Robustness and long term reliability of Arduino boards and AVR chips

If you created some application with the arduino boards or using AVR or PIC chips, how good is their long term reliability and robustness?

For example, PLCs are engineered specially to take a beating, long continuous cycle times, electrical circuit protection, and good mechanical protection against shock etc. So ultimately PLCs tend to be workhorses and great for automation applications because they are reliable for long term continuous use and are very robust.

Are arduino boards and AVR and PIC chips as good in terms of long term continuous operation? How reliable are they? Provided of course you are supplying power constantly and it wont be in a situation where it will get knocked around. For example, will the internal circuit inside the microcontroller or on the PCB reach end-of-life or premature failure after being cycled for a long time? What does your experience tell you? How long can you run them continuously before you get 'bugs' or hickups or even total failure?

It depends on what kind of environment you have in mind. Running on my desk flashing LEDs the Arduino is perfectly reliable and robust. But it is certainly not built to withstand an industrial or laboratory environment. There is almost no protection circuitry, so the processor and anything connected to it can be killed by a glitch or a short. The connectors would not stand up to knocks, vibration, or any mechanical stress.

Honestly, the 8 bit AVR controllers seem to be used mainly by hobbyists. I have looked for a list of professional devices that use AVRs but I found no convincing evidence they have serious applications in practice.
Regarding PICs, I have found some credible commercial devices, not related to hobbyists, that use them.

I am not aware of any PLC that makes use of 8 bit AVRs, maybe because the PLC designers do not want to become dependent of a single chip supplier (Atmel). PLCs use ARMs, 8051 or other processors that are produced by many manufacturers.
Also, 8 bit AVR are weak for industrial applications. You could hardly implement a serious modern PLC using AVRs.

How long can the arduino or those AVR/PIC chips run continuously? Say blinking LEDs on Pin 1. How long can it last until it skips a beat or stops working properly or downright fails?

For example, if you make an RFID door to unlock it for a home automation application, you'd want the ensure your controller is quite reliable lol.

Yeah thats interesting too. I always wondered what chips industrial PLCs use. Do they use commercial ones manufactured by large fab houses or do they program their own controllers with FPGAs, ASICs, etc?

PeekabuPi:
Yeah thats interesting too. I always wondered what chips industrial PLCs use. Do they use commercial ones manufactured by large fab houses or do they program their own controllers with FPGAs, ASICs, etc?

They use also FPGAs and ASICs on PLC boards but this does not mean they are implementing their own processors in ASICs. That would be highly inefficient as long as there are proven processors on the market with good compilers and IDE.

PeekabuPi:
How long can the arduino or those AVR/PIC chips run continuously? Say blinking LEDs on Pin 1. How long can it last until it skips a beat or stops working properly or downright fails?

As long as you don't subject the chip to operating temperatures beyond their ratings and power supplied is always within proper ratings I see no limit to continuous operating, there is no 'wear out' factor for the chip. I have a 5x5x5 led cube run with a 328p chip that has been running non stop for over 2 years now, with only a house power outage stopping it momentary a couple of times.

For example, if you make an RFID door to unlock it for a home automation application, you'd want the ensure your controller is quite reliable lol.

The chip will not be the limiting factor to the system reliability, the power supply most likely will be or perhaps the electro-mechanical locking mechanics.

Yeah thats interesting too. I always wondered what chips industrial PLCs use. Do they use commercial ones manufactured by large fab houses or do they program their own controllers with FPGAs, ASICs, etc?

I have an RFID door opener. It works reliably month after month. The only time it failed was when we had a power outage and then a "brown-out" where I measured about half the normal line voltage. I think the board (temporarily) stopped working because the power company brought the power back up slowly. Possibly the RFID reader had failed and not the Arduino. However simply powering it off and on again fixed it.

I'm not sure if I had brown-out enabled on that particular board, possibly not.

If I was worried (I'm not because there are other ways into the house) I would make sure the brown-out was configured correctly, and make the processor turn the RFID reader on (via a MOSFET) in an orderly way. Also I would have a watchdog timer set up (which I haven't at present) to force a reboot if it gets into some loop.

