Microcontroller based avionics

Crap off the shelf (in some cases!)

I think a lot of “consumer IT” gear today really is crap, but appears statistically very reliable in normal usage because

• electronics which doesn’t get warm almost never fails
• most of it is so cheap that if it fails nobody reports it
• the tech is fast moving so people replace it after a short time
• customer service is usually crap so even if you want to get a resolution you hit a brick wall
• the stuff does have to have a reasonable factory QA otherwise (among other factors) the internet will ensure your name stinks
• at the top end of the volume spectrum, there is a high level of consistency via the use of multi-million $ SMT assembly machines (and diligent slaves)

So the vendor can truthfully boast a fantastic reliability record

I believe a famous British “electronics boffin / enterpreneur” (who even got a knighthood) perfected this principle decades ago

The danger is extrapolating the lack of failure data (as reported by the manufacturer ) to an application where the above factors don’t apply!

Andreas wrote:

highly safety-critical systems like a flight control computer.

In GA this means the pilots brain The only thing I can think that maybe could be level A/B? is a FADEC systems for single engine. Besides Austro and Thielert, those are only found for the non certified market. I don’t think they use straight off the shelf consumer type components though. They could use parts for the automobile industry, but only the “racing” quality parts, probably because the people making those systems for non certified GA comes from the racing industry.

Peter wrote:

I believe a famous British “electronics boffin / enterpreneur” (who even got a knighthood) perfected this principle decades ago

Clive Sinclair?

Speaking of reliability, I strongly suspect we have finally arrived at the point where triple modular redundancy not only becomes affordable even for inexpensive GA hardware, but may actually be the least expensive way to go for higher DALs.

Ultranomad wrote:

I strongly suspect we have finally arrived at the point where triple modular redundancy not only becomes affordable even for inexpensive GA hardware, but may actually be the least expensive way to go for higher DALs.

When I worked on the F-16 in the 80s it had triple redundancy flight computer. Today it has 5 touple (or so I have heard), and the computers run 100 times as fast and do a whole lot more. Computers is only one thing, and the easy part, you also need hydraulics and actuators to move the surfaces. These are double redundant on the F-16, and these things are definitely not cheap, and are not likely to become cheaper as time goes by.

I don’t think fbw kind of things will ever enter GA. Scaled down to model airplane and drone size, no problem. A drone configuration, fixed pitch multi-propeller scaled up to real size could probably work, because you only need a computer and electric motors.

Peter wrote:

I believe a famous British “electronics boffin / enterpreneur” (who even got a knighthood) perfected this principle decades ago

The electronics in his computers was pretty reliable. I have multiple old Sinclair machines that still work just fine. Electronic failures on those were usually due to use abuse (and that 4116 memory required 3 voltages, and the lack of the -5V rail would kill them pretty much instantly, so anyone shorting the wrong pins on the edge connector even briefly would cause the memory to fail).

I recently worked on an ARM based SoC, it’s cheap (about £4 per chip in quantity), powerful enough to run Linux (has an MMU and supervisor mode, runs at 400MHz), and needs minimal external components (an external SDRAM if you need more memory). Will natively drive an LCD, provides up to 32 bits per pixel display depth, and a bunch of GPIOs. I’d have to imagine it would be enough to drive many kinds of avionics. Also it’s a TQFP not a BGA so you can fix a device using it with suitable SMD rework hand tools.

The data"sheet" is however over 1000 pages long!

I think you could build almost any major avionics item with a 100 quid evaluation kit with a big LCD

And an awful lot of homebuilt stuff exhibited around Friedrichshafen looks like it is done that way. Plus a bought-in AHRS module.

The graphics is done with a graphics library.

An EFIS would be a fun project. One day when I retire I will get back into software

TSO would be a lot more tricky because of the environmental requirements. I have seen conformal coatings used in some cases, but mostly not. OTOH most currently flying avionics was done with leaded solder which doesn’t have the famous “whisker problem” which reportedly gets worse with humidity.

And practical reliability in unpressurised planes parked outdoors is another factor (I am not sure we have yet seen how the G1000 lasts in this situation). But the product will always be out of warranty by then

Peter wrote:

TSO would be a lot more tricky because of the environmental requirements.

Cabin installed avionics (i.e. everything in the light GA context) is seldom tested to much more than a temperature range of -20°C…55°C, which is much less demanding than even the typical industrial electronics spec of -40°C…85°C.

tomjnx wrote:

Cabin installed avionics (i.e. everything in the light GA context) is seldom tested to much more than a temperature range of -20°C…55°C

Try powering up the G1000 in -20 deg C. If you are lucky it will come to life in an hour or two.

At which point do the displays need heaters?

The coldest I have started in was about -6C and my KLN94/KMD550 were very marginal then. Took maybe 10 mins to be readable.