1989worstyear
Topic Author
Posts: 24
Joined: Sun Dec 04, 2016 6:53 pm

The A320-200ceo and its Ancient CPU's

Tue May 02, 2017 4:08 am

Given that nothing has changed on the A320-200 since it's certification back in November of '88 (the same month the USSR launched Buran), where do the A320 FBW computer vendors (Thales etc...) still continue to find the 70's/80's era microprocessors - like the old Motorola 68000 - in new -200's that are rolling off the line nearly 30 years later?

**If you're going to say "oh, there have been 'incremental' improvements", please show me proof (type design drawings, functional schematics, etc..). Thanks!
Stuck at age 15 thanks to the certification date of the A320-200 and my parents' decision to postpone having a kid by 3 years. At least there's Dignitas...
 
VS11
Posts: 1114
Joined: Mon Jul 02, 2001 6:34 am

Re: The A320-200ceo and its Ancient CPU's

Tue May 02, 2017 4:41 am

They don't need to use the same processor. If the processor is backward-compatible to run the same software then there are no issues. In the Intel world, you can run PC games for the 8086 even on Pentium in XP and probably even on 32-bit Windows 7.
 
oOfredOo
Posts: 21
Joined: Mon Oct 03, 2016 9:07 pm

Re: The A320-200ceo and its Ancient CPU's

Tue May 02, 2017 5:28 am

 
User avatar
77west
Posts: 826
Joined: Sat Jun 13, 2009 11:52 am

Re: The A320-200ceo and its Ancient CPU's

Tue May 02, 2017 5:44 am

Aerospace rated chips tend to have longer production lives than our desktop and mobile chips due to certification requirements etc. I doubt they are the exact same units, but probably still pretty old designs. If it aint broke don't fix it...
77West - AW109S - BE90 - JS31 - B1900 - Q300 - ATR72 - DC9-30 - MD80 - B733 - A320 - B738 - A300-B4 - B773 - B77W
 
User avatar
zeke
Posts: 11199
Joined: Thu Dec 14, 2006 1:42 pm

Re: The A320-200ceo and its Ancient CPU's

Tue May 02, 2017 6:30 am

1989worstyear wrote:
**If you're going to say "oh, there have been 'incremental' improvements", please show me proof (type design drawings, functional schematics, etc..). Thanks!


You can simply look at the part numbers, for example on the ELAC these are some different part numbers that have arisen from incremental improvements 394512250, 3945122608, 3945128206, 3945128209, 3945128215. There is also comparability matrices in the maintenance manuals, you cannot intermix new and old part numbers in all circumstances.

FBW is a software application that runs on hardware, it is not done in hardware (eg linux is a software application that can be run on various hardware applications, and various CPU manufacturers, whereas something like a FPGA/ASIC the logic is "fixed" in hardware).. Each software version has its own part number, ie the FBW software application has a part number, and the FBW hardware computer has a part number. Software is loaded via a 3.5" floppy drive.

Airbus continuously upgrades the FBW software, it might do it on the A320, flight test it, and then deploy the software to the A330/A340, or they might update it on the A3340/A340 and then deploy it on the A320 series vv.

This differs from "steam driven" aircraft where the flight control and autopilot was implemented in hardware, the A350 has something which is more like a blade server with a number on applications running on each server.
Last edited by zeke on Tue May 02, 2017 6:32 am, edited 1 time in total.
Human rights lawyers are "ambulance chasers of the very worst kind.'" - Sky News
 
thepinkmachine
Posts: 193
Joined: Tue Apr 28, 2015 4:43 pm

Re: The A320-200ceo and its Ancient CPU's

Tue May 02, 2017 7:09 am

I seem to remember talking to an Airbus Engineer in Toulouse a while ago. His stance was that the latest models of the 320 are a completely different a/c under the hood than the original one.
 
User avatar
SomebodyInTLS
Posts: 710
Joined: Wed Jun 15, 2016 12:31 pm

Re: The A320-200ceo and its Ancient CPU's

Tue May 02, 2017 12:01 pm

VS11 wrote:
They don't need to use the same processor. If the processor is backward-compatible to run the same software then there are no issues. In the Intel world, you can run PC games for the 8086 even on Pentium in XP and probably even on 32-bit Windows 7.


It's my understanding that what you write does not apply at all to aerospace applications. Every chip has the potential for thousands of new bugs to appear.

That's why all hardware must comply to its CERTIFICATION STANDARD - in other words, unless you want to re-certify every new processor (very expensive) you just buy up thousands of spare chips when the production run of your current standard chip is ending...
"As with most things related to aircraft design, it's all about the trade-offs and much more nuanced than A.net likes to make out."
 
VS11
Posts: 1114
Joined: Mon Jul 02, 2001 6:34 am

Re: The A320-200ceo and its Ancient CPU's

Tue May 02, 2017 1:57 pm

SomebodyInTLS wrote:
VS11 wrote:
They don't need to use the same processor. If the processor is backward-compatible to run the same software then there are no issues. In the Intel world, you can run PC games for the 8086 even on Pentium in XP and probably even on 32-bit Windows 7.


It's my understanding that what you write does not apply at all to aerospace applications. Every chip has the potential for thousands of new bugs to appear.

That's why all hardware must comply to its CERTIFICATION STANDARD - in other words, unless you want to re-certify every new processor (very expensive) you just buy up thousands of spare chips when the production run of your current standard chip is ending...


