What are chips like the MEGA2560 used for in real-world applications?

I see the applications for the low-end 8 bit controllers that Atmel makes and the ARM processors too, but I don't understand the niche for the ones in the middle like the 2560. It's fairly expensive and has a small number of pins and raw compute power and memory compared to ARM. What is it used for? It seems the only slam-dunk application would be legacy low volume apps where compatibility with existing code is paramount. What sort of products do we know the 2560 is in except for the Arduino Mega? And does the go for XMEGA and AVR32 as well which seem to fall in the same boat?

JoeN:
It's fairly expensive and has a small number of pins and raw compute power and memory compared to ARM.

It's not all about "raw compute power and memory". Sometimes you just need an Uno with a bit more RAM and/or a few extra serial ports and I/O pins.

I've just designed a system for a company using the 2560, not normally my choice but they have a current system using Megas and they want to keep the code and I guess they are comfortable with the chip.


Rob

fungus:

JoeN:
It's fairly expensive and has a small number of pins and raw compute power and memory compared to ARM.

It's not all about "raw compute power and memory". Sometimes you just need an Uno with a bit more RAM and/or a few extra serial ports and I/O pins.

Great. What real world application did it make it into under these circumstances? If this was a hobby application, I totally understand why someone would use it. I've used them. But I can't believe that Atmel designed this chip just for hobby apps. There just aren't enough hobbiests in the world to make it worthwhile. They had to have designed it with real world applications in mind, something with at least somewhat of a higher volume in mind. I am at a loss in figuring out what it might be though. I know the lower end chips go out in hundreds of millions of products a year - toys, remotes, small appliances, etc.. I am just trying to figure out what the 2560 is used for. The way it is priced, I just can't see how it would make it into many real high-volume products.

JoeN:

fungus:

JoeN:
It's fairly expensive and has a small number of pins and raw compute power and memory compared to ARM.

It's not all about "raw compute power and memory". Sometimes you just need an Uno with a bit more RAM and/or a few extra serial ports and I/O pins.

Great. What real world application did it make it into under these circumstances? If this was a hobby application, I totally understand why someone would use it. I've used them. But I can't believe that Atmel designed this chip just for hobby apps. There just aren't enough hobbiests in the world to make it worthwhile. They had to have designed it with real world applications in mind, something with at least somewhat of a higher volume in mind. I am at a loss in figuring out what it might be though. I know the lower end chips go out in hundreds of millions of products a year - toys, remotes, small appliances, etc.. I am just trying to figure out what the 2560 is used for. The way it is priced, I just can't see how it would make it into many real high-volume products.

Field of Dreams, built it and they will come. :wink:

Lefty

JoeN:
I know the lower end chips go out in hundreds of millions of products a year

Name one... :slight_smile:

fungus:

JoeN:
I know the lower end chips go out in hundreds of millions of products a year

Name one... :slight_smile:

I know the majority of radio control ESC speed controllers are based on various versions of the mega8 chip. It's a very big market but probably not hundreds of millions?

Lefty

Sometimes you need something that can run for months/years on a watch battery; 8-bit micros are great at that. You can't do that (as easily) with an ARM. When you look at the line-up that Microchip has, you can see that there are some pretty specific requirements out there and if you can save half a cent on the chip by finding some incredibly obscure part number, it might be worth tens of thousands of dollars in extra profit. I'm thinking smart thermostats are a good location for a 2560, what with complex, custom LCD screens that aren't x by y pixel oriented graphics.

A guy told me once that he'd never seen people become so bitter as when they told the board that they needed to spend 3 cents more in total cost per toy on electronic parts. This for a fairly expensive toy. It's all about competition and the profit margin out there in the real world; every penny counts. It's dog eat dog in toy land.

fungus:

JoeN:
I know the lower end chips go out in hundreds of millions of products a year

Name one... :slight_smile:

That is a good point. I have not done a teardown on enough stuff to know this for sure. Atmel has a positive EBITDA, $300 million in cash, $0 in debt. $1.43B in total revenue. You have to sell a lot of cheap microcontrollers to make that kind of money. It looks like a healthy company for the time being. Their trailing PE is a lot higher than their forward PE. Did they just reorganize something to lower costs? Anyway, I am guessing they sell a ton of the lower end products. Does anyone have better knowledge of specific applications of Atmel microcontrollers over all of their lines?

