the most reliable low-end microcontrollers

[Simon]

hi guys, I'm being promoted at work. I think I have pissed the works manager of to the point that he wants to unload me to the engineering Department, where they will be very happy to use my electric and electronic skills amateurish as they may be (although the current skills level in that department is lower). I've not done a lot with microcontrollers, and I guess most designs will benefit from them in what I will do (air conditioning control). So maybe it is time to make a new series start with microcontrollers, considering many of our designs go into military vehicles I would want the hardest thing possible without of course going nuclear proof and expensive. Which of the basic commonly available microcontrollers is generally preferred in these environments?what should I start to learn?

[SeanB]

Use no micros, just a simple comparator and hysteresis.

Though you can add more error checking with a micro, almost any one will do at the speed of the control loop. More important is watchdog, good noise immunity ( no 1.8V parts then) and providing a lot of input filtering and overvoltage protection on all inputs and outputs ( including parts that are disconnectable like displays and remote receivers), along with providing a power supply that is capable of surviving all the input will throw at it without failing.

[Simon]

yes I would use Standard electronics where possible. But in some situations a microcontroller would make the design infinitely more flexible and much cheaper. I am just looking at the best of what is out there in the consumer market. At the moment all electronic devices we use are just consumer grade for the automotive market. As this is only air conditioning systems it is not a matter of life or death, I'm just looking at what is known to be the most reliable brand or which is the most not recommended.

[SeanB]

If you think AC is not important you have not sat in a Saracen with a broken AC and with some deeply unhappy people outside asking to come in, and you are in the sun and are thinking that being a lobster in a pot is a lot more desirable.

[Simon]

yes of course and we are already supplying units that have PIC microcontrollers in them. But then of course the supplier is not to know that I looked inside. We have not had any of these actually failed due to the microcontroller over the last three years. I am currently using PIC microcontrollers for the basic things I do, but have heard somebody mention not to use them in things like military. Or perhaps that person was referring to the fact that microchip do not guarantee their parts for military or medical use. This is despite the fact they produce parts for the military temperature range.so basically the question is should I favour Pic, atmel or any others or are they all about as reliable as each other?

[SeanB]

Use whatever you are most comfortable with, and which has the right feature set and the right price point. Most important is to make the firmware as bug free as possible, and to reduce the chances that improper inputs and noise can cause it to malfunction. Range checking and sanity testing are good.

[Simon]

so what you're saying is that most manufacturers are going to be all about the same for reliability. At the moment I'm using PIC microcontrollers, I'm not really sure why and I have no problem with them. I suppose really I need to start learning C now.

[SeanB]

Silicon, mostly from the same suppliers. Wafers made to pretty much the same exacting quality standards and tested multiple times. Assembled into packages by a small number of specialist companies, and to pretty much similar high reliability specs.

Yes, pretty much any big manufacturer will supply a good product, and parts that have been in production for a few years will have a good record for failure rates and reliability.

I would suggest that physically larger parts are more reliable from just having more epoxy keeping moisture out, so the older PIC devices, which also are cheaper, are more likely to be reliable. As you say, you do not need super fast, all bells, just reliable, low cost and usable. If you have space and pins build in simple easy to use test routines, and document them. Works for both production and servicing, very important to think of the end user and the service tech who will either thank you or curse you for that.

[free_electron]

Definately go with AUTOMOTIVE grade parts (-40 to 125) . don't use commercial stuff ( 0 .. 80 degree C ).

Automotive grade parts are tested to more stringent parameters and may have a different internal construction for mechanical stability. Or go with industrial( -55 to 150 ) .

The key in desiging automotive circuitry is the power supply and i/o protection. you need to be able to survive a load-dump ... like when disengaging the starter motor. those are enormous spikes !

your entre design should be made automotive grade parts. from capacitor , resistor all the way to cpu...
it is the ONLY way... al the rest is a kludge..

[Zero999]

The trouble is, if there's any risk of people being killed by your code going wrong or a transistor failing short circuit, you could be in serious trouble.

