Should the Government Adopt Commercial Best Practice?

There is something called Betteridge's Law of Headlines that if a headline or title asks a question, then the answer is "no". I think the title of today's blog post, however, is a definitive "yes". I'm thinking about designing the electronics in something like the Joint Strike Fighter, not buying PCs for the DMV.

One problem that the government has is that they don't buy much. The Joint Strike Fighter is projected to create 2,400 planes among all the participant countries. That's roughly the number of iPhones shipped every five minutes. Linton Salmon, then a DARPA program manager, but previously having worked for Texas Instruments in mobile, said in a keynote I attended:

In DoD, I've yet to see a total production that is more than the samples that we would send out in the commercial world.

The reality in most government programs is that all the cost is in design, not production. This makes it all the more important to use learn from the commercial world, which regularly designs the most advanced SoCs in the world and delivers them functioning correctly in time to hit crucial market windows. What we have with government is exponential increasing costs but only linear growth in capability, and not achieving the same power, weight, and cost reduction that we see in the commercial world. In Norman Ralph Augustine’s tongue-in-cheek book Augustine’s Laws, one of the laws states:

Law Number XIV: After the year 2015, there will be no airplane crashes. There will be no takeoffs either, because electronics will occupy 100 percent of every airplane's weight.

This book was written in 1984, nearly 40 years ago, so this is by no means a new problem.

Cadence's James Chew recently contributed an opinion piece How To Break Exponential Pentagon Cost Growth to Aviation Week. James chairs the National Defense Industrial Association’s Science and Engineering Technology Division and is a global group director for Cadence. 

Two key paragraphs are:

For example, the commercial electronics industry has enabled electronic systems companies to develop high-quality, sustainable and modernizable products on a “can’t-miss Christmas” schedule. Much of the industry’s success is due, in large part, to an adherence to “first-pass success” and the computational software tools and processes that enable it. These tools and processes have been developed by companies that invest significant portions of their annual sales—some up to 40%—into research and development (that’s IR&D to you in the DoD), and are a result of the intense competition within the unforgiving consumer electronics market. These tools and processes, which have institutionalized the product development practice of “emulate before you fabricate”, make up the foundation of on-schedule, on-cost product development.

The best-case scenario is that the current DoD and defense industry electronic development process matches up with the commercial electronics development process, where they both seek to achieve “first-pass success”. Even if all things were equal, which they aren’t, the commercial timeline would still be around 30% that of the defense timeline. Eliminating the need for prototype hardware, and the various tests and reworks are associated with it, is a major reduction in design time and cost.

Aerospace programs are usually very long-lasting and sometimes it gets pointed out that a program like the Joint Strike Fighter is nothing like a smartphone. That is true. A smartphone generally is not life-critical and only needs to last a few years. But one area that is more like aerospace is the automotive market. As advanced driver assistance systems (ADAS) have become prevalent, vehicles required much more advanced electronics than was needed a decade or so ago. This has meant that the whole semiconductor ecosystem, including EDA and IP where Cadence operates, has had to learn how to create products with traditional automotive reliability and lifetimes, but in some of the most advanced semiconductor processes available. ADAS requires cameras and radar, high-performance networking, high-performance artificial-intelligence-based processing, and so on. An automotive platform like a latest model car, is much more like an aerospace product. There is also an enormous software component in such a vehicle, up to 100M lines of code—in fact it is not much of an exaggeration to say that much of automotive electronics is simply to provide a high-performance platform to run the software. A chip is only correct if it runs that software correctly, which requires emulation like Palladium and FPGA prototyping like Protium to ensure that the chips are correct when they are finally fabricated.

The government has recognized this as a big issue. Another quote from James' piece:

The fiscal 2017 National Defense Authorization Act, reinforced by the fiscal 2021 Defense Appropriations Act, has an entire section on transitioning to commercial electronics best practices. Program offices and some individuals within the defense industrial base are seeking to better understand the commercial industry-proven way to design electronics that reduce design schedules by at least 70%, producing “first-pass success” electronic system designs that are immediately sustainable and agilely modernizable.

James's piece can be found on the Aviation Week website. It seems to be accessible without a subscription.

Learn More

See the Cadence white paper Meeting the Challenges of the National Defense Strategy also by James (along with Steve Carlson and Charlie Schadewitz) 

Or attend CadenceCONNECT Mission Critical on October 13. Click on the image below for details including a link to register.

Sign up for Sunday Brunch, the weekly Breakfast Bytes email.