You are probably right - I do not know the actual details for certification. It is probably not worth it to certify each processor generation but every once in a while it makes sense to. Processors do become faster over time so it makes sense for the FBW software to be recompiled/ recertified.
 
kurtverbose
Posts: 458
Joined: Thu Jun 26, 2014 9:33 pm

Re: The A320-200ceo and its Ancient CPU's

Tue May 02, 2017 3:42 pm

zeke wrote:
Software is loaded via a 3.5" floppy drive.


No wonder you have to turn off all your gadgets on a flight - it's because the airlines are jealous of your technology.
 
User avatar
SomebodyInTLS
Posts: 710
Joined: Wed Jun 15, 2016 12:31 pm

Re: The A320-200ceo and its Ancient CPU's

Tue May 02, 2017 4:17 pm

VS11 wrote:
SomebodyInTLS wrote:
VS11 wrote:
They don't need to use the same processor. If the processor is backward-compatible to run the same software then there are no issues. In the Intel world, you can run PC games for the 8086 even on Pentium in XP and probably even on 32-bit Windows 7.


It's my understanding that what you write does not apply at all to aerospace applications. Every chip has the potential for thousands of new bugs to appear.

That's why all hardware must comply to its CERTIFICATION STANDARD - in other words, unless you want to re-certify every new processor (very expensive) you just buy up thousands of spare chips when the production run of your current standard chip is ending...


You are probably right - I do not know the actual details for certification. It is probably not worth it to certify each processor generation but every once in a while it makes sense to. Processors do become faster over time so it makes sense for the FBW software to be recompiled/ recertified.


Unless you're making a design change which will significantly improve the aircraft, it's always cheaper and faster to keep everything the same as much as possible. Especially with critical systems like the fly-by-wire. AFAIK even replacing a simple electronic component with an identical component produced by a different manufacturer will require $$$$, testing, many man-hours and much documentation since you will have to prove to the authorities (through paper trails, analytical proof and physical testing) that the raw materials are up to standard, that the manufacturing process and quality control is up to standard, that the performance of the finished product is up to standard, that its performance under different environmental conditions is within tolerances, that its degradation with age is within tolerances, etc. etc.

I imagine that the certification for a modern processor would be fantastically expensive because there would be so much to test and validate. The natural conservatism of the aerospace industry means that it is normal for aircraft and spacecraft to use flight-critical electronics which are actually several generations out of date even when they're being designed (since they have a proven track record and tend to be more robust than newer equipment).
"As with most things related to aircraft design, it's all about the trade-offs and much more nuanced than A.net likes to make out."
 
User avatar
77west
Posts: 826
Joined: Sat Jun 13, 2009 11:52 am

Re: The A320-200ceo and its Ancient CPU's

Tue May 02, 2017 8:10 pm

An example of the operating system that runs on these "ancient" CPU's is VXWorks https://en.wikipedia.org/wiki/VxWorks

Which is used by the 787, among other aircraft and spacecraft. First release was in the 80's.

Some of the radiation-hardened CPUs for Aerospace use can be over $100,000 for one little, out of date chip. Such is the industry.
77West - AW109S - BE90 - JS31 - B1900 - Q300 - ATR72 - DC9-30 - MD80 - B733 - A320 - B738 - A300-B4 - B773 - B77W
 
WIederling
Posts: 3765
Joined: Sun Sep 13, 2015 2:15 pm

Re: The A320-200ceo and its Ancient CPU's

Wed May 03, 2017 7:39 am

77west wrote:
An example of the operating system that runs on these "ancient" CPU's is VXWorks https://en.wikipedia.org/wiki/VxWorks

Which is used by the 787, among other aircraft and spacecraft. First release was in the 80's.

Some of the radiation-hardened CPUs for Aerospace use can be over $100,000 for one little, out of date chip. Such is the industry.


That's why nonvolatile antifuse programmable FPGA ( like Actel ACT* ) are rather popular in space electronics. ( I've used them on Cassini and Rosetta )

You have the MIL specced RadHard parts and stuff them with your (compiled for the target device ) app specific HDL code.
A big step in that domain. ( The ramcell based devices like the original XILINX line of products require too much periphery to
protect against functional changes from flipping Ram Cells.)
Murphy is an optimist
 
User avatar
Revelation
Posts: 15859
Joined: Wed Feb 09, 2005 9:37 pm

Re: The A320-200ceo and its Ancient CPU's

Wed May 03, 2017 11:53 am

So, does anyone know what CPU family/families was/are used to run the A320 flight control software, and if shortages are an actual issue? It seems so far this thread is based on supposition rather than fact. It could very well be that Airbus has migrated to a newer CPU that is not suffering any shortages.

Google led me to http://www.davi.ws/avionics/TheAvionics ... Cap_12.pdf which says:

Consequently, two types of computers may be distinguished:
2 ELAC (elevator and aileron computers) and 3 SEC (spoiler and elevator computers) on A320/A321
and,
3 FCPC (flight control primary computers) and 2 FCSC (flight control secondary computers) on A330/A340.

Taking the 320 as an example, the ELACs are produced by Thomson-CSF around 68010 microprocessors
and the SECs are produced in cooperation by SFENA/Aerospatiale with a hardware based on the 80186
microprocessor.


Both of those are indeed common embedded processors of the 1980s.

I couldn't find a reference to say if these had been updated since, or if they continue to use them.
Inspiration, move me brightly! Light the song with sense and color.
Hold away despair, more than this I will not ask.
Faced with mysteries dark and vast, statements just seem vain at last.
Some rise, some fall, some climb, to get to Terrapin!
 
User avatar
zeke
Posts: 11199
Joined: Thu Dec 14, 2006 1:42 pm

Re: The A320-200ceo and its Ancient CPU's

Wed May 03, 2017 12:18 pm

Revelation wrote:
I couldn't find a reference to say if these had been updated since, or if they continue to use them.


