100M lines (about 10X what is in a jet fighter, twice as much as Facebook). Moore's Law flattening is not helping, since processors are not getting faster. Power is the biggest challenge of all. A human brain is 20-40W, an autonomous prototype is more like 400W.

Another challenge is that more automation means more sensors (and actuators). The picture compares a Level-2 Mercedes E-class in 2016, to an Level-3 Audi A8 in 2019, to a Level-4 Renault Espace prototype in 2020. The Espace is not sold in the US, but I happen to know a lot about it since I owned one the whole time I lived in France. It is a minivan that drove very much like a car since its bodywork (at least back then) was all made of fiberglass so its weight and center of mass were both low. The Level-4 Espace is equipped with 4 medium range radar, 1 long range radar, 3 lidar, 4 surround cameras, and 3 frontal cameras (short, medium, and long-range).

Sanjay wrapped up with the key technical challenges:

• Programming these beasts is hard
• Computation performance is the current battlefield
• Power is too high and is the largest missed promise
• Safety and security is possible but not there yet
• Throughput and latency mostly achieved
• Data bandwidth is often a gap between available bandwidth and requirements
• OTA (over-the-air update of software) is possible (indeed Tesla, not to mention every smartphone in the world, has shown it can be done)

Unusual Time Zones

I mentioned in my post about the first day that India has an 11½-hour time difference from California (GMT+5:30). The reason for this unusual time zone on the half-hour is that India is a single time zone, and by nudging it an extra 30 minutes, it makes much of the country be closer to the "correct" time zone where the sun is overhead at noon. Before Independence, India was bigger and so had two time zones, Bombay Time and Calcutta Time, and the current Indian Standard Time is the average of the two, introduced at Independence although only gradually observed. Pakistan is still half an hour behind India, and Bangladesh is still half an hour ahead.

I had heard that Iran is the only other country like this, another big country with just one time zone. It is indeed on a half-hour, but It turns out that a lot more places are on these unusual time zones.

In North America (or just off the coast, depending on how you count islands), Newfoundland is on a half-hour time zone—just the island, not the whole province. Afghanistan, Myanmar, and several islands—Sri Lanka being a big one—are all on the half-hour. Australia is a big multiple-time-zone country (it is roughly the size of the lower 48 US states), but Central Australia Standard Time is on a half-hour (GMT+9:30). To further add to the confusion, Australia has daylight savings, so shifts its clock one hour in summer. But obviously in their summer. However, since they are in the southern hemisphere, that means they come off daylight saving about the same time as the US goes on, meaning the time difference changes by 2 hours.

An even-more-unusual time zone is Nepal, with a time zone 45 minutes off (GMT+5:45). Since Nepal adjoins India, presumably you have to adjust your watch by just 15 minutes when you cross the border.

Also on a 45-minute offset are the Chatham Islands (GMT+12:45), big enough for there to be a Chatham Islands Standard Time, but not big enough that you know where they are...they are out in the Pacific 500 miles east of New Zealand, and a part of that country administratively if not temporally.

But the Breakfast Bytes award for the most unusual time zone of all is Eucla, which is part of  Western Australia but abuts South Australia, and has its own time zone, Australia Western Central Standard Time (GMT+8:45). The population of Eucla is 53 people (2016 census). Not only do they have their own named time zone, but it is also on a 45-minute time difference, and so those 53 people don't share that GMT offset with anyone else.

Sign up for Sunday Brunch, the weekly Breakfast Bytes email