When you read the recent headlines on global chip shortages, does R.E.M. or David Bowie spring to mind? On one hand it could be The End of the World as we Know it. Or maybe it’s just another round of Changes. No this is not a test to see when you started listening to the radio. Rather, this is simply a musing on the current global semiconductor shortage. The situation has gotten to the point that the auto industry has seen a meaningful slowdown.
In the news…
Depending on where you access your news, you will find a variety of opinions. There are those that are still predicting the End of the World as we Know It.
For a slightly more positive spin on the Changes that look for a silver lining.
Personally, I side with the list of optimistic authors. Now, this is a blog that often sings the praises of virtualisation. And you probably expect a plug about now for VLAB solving your silicon scarcity issues. Today I will try to refrain and instead point to what you might see happening across the Automotive industry.
The state of the Automotive industry
Getting back to the pseudo-optimism of R.E.M.’s the End of the World as we Know It. Buried away in many of today’s popular cars there are sometimes hundreds of individual automotive CPUs. You could be mistaken for thinking that most are in the touch screen that now controls your radio. Unified infotainment systems including radio, A/C, navigation, and general driver distraction generally use a single high performance multi-core CPU. In contrast, the power steering, anti-lock brakes, fuel injectors, and traction control generally each have their own low performance CPU. Add to that the head lights, tail lights, emissions controls, power windows, turn signals, heated mirrors, and power seats. Suddenly you have more processors in passenger cars with basic cruise control than in passenger airliners with auto pilot.
This is where Bowie’s Change comes in. Instead of spreading numerous low performance chips, consider using high performance controllers in each area, or zone of the car? Perhaps go even further and consolidate to a very few very high-performance compute CPUs. Take another step and locate CPUs in the centre of the car to reduce cabling. Maybe one CPU per software domain, with I/O gateways in the various zones can also reduce chip count? It all seems blindingly obvious. But as aviation discovered in the early 2000s, this transition involves a large amount of system architecture and software development. Unfortunately articles on how VLAB makes such transitions much easier (see below) weren’t available at the time. But you might be interested…
Motivation for change
Welcome back! Now that you understand that VLAB can solve all the technical challenges you may be scratching your head. “Why isn’t everyone already doing this?” I hear some of you ask. If you did then you probably belong to the engineering fraternity who believes in logic, efficiency, and elegant design. That being the case you probably aren’t a devoted student of the economics of automotive supply chain.
In Aviation, every gram translates into large operating costs. Consequently Boeing and Airbus have a distinct financial incentive in dealing with their suppliers. For example, “All software will be ported to this standards complaint central computer no matter what the development cost”. Automakers have had relative low consumer pressure to invest in high development cost efficiency improvements. This is particularly true when renegotiating contracts with Tier 2 or 3 suppliers who historically choose the ECU on which to build their software. You may have seen proposals for zone controllers and domain controller based automotive networks from Visteon and Aptiv (see here) before. They have been around for years but adoption has been slow.
Winds of change
Enter the chip shortage. Now there is a solid economic reason to consider a change in software architecture. Swap trying to coordinate the delivery of hundreds of scarce chips, for only a dozen or so. This reduces the scope of the problem, but I hear questions of how? Here we must look at the details of chip shortage. This isn’t a case of many totally independent vendors all having manufacturing problems all at once. This is a case of many vendors outsourcing manufacturing to only a few places creating a general bottleneck that affects seemingly unrelated chips.
From the chip manufacturer’s perspective, one chip is more or less equivalent to another from a capacity point of view. This is only slightly an exaggeration. There can be a factor of two or three difference in chip area between high and low performance chips. But if we can reduce the demand per car by an order of magnitude that’s still a win. Going one step further will see a move to a more standardized approach to the processors. Ultimately reducing the number of different chips any given vehicle manufacturer needs, improves the chance of on-time delivery.
Now for the disclaimer… I have completely ignored the argument that there is no way Automotive software can be consolidated as per the Aviation Federated Electronics to IMA migration. The common argument is that the comparison is not a fair one. Internal combustion engines running at a few thousand RPM, up against a jet with compressor stages turning at 10s of thousands of RPMs. Similarly, full Level 5 autonomy for vehicles operating at 10s of MPH in 2 dimensions, cannot be similar to navigation in 3 dimensions at 100s of MPH. Therefore, it is entirely possible that I have been completely wrong in my opinion that this is a solvable problem.
To summarise. The glass half full perspective sees a pandemic initiated chip shortage resulting in full level 5 autonomy. The alternative is one I prefer to leave to another blog. In the mean time join me in kicking back, and listening to some classic rock!