All I can say with certainty is the part numbers for those various computers are not the same today as it was when the A320 first came off the production line, there has been a number of new models. Even the older models have gone through SB process a number of times due to AD etc.
Human rights lawyers are "ambulance chasers of the very worst kind.'" - Sky News
 
Tristarsteve
Posts: 3420
Joined: Tue Nov 22, 2005 11:04 pm

Re: The A320-200ceo and its Ancient CPU's

Wed May 03, 2017 6:37 pm

As a topical comment.
On the BA B747-400 the FMS was getting old a few years ago. The BA engineers had to actively delete unused way points and airports to get them into the memory. The memory was in kilobytes. There was an active program to split the FMS database in two. One for westbound, and one for eastbound, and try and split the fleet in half.The solution was to adopt the FMS for the B747-8 which was an easy mod. But the certification was delayed and delayed and things got critical in the last six months before the new FMS could be fitted to the -400s.
THere was never any talk of fitting an improved memory, the flight testing would be too expensive. It would be much cheaper to have a team of technicians on the ramp at LHR interchanging east and west databases on the aircraft on the transits.
 
Flighty
Posts: 8676
Joined: Thu Apr 05, 2007 3:07 am

Re: The A320-200ceo and its Ancient CPU's

Thu May 04, 2017 5:55 am

I wrote some comment of pure genius in my own mind that was deleted, but the gist of it was, they probably (should) run 5,000 copies of each of the specialized electronic parts. Imagine if the 787 instruments, designed in the early 2000s, are no longer available from chip fabrication factory. 787 program still has a couple thousand frames to go IMO, and 40 years to go. It is only prudent to spend millions here and there to strengthen the supply chain in these volatile areas. Without a supply of these certified parts, future builds could be severely threatened and maintenance could be impossible.

IMO it makes more sense to just run the factory for the full potential production and close it out, rather than keep a factory open. Probably this is done already and I am just ignorant (which is true here).
 
gloom
Posts: 139
Joined: Thu Jun 30, 2016 4:24 pm

Re: The A320-200ceo and its Ancient CPU's

Thu May 04, 2017 6:57 am

There's one thing against making it "build them all and have a few thousands spares".

All electronics (so, CPUs as well) do suffer from time. There's an ongoing oxidation process, and plastic degeneration as well. So, when we talk 3 or 5 years, it's not a problem. If it's 20 or 30, it could get tough.
I know there are some old Ataris, or Commodores from 80s, and they seem to be running after 35 years, but there's a huge difference between "it can be run", and "it will run troublefree". Especially in case of certified CPUs and apps where lives are at stake.

Cheers,
Adam
 
WIederling
Posts: 3765
Joined: Sun Sep 13, 2015 2:15 pm

Re: The A320-200ceo and its Ancient CPU's

Thu May 04, 2017 8:05 am

gloom wrote:
There's one thing against making it "build them all and have a few thousands spares".

All electronics (so, CPUs as well) do suffer from time. There's an ongoing oxidation process, and plastic degeneration as well. So, when we talk 3 or 5 years, it's not a problem. If it's 20 or 30, it could get tough.


A question of qualified storage. ( dry. moisture creeps into the package/pin interface. If you solder the "wet chips" in an IR or reflow oven they may go "poof". )
The usual way is to go for an expected lifetime procurement when product end of life is announced from the manufacturer.
( first tier customers get notice on this.)
Murphy is an optimist
 
User avatar
flyingturtle
Posts: 5104
Joined: Mon Oct 31, 2011 1:39 pm

Re: The A320-200ceo and its Ancient CPU's

Thu May 04, 2017 8:32 am

zeke wrote:
Software is loaded via a 3.5" floppy drive.


The seat reservations for the German ICE 1 and ICE 2 trains are still loaded with floppies.

https://www.digitaltrends.com/computing ... o-move-on/ for an interesting read on some people who still sell floppies... today.

David
Keeping calm is terrorism against those who want to live in fear.
 
User avatar
Faro
Posts: 1572
Joined: Sun Aug 12, 2007 1:08 am

Re: The A320-200ceo and its Ancient CPU's

Thu May 04, 2017 8:52 am

From this older thread on FMS operating systems the following quote by MD11Engineer in Reply N° 10:

viewtopic.php?f=5&t=761577&p=10993761&hilit=operating+system#p10993761

"I understand that often older processor types are used in aviation and spaceflight applications. Two reasons:
Most computers only take care of one specific task, the free programability and high speed required e.g. for desktop applications is not needed (nobody will play a game or watch a movie on them) and older processors, which have been used for many years are wellknown and have no more hidden problems left."


Faro
The chalice not my son
 
gloom
Posts: 139
Joined: Thu Jun 30, 2016 4:24 pm

Re: The A320-200ceo and its Ancient CPU's

Thu May 04, 2017 9:35 am

WIederling wrote:
A question of qualified storage.


Exactly, and that means extra costs. The only question is whether it costs more than for example keeping a low-intensity production line. Probably not, but who knows. After all, even 6502/6510 are still in production on low-intensity lines :)

Cheers,
Adam
 
User avatar
rjsampson
Posts: 205
Joined: Sat Jul 31, 2010 12:00 am

Re: The A320-200ceo and its Ancient CPU's

Thu May 04, 2017 10:07 pm

To expand on everyone's comments: Processors on the A320 (or any other aircraft) do not need to be advanced. They need to be robust, and they are.