I don't think there is any reason to suppose the the AVR chips are unreliable. In any case as Retrolefty said, I would be more worried that the RFID reader failed, or the door lock failed.

How long can it last until it skips a beat or stops working properly or downright fails?

Operating within design parameters, there is no reason for it to "wear out" or "skip a beat".

Beside AVRs, Atmel also manufactures Intel 8051 compatible processors and many other IC which have heavy industrial or domestic applications.

The microelectronic technology Atmel uses for all of its chips is reliable.

However, for newly released AVRs there could be problems:

Bad Remark:
Atmel AVR microcontrollers are inexpensive and have an inexpensive development tools but the AVR microcontrollers are not up to par for commercial applications. I used a AVR32 in a comical application and while the chip did have some nice features is also had several bugs in the board support package, drivers, development tools and in the chip itself. To make matters worse each shipment of parts not only had bugs but different bugs requiring a hardware and software change for each shipment. This may not be a big problem for a hobbyist but it’s a disaster for a company that wants to put its product into production. I would not recommend anyone attempt to use an Atmel part in a commercial product unless they want to lose there job.
_Answer: _
Sounds like you had some trouble with AVR32, which is relatively new. However, standard 8-bit AVRs are used all over in successful industrial and commercial products, and this has been the case for many, many years. – wjl Oct 15 '11 at 1:31

subway:
However, standard 8-bit AVRs are used all over in successful industrial and commercial products, and this has been the case for many, many years. – wjl Oct 15 '11 at 1:31[/i]
microcontroller - Why are Atmel AVRs so popular? - Electrical Engineering Stack Exchange

There was a thread a while back about products that folks had torn down for parts or whatever and found Atmel MCUs inside. I happen to know that they are used in automotive applications. I expect that they are very robust indeed. I would not expect MCUs from Microchip, TI, Maxim, NXP, ST Micro, etc. to differ significantly in that regard.

Arduinos are "prototyping platforms" (first sentence on the home page) so right there that tells us they are not necessarily meant for deployment in industrial or even commercial applications. That said, they do the job of prototyping platform very well. Even so, I would still expect them to be very robust in reasonable environments.

subway:
To make matters worse each shipment of parts not only had bugs but different bugs requiring a hardware and software change for each shipment. This may not be a big problem for a hobbyist but it’s a disaster for a company that wants to put its product into production.

I seem to recall an EEVblog about PIC chips also having bugs.

Ah yes, here:

Title: "EEVblog #53 – Mr Murphy and Microchip PIC Silicon Bugs".

In the middle of the episode he finds that one of the sets of the ICSP programming pins are not connected internally in the chip! And that was the one he chose to wire up on his board.

So it's not just Atmel that have production problems.

I expect that [AVRs] are very robust indeed. I would not expect MCUs from Microchip, TI, Maxim, NXP, ST Micro, etc. to differ significantly in that regard.

That would match my expectations as well. Semiconductor processes are pretty uniform, and you don't get to be the size of Atmel by selling chips that are not "Robust" (where "robust" means that they don't fail without encountering circumstances where you'd expect them to fail.)

There's a whole science to "reliability engineering." Spontaneous chip failures are not a likely cause of failures.

(Also, you can't be an Atmel (or a chip manufacturer at all) by selling chips just to hobbyists. Just because you can't find a list of products that use AVRs doesn't mean that there aren't any.)

[Atmel] Microcontroller segment net revenues increased 95% to $892 million for the year ended December 31, 2010 from $458 million for the year ended December 31, 2009. The increase in net revenues was primarily related to increased volume shipments from customers for both AVR and ARM-based 8-bit and 32-bit microcontrollers. Microcontroller net revenues represented 54%, 38% and 33% of total net revenues for the years ended December 31, 2010, 2009 and 2008, respectively.

Yes, I can't seriously see Atmel staying afloat on the hobbyist market. They have a huge product range. To keep hobbyists happy they would only need two or three chips, and tell them to suck it up.