In a mission/safety critical application an ordinary micro will not do. It needs to fail safe and have a very level high reliability. The maximum acceptable failure rate will depend on factors such as the mission time and the consequences of it going wrong, i.e. severity of injury, loss of life, damage to property etc. The software has to be approved by an external body and there needs to be multiple levels of protection, i.e. a supervisor CPU monitoring the main CPU to check it's behaving and the hardware needs to be designed so a software fault shouldn't be deadly. In fact in a safety critical system the same level of scrutiny also applies to any component from a relay to a switch.

I suggest you read more up on industrial controls and articles by manufactures of safety systems: Piltz, Allen Bradley, Siemens etc.

[vxp036000]

Having worked in defense before, I can tell you that it will be nearly impossible to overhall the design like you're describing without an enormous expense of re-design and qualification testing.  And you'll need to get the customer to buy off on it.  This could take many months, if not years.

hi guys, I'm being promoted at work. I think I have pissed the works manager of to the point that he wants to unload me to the engineering Department, where they will be very happy to use my electric and electronic skills amateurish as they may be (although the current skills level in that department is lower). I've not done a lot with microcontrollers, and I guess most designs will benefit from them in what I will do (air conditioning control). So maybe it is time to make a new series start with microcontrollers, considering many of our designs go into military vehicles I would want the hardest thing possible without of course going nuclear proof and expensive. Which of the basic commonly available microcontrollers is generally preferred in these environments?what should I start to learn?

[Psi]

In some situations you might be able to implement a system with 3 micros. (Like how airplanes have 3 flight computers).

All micros run the same code and have their own dedicated powersupply.
The system inputs connect to all 3 micros and the outputs are compared together with external logic to check for faults. etc.
This way the system can stay running without interruption if one micro dies.

It's overkill for most projects.
but the hardware cost isn't huge and it does provide a good level of protection.

Just a crazy idea.

[IanB]

...military...

With military contracts there will be very stringent and precise specifications ahead of time. If your company already has an military contracts these specifications will be on file from previous jobs and you can read up on the requirements. On the other hand if this is a new area you are trying to get into, you will likely find there is a long and arduous qualification process to get through before you can be considered as a bidder. Either way this is something that will be of most concern to your executive management. Hate to say it, but you are just a pawn in the game...

[free_electron]

eh .. these are not mission critical. OP talks about airconditioning ...
then again .. you are dealing with the military...

And those can be totally bonkers.

[Psi]

eh .. these are not mission critical. OP talks about airconditioning ...
then again .. you are dealing with the military...

And those can be totally bonkers.

Air Conditioning can definitely be mission critical.

Refrigerated/climate controlled shipping containers for example.
If they fail on a trip in the middle of the ocean the goods inside are ruined very quickly.
It's such an issue that shipping companies often pay electrical, mechanical and software engineers to be at sea the whole trip even though there's normally nothing at all to do.
You have 1000's of containers stacked up to many levels that are not easily accessible so they require constant monitoring and the software/hardware to do that.
They need people there who can fix any issue instantly, just in case something goes wrong.

And there are many other areas too, like medical storage.
Also museum storage where old books/artifacts have to be kept at a constant temp/humidity to prevent deterioration

[rsjsouza]

Did you consider the Cortex R4's? They are usually automotive grade (as mentioned by free_electron before) and several have great peripherals (high resolution PWMs, CAN are among them).

Although a lot more complex than your typical 8 or 16-bit micro, they will give you a lot of extra room if the specifications change (or keep changing every week...

Not to mention there is a great variety of dev tools around...

[Rerouter]

though its a more obscure brand of micro, renasis micros are quite popular in electronic throttles in vehicles (i would guess second most important thing in a vehicle following brakes)

[Simon]

well this is all bearing in mind that I am not very good at programming, so I'm really looking for the simplest possible system. Performance should not really be a problem, as all this stuff should be doing is working a control panel, press a button to turn something on or off. I am actually thinking is the arduino an option? it's based on an atmel microcontroller so fairly reliable, and want to have a good program it should be good for good?