The tasks aircraft computers must perform are rudimentary by our own standards. Best to have as simple, proven, low transistor-count as a microcircuit as you can, to eliminate points of failure, and adhere to strong AIRINC standards (in the US anyway).

The most powerful processors assisting pilots in safe operation of their aircraft today, for the most part, are in the iPads they carry. Since the first CAT-IIIc and FBW logic were implemented in passenger aircraft, there hasn't been a whole lot processing power demanded for an airplane to aviate and nagivate itself, at the command of its pilots. Today's Airbus 320 IFE probably has more processing power than its FMS.

To 77west's comment about VxWorks: He's right. It is an antiquated, ancient embedded Linux operating system, that powers the 787, Boeing's most advanced aircraft. You would be amazed how many things in your life run on VxWorks. The basis for Android was developed on that platform! [Disclosure: Wind River is a client of mine]
"..your eyes will be forever turned skyward, for there.." yeah we know the DaVinci quote. But GA is so dang expensive these days! :(
 
kitplane01
Posts: 512
Joined: Thu Jun 16, 2016 5:58 am

Re: The A320-200ceo and its Ancient CPU's

Fri May 05, 2017 12:40 am

rjsampson wrote:
To 77west's comment about VxWorks: He's right. It is an antiquated, ancient embedded Linux operating system, that powers the 787, Boeing's most advanced aircraft. You would be amazed how many things in your life run on VxWorks. The basis for Android was developed on that platform! [Disclosure: Wind River is a client of mine]



VxWorks is not Linux, nor is it ancient Linux, nor is it based on Linux.
https://en.wikipedia.org/wiki/VxWorks

I don't know what the last sentence means, but Android uses the Linux kernel, and not VxWorks. It has since Android TV, which pre-dates Android 1.0.
http://elinux.org/Android_Kernel_Versions
https://en.wikipedia.org/wiki/Android_(operating_system)

I teach this stuff for a living.
 
gloom
Posts: 139
Joined: Thu Jun 30, 2016 4:24 pm

Re: The A320-200ceo and its Ancient CPU's

Fri May 05, 2017 6:42 am

Linux is generally everything that develops on old kernel built by Linus Thorvalds (hence the name), and Linux predecessor was Minix. Linus did the job to rewrite the kernel.

I'd say VXWorks is more of a Unix style, but the need to be RTM (real time machine) changes quite a lot in internal architecture of the kernel. It reminds me of older pre-PC systems like VAX/VMS where kernel-level processes were RTMs as well.

But it's probably going too far for a.net forum :)

Cheers,
Adam
 
WIederling
Posts: 3765
Joined: Sun Sep 13, 2015 2:15 pm

Re: The A320-200ceo and its Ancient CPU's

Fri May 05, 2017 7:13 am

gloom wrote:
Linux is generally everything that develops on old kernel built by Linus Thorvalds (hence the name), and Linux predecessor was Minix. Linus did the job to rewrite the kernel.


Minix was the host system for the "first" kernel written by Linus Thorvalds. for obvious reasons it supported the original Minix native file system.
That and its functionality as a "unixy" clone scopes its relationship to Minix.

What became Linux is a moving target and as some Microsoft guy noticed long ago "best of breed" in its domain.
I wouldn't be surprised if Microsoft leverages this to step forward .. and away from their own miserable piece of code.
( Raymarine Chart Plotters are Linux based for quite a while now, same for B6G. Haven't looked into Garmin Hardware.)

What VXworks provides is a Real Time framework and that used to be an overlay over a Linux system and not intrinsic.
( .. and base Linux still is not a hard Real Time OS.)
Murphy is an optimist
 
gloom
Posts: 139
Joined: Thu Jun 30, 2016 4:24 pm

Re: The A320-200ceo and its Ancient CPU's

Fri May 05, 2017 7:22 am

Nice we seem to agree, just a quick OT to finish.

WIederling wrote:
What became Linux is a moving target and as some Microsoft guy noticed long ago "best of breed" in its domain.
I wouldn't be surprised if Microsoft leverages this to step forward ..


It actually is. Linux is standard OS for machines created within Azure. I was quite surprised, when I saw that and talked to Microsoft "Linux expert". The end of the world must be near ;)

Cheers, Adam
 
User avatar
Revelation
Posts: 15859
Joined: Wed Feb 09, 2005 9:37 pm

Re: The A320-200ceo and its Ancient CPU's

Fri May 05, 2017 11:53 am

I've been doing unix-based systems since the 80s and met Linus when he was doing the port to the DEC Alpha back in the 90s. The fact that he did the port relatively early in Linux's life cycle allowed a lot of things to be re-designed to work on multiple architectures, which is why over time it has been able to support AMD64, x86-64, and these days, ARM in all its various flavors.

Having said that, I agree with the point of view that a smaller OS focused on real-time is a better approach for flight control.

I also agree that an older, simpler chip is adequate for the tasks being asked in the flight control space.

Newer chips with multiple levels of cache of varying size and speed make it really hard to predict and prove out worst case latency unless the software is relatively simple.

For example. a few years ago I was working on a computer benchmarking/performance team. While going too slow is always a problem, we also had some benchmark runs that went unbelievably fast. One run out of ten or twenty would go 20% faster than the rest, which is a huge difference given the workload. The only factor (presuming no measurement error) that could explain that much of an improvement was ideal cache layout for the one fast run, but unfortunately we could never prove that was the case, and we could never tune our way into getting the ideal layout all of the time. It was very frustrating! It's very hard to get useful cache performance information in real time without disturbing the caching itself (i.e. Heisenberg issues), and it's very hard to then rearrange code and data to get the ideal cache behavior repeatedly, unless your system is far simpler and far more constrained than ours was. In fact Linux itself is too complicated to get such results. There's too many code paths and data structures that have different caching properties and the allocators don't usually produce memory with known cache alignments, etc.