The net revenues reported by a company or in the name of a company in the media do not weight too much. Figures like this "[Atmel] Microcontroller segment net revenues increased 95% to $892 million for the year ended December 31, 2010 from $458 million for the year ended December 31, 2009." can be misleading.

That was from the company "Annual Report", which is one of those things that had better be pretty accurate (or you get lawsuits for misleading the investors.) It can still be misleading, though. (For instance, I notice that it doesn't break down the multiple types of microcontrollers that Atmel sells, so it's presumably everything from 8051s to AVRs to AVR32s to ARMs.)

The point is that you don't sell a billion dollars worth of chips in a year to hobbyists...

And they do sell a bunch of 'em, according to Wikipedia,

About 55% of all CPUs sold in the world are 8-bit microcontrollers and microprocessors. According to Semico, over four billion 8-bit microcontrollers were sold in 2006.[4]

A typical home in a developed country is likely to have only four general-purpose microprocessors but around three dozen microcontrollers. A typical mid-range automobile has as many as 30 or more microcontrollers. They can also be found in many electrical devices such as washing machines, microwave ovens, and telephones.

Perhaps not surprising but still impressive numbers. I imagine a lot of the folks here on the forum don't have "typical" homes with a mere three dozen MCUs :smiley:

It can be calculated from the report below how much money Atmel got from 8 bit microcontrollers (likely 8051+AVR8+others), this year and in 2011. I made the calculation and it appears that they earn two - three times more money from 8 bit families than from 32.

"Moving to a discussion of our business segments. For our Microcontroller business unit, Microcontroller revenue of $226 million increased 3% sequentially and was down 25% as compared to the third quarter of 2011.
For the second quarter in a row, both our core microcontroller revenue and maXTouch revenue grew sequentially.
By product family, during the second quarter, our 8-bit micros were down 7% sequentially, and were down 35% year-over-year. While 32-bit microcontrollers increased 33% sequentially and increased 11% year-over-year."

Source: Atmel Management Discusses Q3 2012 Results - Earnings Call Transcript | Seeking Alpha

PeekabuPi:
If you created some application with the arduino boards or using AVR or PIC chips, how good is their long term reliability and robustness?

Excellent.

PeekabuPi:
For example, PLCs are engineered specially to take a beating, long continuous cycle times, electrical circuit protection, and good mechanical protection against shock etc. So ultimately PLCs tend to be workhorses and great for automation applications because they are reliable for long term continuous use and are very robust.

Read these forums for a while. Arduinos take an awful lot of abuse. Remember: the connectors on the Arduino board go directly to the chip, there's no protection circuitry.

PeekabuPi:
Are arduino boards and AVR chips as good in terms of long term continuous operation? How reliable are they?

AVR chips are designed for industrial operation. They have built-in watchdogs, brown-out detection, everything you need.

The Arduino boards could be improved, but not terrible. They're designed for hobbyist price levels. There's more robust versions out there if you look.

The Arduino IDE/software doesn't enable stuff like brownout detection, watchdog. You can do it but not many people will. In spite of that the programs seem to run for months/years without problems so long as the power supply is good.

twice:
Honestly, the 8 bit AVR controllers seem to be used mainly by hobbyists. I have looked for a list of professional devices that use AVRs but I found no convincing evidence they have serious applications in practice.

That's complete rubbish. There's no way Atmel would survive if that was true.

They also wouldn't be making all those six-pin, surface mount variations of the AVR chips, etc. Do "hobbyists" use those?

fungus:
That's complete rubbish. There's no way Atmel would survive if that was true.

Are you clawson, the main moderator from avrfreaks?

The main problem is the awful connectors on the board. Wires poked into pin headers are brilliant for prototyping but awful for producing reliable installations.

If you take those connectors off and connect directly to the board you'll have something pretty damn robust IMHO. If you want to build you own you could start with something like this Diavolino . It allows you to choose your own power components, capacitors, etc.

With a decent power supply the difference between a slightly uprated Arduino and 'industrial' would mostly be down to the software.

twice:

fungus:
That's complete rubbish. There's no way Atmel would survive if that was true.

Are you clawson, the main moderator from avrfreaks?

No...