[EEVblog]

hi guys, I'm being promoted at work. I think I have pissed the works manager of to the point that he wants to unload me to the engineering Department, where they will be very happy to use my electric and electronic skills amateurish as they may be (although the current skills level in that department is lower). I've not done a lot with microcontrollers, and I guess most designs will benefit from them in what I will do (air conditioning control). So maybe it is time to make a new series start with microcontrollers, considering many of our designs go into military vehicles I would want the hardest thing possible without of course going nuclear proof and expensive. Which of the basic commonly available microcontrollers is generally preferred in these environments?what should I start to learn?

Well done!, enjoy the new role and never step back!

If you want ultra-reliable then as a general rule you'd want to stick with the larger geometry devices, which are invariably the oldest (e.g. an 8051, 80186 et.al).
And also companies that you can throw extra money to characterise them better.
I think the 80x386 series are still popular in those circles?

Here is an interesting list of what space probes have used over time:
http://www.cpushack.com/space-craft-cpu.html

Dave.

[EEVblog]

your entre design should be made automotive grade parts. from capacitor , resistor all the way to cpu...
it is the ONLY way... al the rest is a kludge..

That's pretty much the best easy way to go without having to get into special test/verification deals with the manufacturers.
So you can get better than average reliability ensured, but still be able to buy your parts off the shelf.

Dave.

[Simon]

your entre design should be made automotive grade parts. from capacitor , resistor all the way to cpu...
it is the ONLY way... al the rest is a kludge..

That's pretty much the best easy way to go without having to get into special test/verification deals with the manufacturers.
So you can get better than average reliability ensured, but still be able to buy your parts off the shelf.

Dave.

cough cough my lot of very lax on that sort of thing. At best they just throw something so massive at it it will never go wrong. At worst they just take a chance on some shit they got from China or the bloke down the road that is no better than a hobbyist. I'm hoping to strike a healthy balance. Basically as I am at the very beginning with microcontrollers, it is prime time to decide which branch to go for with this new perspective in mind.
Do I need to consider the software aspects as well? It seems most microcontrollers are used in automotive systems so once I have some working software is that it? I don't really think I want to go for a full blown CPU, as in a 386 or anything like that. This should be really basic stuff where a microcontroller will just be replacing a few chips to make better use of space or add that little function that would otherwise require lots of chips.

In the little amount of work I have done for them electrically in the past I did go towards a microcontroller because the guy could never make his mind up what he wanted and kept changing his mind. So I know that a microcontroller might be a good idea even if just for that so that I can make that simple change in software and not have to redesign my whole PCB layout.

yes really the most important thing in the design is going to be protecting against transience and load dumps, we have had issues with this in the past because we had an automotive grade thermostat with no protection, the customer then came up with a whole load of garbage to protect it from that and we had to accept it. I will be doing better than that.

[EEVblog]

If you have experience in PIC, then I'd just stick with that. Tons of options in the Microchip range, and one of the most popular in the industry.

Dave.

[Simon]

well yes I've used pic but I need to start afresh because the mikroe basic software is crap so I need a fresh start as a bobbiest as well as for work. Which is why I'm willing to switch. I don't know enough about pics for abandoning them a bit to be an issue for me, as the basics I do know are about valid on any microcontroller.

I've ordered an arduino uno just to get me back into things at home at least, after all it is based on C which is better than messing about with basic.

[Rufus]

yes really the most important thing in the design is going to be protecting against transience and load dumps

Then you should read "Def Stan 61-5  Part 6". I think Issue 6 is current.

[Simon]

yes really the most important thing in the design is going to be protecting against transience and load dumps

Then you should read "Def Stan 61-5  Part 6". I think Issue 6 is current.

Yes I have already done some work for them in measuring spikes from their own system to demonstrate to the customer that we are not blowing up our own thermostats. If I design anything I will build in spike protection which so far they have not done and relied on the customer's power supply to be protected. We had issues in the past where the control box would get blown up the first time the vehicle was turned on, then they would replace it and from then on it would be fine. Obviously a problem with the customers vehicle but they tried to blame us.