Another factor is developers themselves and the organizations they work in. Most developers are concerned with getting the logic right, and can't be bothered to learn about the performance implications of how they write code. It's hard to find engineers who can and will do both. It's hard to find an organization that will, for instance, reject code being reviewed because it didn't consider how the code change impacted the cache alignment when it added a few fields to a data structure. Most organizations use tools that automatically merge streams of code and these tools at best produce code that compiles cleanly but have no support at all for dealing with the resulting performance impacts. Thus most code development environments just aren't suitable to producing code with hard real time performance requirements.

The main reason Linux is used in embedded applications these days is that the industry now produces CPUs and memory devices with the required speed and capacity at low prices. Once this is true, then the fact that Linux is free and largely maintained by people who work for free or are paid by others makes it an easy choice for an operating system, especially where there are no hard real time requirements. The fact that it works on multiple architectures allows one to avoid paying the prices that Intel has been able to demand for its processors.

The funny thing to consider is that in the late 80s / early 90s a lot of us making a living in the software space were picking up books to learn about Windows NT and the OSI protocol stack. Those were supposed to be the new hot skills to have moving forward. I'm forever thankful that Linux and TCP/IP have grown to the point to be the software of choice in so many roles. I think if Linux didn't come along one of the various BSDs would have filled the role to some degree. One big thing working in Linux's favor was early adoption of true open source approaches and a low barrier to entry (the BSDs tended to be clanish and less welcoming to outsiders).
Inspiration, move me brightly! Light the song with sense and color.
Hold away despair, more than this I will not ask.
Faced with mysteries dark and vast, statements just seem vain at last.
Some rise, some fall, some climb, to get to Terrapin!
 
User avatar
rjsampson
Posts: 205
Joined: Sat Jul 31, 2010 12:00 am

Re: The A320-200ceo and its Ancient CPU's

Fri May 05, 2017 5:33 pm

kitplane01 wrote:

VxWorks is not Linux, nor is it ancient Linux, nor is it based on Linux.
https://en.wikipedia.org/wiki/VxWorks

I don't know what the last sentence means, but Android uses the Linux kernel, and not VxWorks. It has since Android TV, which pre-dates Android 1.0.
http://elinux.org/Android_Kernel_Versions
https://en.wikipedia.org/wiki/Android_(operating_system)


Thanks for pointing out my error. Indeed VxWorks is not based on Linux :banghead: It's a UNIX-like RTOS developed nearly 30 years ago that has been well-supported with updates ever since. Regarding Android, it does indeed use the Linux kernel and (again, thank you for pointing out my error). I suppose I should have indicated that the basis for the hardware architecture in the early days of Android Development were developed and validated using VxWorks and Greenhills. The first production Android device was, and is, based on Linux Kernel.

OK... back to aviation :)
"..your eyes will be forever turned skyward, for there.." yeah we know the DaVinci quote. But GA is so dang expensive these days! :(
 
Chaostheory
Posts: 956
Joined: Wed Mar 06, 2013 2:09 am

Re: The A320-200ceo and its Ancient CPU's

Fri May 05, 2017 7:17 pm

Revelation wrote:
I've been doing unix-based systems since the 80s and met Linus when he was doing the port to the DEC Alpha back in the 90s. The fact that he did the port relatively early in Linux's life cycle allowed a lot of things to be re-designed to work on multiple architectures, which is why over time it has been able to support AMD64, x86-64, and these days, ARM in all its various flavors.

Having said that, I agree with the point of view that a smaller OS focused on real-time is a better approach for flight control.

I also agree that an older, simpler chip is adequate for the tasks being asked in the flight control space.

Newer chips with multiple levels of cache of varying size and speed make it really hard to predict and prove out worst case latency unless the software is relatively simple.

For example. a few years ago I was working on a computer benchmarking/performance team. While going too slow is always a problem, we also had some benchmark runs that went unbelievably fast. One run out of ten or twenty would go 20% faster than the rest, which is a huge difference given the workload. The only factor (presuming no measurement error) that could explain that much of an improvement was ideal cache layout for the one fast run, but unfortunately we could never prove that was the case, and we could never tune our way into getting the ideal layout all of the time. It was very frustrating! It's very hard to get useful cache performance information in real time without disturbing the caching itself (i.e. Heisenberg issues), and it's very hard to then rearrange code and data to get the ideal cache behavior repeatedly, unless your system is far simpler and far more constrained than ours was. In fact Linux itself is too complicated to get such results. There's too many code paths and data structures that have different caching properties and the allocators don't usually produce memory with known cache alignments, etc.

Another factor is developers themselves and the organizations they work in. Most developers are concerned with getting the logic right, and can't be bothered to learn about the performance implications of how they write code. It's hard to find engineers who can and will do both. It's hard to find an organization that will, for instance, reject code being reviewed because it didn't consider how the code change impacted the cache alignment when it added a few fields to a data structure. Most organizations use tools that automatically merge streams of code and these tools at best produce code that compiles cleanly but have no support at all for dealing with the resulting performance impacts. Thus most code development environments just aren't suitable to producing code with hard real time performance requirements.

The main reason Linux is used in embedded applications these days is that the industry now produces CPUs and memory devices with the required speed and capacity at low prices. Once this is true, then the fact that Linux is free and largely maintained by people who work for free or are paid by others makes it an easy choice for an operating system, especially where there are no hard real time requirements. The fact that it works on multiple architectures allows one to avoid paying the prices that Intel has been able to demand for its processors.

