8
$\begingroup$

The development of computer flight systems is characterised by lengthy certification processes, exceptionally rigorous testing regimes and a default position of extreme caution.

A change in any component at any layer in the hardware/software stack could introduce new incorrect behaviour. For example, microprocessors are so complex that different manufacturing revisions of the same model can introduce bugs.

As a result, these systems are very tightly specified at all levels and very slow to change, so that many systems still in production use what in most other contexts would be considered obsolete technology.

CPUs such as the 68040 and 80486 (both introduced around 1990) were designed into systems a quarter of a century ago that are still being produced today.

Summary of the problem

The pace of development of the electronics industry has accelerated hugely since the first generation of computerised flight control systems were introduced in the early 1970s, and is becoming increasingly incompatible with the needs of the aviation industry. It means that for example:

  • Devices now have manufacturing lifetimes of just a few years (compared with several times that a few decades ago).

  • Devices are now intended to have much shorter service lives than they used to (by orders of magnitude).

  • Increasingly tiny widths in microprocessors mean they succumb increasingly early to problems such as oxide breakdown.

Aerospace represents a very small portion of the manufacturers' business: their main customers make consumer devices, and have utterly different demands on the electronics industry from those of customers in aviation.

Research into the issue

These and other issues are outlined in some detail in Microelectronics Reliability: Physics-of-Failure Based Modeling and Lifetime Evaluation (and no doubt elsewhere - this was just one I found while trying to answer my own question).

The Aerospace Vehicle Systems Institute (AVSI) Consortium Project #17 Methods to Account for Accelerated Semiconductor Device Wear Out has attempted to quantify these risks.

The question

What can the aerospace industry do - or what is it already doing - to mitigate the problems and risks this poses?

$\endgroup$
  • 3
    $\begingroup$ Can you cite the use of 68040 and 80484? Are you talking about simply "hobbyist" avionics such as handheld GPS or commercial aviation? What are the problem and risks that you see? Why does avionics need to move more quickly? If it does the job and does it well, why change? The drivers for aviation are very different than for consumer goods. I think you would need to cite concrete problems and risks in order to start to get answers. $\endgroup$ – Simon Jun 9 '16 at 4:28
  • 4
    $\begingroup$ You don't really point out any problems here, just that the technology is older. Are you talking about availability of parts or what. Seems like it's not really a problem that aviation specific items have to be made to higher tolerances. That's why they cost more. $\endgroup$ – TomMcW Jun 9 '16 at 6:14
  • 2
    $\begingroup$ Aerospace isn't the only user of these parts: automotive, medical, petrochem, and other industries also need hardened, certified microprocessors. Like Tom, I'm not seeing any "problems and risks" here. Consumer electronics is just a different industry. $\endgroup$ – Dan Hulme Jun 9 '16 at 8:30
  • 2
    $\begingroup$ The problem is not that aerospace uses older technology, nor that it should change this. The problem is that trends in the electronics industry are becoming increasingly at odds with those of the aerospace industry. Aviation relies on service and manufacturing lives of devices that run into decades; the electronics industry wants to produce devices with lives measured in months. $\endgroup$ – Daniele Procida Jun 9 '16 at 8:37
  • 2
    $\begingroup$ The chip foundries and the device manufacturers are not the same industries. I believe foundries are happy to produce "simple" chips with modern tooling as the rate of defects is minimal and the profit higher. But there are also more recent circuits developed for avionics. This GPU SoC is based on ARM Cortex A9, which is also used on many smartphones. The i.MX 6 itself is a 40nm-VLSI announced in 2011. $\endgroup$ – mins Jun 9 '16 at 9:33
4
$\begingroup$

The aviation industry is strongly cognisant of these risks.

At least one research programme has attempted to analyse them in depth, the Aerospace Vehicle Systems Institute (AVSI) Consortium Project #17 Methods to Account for Accelerated Semiconductor Device Wear Out.

The project ran for six years until 2007 with a budget of ~US$1.2 million.

However, the project's scope was analysis, development of guidelines, testing methods and so on, and did not include ways of addressing the fundamental problem.

A paper from the project Microelectronics Reliability: Physics-of-Failure Based Modeling and Lifetime Evaluation discusses some of the issues in depth.

| improve this answer | |
$\endgroup$
  • 4
    $\begingroup$ This is only a partial answer to my own question. If anyone has more information about this project or related initiatives, I'd be interested to know more. $\endgroup$ – Daniele Procida Jun 9 '16 at 9:50
2
$\begingroup$

I think you're conflating the PC industry, and its push with Moore's law, with electronics in general.

A huge amount of electronics are used for safety-critical applications. The avionics industry is only a part of it. The automotive industry, in fact, is the leader. Systems like engine computers, electronic power steering, airbag systems, are all long-life safety-critical devices.

Medical devices, industrial systems like robots and elevators, and even appliances like a oven, can cause severe injury and death if they fail. So in many ways, the problem is getting easier, as demand for embedded systems increase.

As an example, you wouldn't use a 80486 or m68k in a safety-critical system. You'd use a chip with an ARM Cortex-R52 core which has specific hardware for lock-step operation, integrated self-test hardware, certification documentation, and more.

Today, there is a big enough market that there are fabrication processes, designs, validation specs, and market availability for long-lived safety-critical devices. Simply search "automotive processor" "SIL CPU" or "automotive grade electronics" and you'll find plenty of systems.

| improve this answer | |
$\endgroup$

Not the answer you're looking for? Browse other questions tagged or ask your own question.