>> Because usually it's not worth the effort preventing them getting out? 
>> It's not because it's impossible.  Compare software released for 
>> home/office use with software released for running cars, or factories, or 
>> power stations, or planes, or space ships.
>>
>
> You don't think there's bugs in these? ;)

Of course there are, but far less than in desktop software.  My point was 
that depending how much effort you want to put in to it, you either end up 
with software like Microsoft, software like Nasa, or somewhere inbetween.

> I read an article somewhere that before any program change is made, there 
> are meetings upon meetings.

I think that happens for a lot of software development, not just in space 
shuttles.

> I'm sure there are glitches in the firmware of cars, you probably don't 
> notice, or think its something mechanically wrong.

It's extremely unusual though, almost unheard of, for there to be any 
noticeable glitch in the core software running a car.  It's simply because 
the consequences are so great, that the effort is put in to ensure that no 
glitches ever can occur.  Especially nowadays with all the stability, engine 
and other control stuff run by software, it would be a total disaster for 
any company if their cars suddenly did something stupid under certain 
conditions.

> I've seen bearings in our equipment crushed because a drive didn't stop 
> pumping current into the motor when it ran against a hard stop.

I saw a solid steel drive bar approx 10cm diameter sheared in two by a 
motor, it was meant to be driving a machine to compress paper, but it got 
jammed and for some reason the controller didn't stop the motor but kept 
giving it more power.

> Entertaining -- it ramped the current up high enough to fry the motor 
> coils. Oops. I've personally stripped toothed belts bare because of a 
> software error.

But if you knew your software was to control something like a car or plane 
with people in it, you would have course checked and tested your software 
more thoroughly.  That's my original point, ditto with Andrew's order, if 
they really cared enough they would have systems in place to make sure that 
incorrect orders could not be delivered.

> Oh, and most avionics have direct linkage to the controls. The drives 
> AFAIK are only used in autopilot, so that can be disengaged if something 
> goes awry.

Maybe on very small civilian aircraft, but on any passenger carrying jet or 
military jet the control system is totally electronic with no mechanical 
linkages.  I suspect it would be near impossible for a person to manually 
control most planes mechanically, both because of the high forces needed, 
and the lack of an electronic controller to help keep things stable.

But usually there are 4 totally separate controllers, so they do have a high 
degree of redundancy.

> Thankfully, the heavy calculus lifting is done in the drive's firmware all 
> the software has to do is give it the factors found from running a tuning 
> routine, then feed it position, velocity or torque (depending on your 
> goals) It will then use the PID loop to achieve the desired value.

Reminds me of doing PID control loops at University, our lecturer had this 
huge example about 2 metres high with a beefy motor controlling this mass on 
cables around a pulley.  He spent almost the whole lecture explaining the 
theory (where most people had fallen asleep due to the horrendous calculus 
involved), and at the end did the demo.  He had set it up to get from point 
A to point B in the shortest time with something like 5% overshoot. 
Everyone was suddenly paying attention now as he was about to switch it on, 
and then as he flicked the switch, the mass moved extremely quickly to point 
B, past point B, kept accelerating and smashed off the end of the demo. 
Everybody laughed.

Post a reply to this message