The funny thing to consider is that in the late 80s / early 90s a lot of us making a living in the software space were picking up books to learn about Windows NT and the OSI protocol stack. Those were supposed to be the new hot skills to have moving forward. I'm forever thankful that Linux and TCP/IP have grown to the point to be the software of choice in so many roles. I think if Linux didn't come along one of the various BSDs would have filled the role to some degree. One big thing working in Linux's favor was early adoption of true open source approaches and a low barrier to entry (the BSDs tended to be clanish and less welcoming to outsiders).


Fascinating post. Thanks.

Now that I come to think of it, I recall the brains of our old 772's are AMD risc powered. I'm note sure about our new 77Ws.
 
kitplane01
Posts: 512
Joined: Thu Jun 16, 2016 5:58 am

Re: The A320-200ceo and its Ancient CPU's

Sat May 06, 2017 4:03 am

gloom wrote:
Linux is generally everything that develops on old kernel built by Linus Thorvalds (hence the name), and Linux predecessor was Minix. Linus did the job to rewrite the kernel.

I'd say VXWorks is more of a Unix style, but the need to be RTM (real time machine) changes quite a lot in internal architecture of the kernel. It reminds me of older pre-PC systems like VAX/VMS where kernel-level processes were RTMs as well.

But it's probably going too far for a.net forum :)

Cheers,
Adam


Linus's last name is spelled Torvalds.

Linux was not developed from Minux's source code, nor was it designed to replace Minux.
 
WIederling
Posts: 3765
Joined: Sun Sep 13, 2015 2:15 pm

Re: The A320-200ceo and its Ancient CPU's

Sat May 06, 2017 9:27 am

kitplane01 wrote:
Minux.


Minix is not spelled Minux. :-)

Andrew S. Tanenbaum's Minix was written as example code for his "Operating System Design" lectures.
Freely available and designed to run on bog standard x86 PC hardware _and_ "selfhosted".
i.e. you could compile the OS on itself.
Murphy is an optimist
 
kitplane01
Posts: 512
Joined: Thu Jun 16, 2016 5:58 am

Re: The A320-200ceo and its Ancient CPU's

Sat May 06, 2017 4:52 pm

WIederling wrote:
kitplane01 wrote:
Minux.


Minix is not spelled Minux. :-)

Andrew S. Tanenbaum's Minix was written as example code for his "Operating System Design" lectures.
Freely available and designed to run on bog standard x86 PC hardware _and_ "selfhosted".
i.e. you could compile the OS on itself.



Crap. While complaining about spelling I mis-spell something else. How stupid am I. :-(
 
User avatar
lightsaber
Crew
Posts: 12815
Joined: Wed Jan 19, 2005 10:55 pm

Re: The A320-200ceo and its Ancient CPU's

Sat May 06, 2017 6:25 pm

Revelation wrote:
I've been doing unix-based systems since the 80s and met Linus when he was doing the port to the DEC Alpha back in the 90s. The fact that he did the port relatively early in Linux's life cycle allowed a lot of things to be re-designed to work on multiple architectures, which is why over time it has been able to support AMD64, x86-64, and these days, ARM in all its various flavors.

Having said that, I agree with the point of view that a smaller OS focused on real-time is a better approach for flight control.

I also agree that an older, simpler chip is adequate for the tasks being asked in the flight control space.

Newer chips with multiple levels of cache of varying size and speed make it really hard to predict and prove out worst case latency unless the software is relatively simple.

For example. a few years ago I was working on a computer benchmarking/performance team. While going too slow is always a problem, we also had some benchmark runs that went unbelievably fast. One run out of ten or twenty would go 20% faster than the rest, which is a huge difference given the workload. The only factor (presuming no measurement error) that could explain that much of an improvement was ideal cache layout for the one fast run, but unfortunately we could never prove that was the case, and we could never tune our way into getting the ideal layout all of the time. It was very frustrating! It's very hard to get useful cache performance information in real time without disturbing the caching itself (i.e. Heisenberg issues), and it's very hard to then rearrange code and data to get the ideal cache behavior repeatedly, unless your system is far simpler and far more constrained than ours was. In fact Linux itself is too complicated to get such results. There's too many code paths and data structures that have different caching properties and the allocators don't usually produce memory with known cache alignments, etc.

Another factor is developers themselves and the organizations they work in. Most developers are concerned with getting the logic right, and can't be bothered to learn about the performance implications of how they write code. It's hard to find engineers who can and will do both. It's hard to find an organization that will, for instance, reject code being reviewed because it didn't consider how the code change impacted the cache alignment when it added a few fields to a data structure. Most organizations use tools that automatically merge streams of code and these tools at best produce code that compiles cleanly but have no support at all for dealing with the resulting performance impacts. Thus most code development environments just aren't suitable to producing code with hard real time performance requirements.

The main reason Linux is used in embedded applications these days is that the industry now produces CPUs and memory devices with the required speed and capacity at low prices. Once this is true, then the fact that Linux is free and largely maintained by people who work for free or are paid by others makes it an easy choice for an operating system, especially where there are no hard real time requirements. The fact that it works on multiple architectures allows one to avoid paying the prices that Intel has been able to demand for its processors.

The funny thing to consider is that in the late 80s / early 90s a lot of us making a living in the software space were picking up books to learn about Windows NT and the OSI protocol stack. Those were supposed to be the new hot skills to have moving forward. I'm forever thankful that Linux and TCP/IP have grown to the point to be the software of choice in so many roles. I think if Linux didn't come along one of the various BSDs would have filled the role to some degree. One big thing working in Linux's favor was early adoption of true open source approaches and a low barrier to entry (the BSDs tended to be clanish and less welcoming to outsiders).

For keeping the same capability, old processors are fine.

Currently, Motorola built PowerPC 7410s are the industry darling, as before smartphones, that was the last CPU built for really low power with great performance.

Everyone must realize the CPU must perform at peak in a cabin depressurization event (poor cooling) when high demand is expected. The CPU and RAM (not storage) must also be certified for under 50% utilization in case an untested event spikes use.

I'm excited about all the ARM processors hitting the market. They shall boost FBW capabilities. There is logic not being implemented due to CPU limits. But what more I know I am not at liesure to discuss. I'll just note self driving car developments will find there was into FBW systems.

I too am glad for Linux as open source has helped tremendously.

Lightsaber
"They did not know it was impossible, so they did it!" - Mark Twain
 
User avatar
SomebodyInTLS
Posts: 710
Joined: Wed Jun 15, 2016 12:31 pm

Re: The A320-200ceo and its Ancient CPU's

Sat May 06, 2017 6:50 pm

I'd just like to thank OP for creating an unusually technical and interesting thread...
"As with most things related to aircraft design, it's all about the trade-offs and much more nuanced than A.net likes to make out."
 
User avatar
Revelation
Posts: 15859
Joined: Wed Feb 09, 2005 9:37 pm

Re: The A320-200ceo and its Ancient CPU's

Sat May 06, 2017 7:02 pm

lightsaber wrote:
For keeping the same capability, old processors are fine.

Currently, Motorola built PowerPC 7410s are the industry darling, as before smartphones, that was the last CPU built for really low power with great performance.

Everyone must realize the CPU must perform at peak in a cabin depressurization event (poor cooling) when high demand is expected. The CPU and RAM (not storage) must also be certified for under 50% utilization in case an untested event spikes use.

I'm excited about all the ARM processors hitting the market. They shall boost FBW capabilities. There is logic not being implemented due to CPU limits. But what more I know I am not at liesure to discuss. I'll just note self driving car developments will find there was into FBW systems.

I too am glad for Linux as open source has helped tremendously.

Lightsaber

Very interesting. I looked at Wiki's page for the 7410 ( https://en.wikipedia.org/wiki/PowerPC_G4#PowerPC_7410 ) and it says:

The chip added the ability to use all or half of its cache as high-speed, non-cached memory mapped to the processor's physical address space as desired. This feature was used by embedded systems vendors such as Mercury Computer Systems.


This goes along with my statement about how caching can make real time analysis more difficult. It is sometimes more desirable to "pin down" parts of the cache memory so it's behavior becomes predictable.

However, going back to the theme of this thread, this chip design was first used back in 2001 by Apple for the G4 PowerMacs. It probably qualifies as "ancient" in most people's books.

It will be interesting to see which ARM chip(s) end up being preferred. Since ARM, Inc licenses their intellectual property quite broadly, there are a lot of them to choose from.
Inspiration, move me brightly! Light the song with sense and color.
Hold away despair, more than this I will not ask.
Faced with mysteries dark and vast, statements just seem vain at last.
Some rise, some fall, some climb, to get to Terrapin!
 
A320FlyGuy
Posts: 167
Joined: Wed May 16, 2012 8:31 pm

Re: The A320-200ceo and its Ancient CPU's

Sat May 06, 2017 8:21 pm

It's also important to keep in mind that it's not like the FBW system is handling detailed graphic design - at the end of the day, the FBW is using algebraic logic to move sidestick motion into control surface deflection - the fact that the processor is old is irrelevant - it is a solid, reliable and predictable platform and there is so much experience working with it that it would be hideously expensive to try and go back now and start all over.
My other car is an A320-200
 
A320FlyGuy
Posts: 167
Joined: Wed May 16, 2012 8:31 pm

Re: The A320-200ceo and its Ancient CPU's

Sat May 06, 2017 8:28 pm

It's also important to keep in mind that it's not like the FBW system is handling detailed graphic design - at the end of the day, the FBW is using algebraic logic to move sidestick motion into control surface deflection - the fact that the processor is old is irrelevant - it is a solid, reliable and predictable platform and there is so much experience working with it that it would be hideously expensive to try and go back now and start all over.
My other car is an A320-200
 
User avatar
Revelation
Posts: 15859
Joined: Wed Feb 09, 2005 9:37 pm

Re: The A320-200ceo and its Ancient CPU's

Sun May 07, 2017 11:16 am

A320FlyGuy wrote:
It's also important to keep in mind that it's not like the FBW system is handling detailed graphic design - at the end of the day, the FBW is using algebraic logic to move sidestick motion into control surface deflection - the fact that the processor is old is irrelevant - it is a solid, reliable and predictable platform and there is so much experience working with it that it would be hideously expensive to try and go back now and start all over.

Correct. I think the concerns were about the supply of the parts, however as mentioned earlier no one has given any verifiable statement here that shows there actually is a problem with the supply of the parts. It's more of the opposite -- examples were given where microprocessors of the 1980s are still being produced.
Inspiration, move me brightly! Light the song with sense and color.
Hold away despair, more than this I will not ask.
Faced with mysteries dark and vast, statements just seem vain at last.
Some rise, some fall, some climb, to get to Terrapin!
 
User avatar
zeke
Posts: 11199
Joined: Thu Dec 14, 2006 1:42 pm

Re: The A320-200ceo and its Ancient CPU's

Sun May 07, 2017 11:59 am

Eventually I see Airbus adopting technology like the A380/A350 into the product.

The A380 introduced the integrated modular avionics concept to the Airbus FBW aircraft. On the A320 currently there is basically dedicated wiring for each computer, and each computer is its own little box.

On the A380 they have the IMA which runs a MPC755 processor as the CPU, and the I/O is done with a TMS320C33 chip. These are still fairly old chips but it is more than enough power to replace a large number of computers with a single computer running the ARINC 653 RTOS and a number of different software applications within their own partitions.

I believe with the ATR-600 Airbus went with IMA system with ADFX replacing the cable runs and LRUs of the earlier ATRs.

I hear they are working on IMA2G which will probably also coincide with the new electric wing and distributed sensors for the "A320".
Human rights lawyers are "ambulance chasers of the very worst kind.'" - Sky News
 
benbeny
Posts: 171
Joined: Thu Nov 03, 2016 1:44 pm

Re: The A320-200ceo and its Ancient CPU's

Thu May 18, 2017 10:42 am

Apology for resurrecting this thread. In any FBW design, does the computing power build 3D models of aircraft position and so forth in the air and then react accordingly, or they just translate input from pilots into output after being mixed with some predetermined parameters?
And is there any potential of errors yet still undiscovered in those ancient processors?
 
kitplane01
Posts: 512
Joined: Thu Jun 16, 2016 5:58 am

Re: The A320-200ceo and its Ancient CPU's

Thu May 18, 2017 10:57 am

benbeny wrote:
Apology for resurrecting this thread. In any FBW design, does the computing power build 3D models of aircraft position and so forth in the air and then react accordingly, or they just translate input from pilots into output after being mixed with some predetermined parameters?
And is there any potential of errors yet still undiscovered in those ancient processors?



I'm not sure if this answers your question. There has been research into modeling the aircraft aerodynamics in real time,including any damage the airplane might have, and controlling the flight control surfaces from pilot input and that model. I think this is jus research though.
 
benbeny
Posts: 171
Joined: Thu Nov 03, 2016 1:44 pm

Re: The A320-200ceo and its Ancient CPU's

Thu May 18, 2017 11:42 am

kitplane01 wrote:
benbeny wrote:
Apology for resurrecting this thread. In any FBW design, does the computing power build 3D models of aircraft position and so forth in the air and then react accordingly, or they just translate input from pilots into output after being mixed with some predetermined parameters?
And is there any potential of errors yet still undiscovered in those ancient processors?



I'm not sure if this answers your question. There has been research into modeling the aircraft aerodynamics in real time,including any damage the airplane might have, and controlling the flight control surfaces from pilot input and that model. I think this is jus research though.

This is exactly what I thought. I don't know how powerful they are, but I don't think those processors are capable to handle such amounts of calculations needed in real time.
 
User avatar
Revelation
Posts: 15859
Joined: Wed Feb 09, 2005 9:37 pm

Re: The A320-200ceo and its Ancient CPU's

Thu May 18, 2017 11:42 am

benbeny wrote:
Apology for resurrecting this thread. In any FBW design, does the computing power build 3D models of aircraft position and so forth in the air and then react accordingly, or they just translate input from pilots into output after being mixed with some predetermined parameters?

It's definitely the later. The reference I gave above ( http://www.davi.ws/avionics/TheAvionics ... Cap_12.pdf ) is pretty readable and gives a good flavor of what is going on.

benbeny wrote:
And is there any potential of errors yet still undiscovered in those ancient processors?

I'd say that given the amount of testing and the number of years of hard usage the odds of finding new errors is very small.
Inspiration, move me brightly! Light the song with sense and color.
Hold away despair, more than this I will not ask.
Faced with mysteries dark and vast, statements just seem vain at last.
Some rise, some fall, some climb, to get to Terrapin!
 
benbeny
Posts: 171
Joined: Thu Nov 03, 2016 1:44 pm

Re: The A320-200ceo and its Ancient CPU's

Fri May 19, 2017 12:46 pm

Revelation wrote:
benbeny wrote:
Apology for resurrecting this thread. In any FBW design, does the computing power build 3D models of aircraft position and so forth in the air and then react accordingly, or they just translate input from pilots into output after being mixed with some predetermined parameters?

It's definitely the later. The reference I gave above ( http://www.davi.ws/avionics/TheAvionics ... Cap_12.pdf ) is pretty readable and gives a good flavor of what is going on.

benbeny wrote:
And is there any potential of errors yet still undiscovered in those ancient processors?

I'd say that given the amount of testing and the number of years of hard usage the odds of finding new errors is very small.

Thank you for your links, it really opens up my mind.

Who is online

Users browsing this forum: jarheadk5 and 4 guests

Popular Searches On Airliners.net

Top Photos of Last:   24 Hours  •  48 Hours  •  7 Days  •  30 Days  •  180 Days  •  365 Days  •  All Time

Military Aircraft Every type from fighters to helicopters from air forces around the globe

Classic Airliners Props and jets from the good old days

Flight Decks Views from inside the cockpit

Aircraft Cabins Passenger cabin shots showing seat arrangements as well as cargo aircraft interior

Cargo Aircraft Pictures of great freighter aircraft

Government Aircraft Aircraft flying government officials

Helicopters Our large helicopter section. Both military and civil versions

Blimps / Airships Everything from the Goodyear blimp to the Zeppelin

Night Photos Beautiful shots taken while the sun is below the horizon

Accidents Accident, incident and crash related photos

Air to Air Photos taken by airborne photographers of airborne aircraft

Special Paint Schemes Aircraft painted in beautiful and original liveries

Airport Overviews Airport overviews from the air or ground

Tails and Winglets Tail and Winglet closeups with beautiful airline logos