JoeN:

fungus:

JoeN:
I know the lower end chips go out in hundreds of millions of products a year

Name one... :slight_smile:

That is a good point. I have not done a teardown on enough stuff to know this for sure. Atmel has a positive EBITDA, $300 million in cash, $0 in debt. $1.43B in total revenue. You have to sell a lot of cheap microcontrollers to make that kind of money. It looks like a healthy company for the time being. Their trailing PE is a lot higher than their forward PE. Did they just reorganize something to lower costs? Anyway, I am guessing they sell a ton of the lower end products. Does anyone have better knowledge of specific applications of Atmel microcontrollers over all of their lines?

7682219 (ATML) Stock Price, News, Quote & History - Yahoo Finance

I think a few years back (when the depression hit?) they sold off all their foundries (fabs) and now contract that part of manufacturing out.

Lefty

Products stick around long after their markets go away. Once a product goes into its mature lifecycle, the cost of continuing to sell it is generally pretty small. It doesn't take a big market (cough Arduino boards, cough) to keep a chip around.

Keep in mind that an existing product may not have competed with all the complementary products it does today, when it was introduced.

JoeN:
Does anyone have better knowledge of specific applications of Atmel microcontrollers over all of their lines?

They are very focused on selling their touch controllers.

What do those go into? POS terminals? Seems like every retail credit card terminal has touch now, you don't even have to use the buttons, just touch the screen.

Anything and everything with capacitive touch.

Spend a few minutes reading atmel's marketing literature.

It would probably find a use in those situations where you need more inputs, and it is cheaper to get the bigger chip than the smaller chip and lots of shift registers.

Plus you may need more RAM or program memory. And the extra hardware serial ports would be useful.

I would guess something like a washing machine or dishwasher might need the extra digital/analog inputs, plus the extra timers would come in handy.

Also the extra EEPROM would be handy for storing logs of what went wrong.

I know that when our dishwasher gets serviced he connects up a gadget to the LEDs on the front panel and downloads the logs of what happened when.

Except the most popular touch device in the world... =(

I'm pretty sure the lowest price bracket for thet 2560 would cause any washing machine or dishwasher manufacturer to look elsewhere. Margins are just too tight on that sort of thing to consider a chip that looks like it costs eight dollars or more in even the largest quantities. But that is just my guess.

JoeN:
I see the applications for the low-end 8 bit controllers that Atmel makes and the ARM processors too, but I don't understand the niche for the ones in the middle like the 2560. It's fairly expensive and has a small number of pins and raw compute power and memory compared to ARM. What is it used for? It seems the only slam-dunk application would be legacy low volume apps where compatibility with existing code is paramount. What sort of products do we know the 2560 is in except for the Arduino Mega? And does the go for XMEGA and AVR32 as well which seem to fall in the same boat?

Its not as expensive as you think - commercial products will be buying multiples of 10,000 and the price breaks for high
volume are substantial - qtys of 500+ already about half unit price... Typically PCB board area (and associated manufacturing
costs and enclosure) is more expensive than the chips on it unless exotic devices are used.

Some of the market will be for gadgets developed on smaller AVR processors, but which have become more featureful in
successive versions and the developers move to a microcontroller that has more memory which is code-compatible with
previous versions. The cost of re-engineering the hardware and software to switch between vendors is not trivial and
carries plenty of risk compared with staying with tried/trusted tech - especially in a competitive market where keeping
up with market developments is a race with time.

Also when a range of similar products is developed it is desirable to base all of them on the same software base - so using a range
of uControllers from the same vendor pays off - the cheaper/less featureful units can shave cost by using cheaper versions. The more
advanced units in the product range will demand more memory, I/O, interfaces etc...

The uControllers vendor that can provide a wide range of software-compatible controllers will be favoured in the long run.

That is only the case if the fab producing the chip has enough volume to keep it running, either churning out the original chip, or multiple chips using the same process technology. If the numbers are really low, the fab's owners will cease production, and either upgrade the fab to support whatever the next generation of chips are or close it.

I shoot with Olympus cameras, and for awhile, Olympus, Toshiba, and Fuji came out with their own memory cards (originally smartmedia and then xD). At one point, the high end xD cards were completely unavailable, and the explanation I heard was Toshiba found it more profitable to move the fab to make memory cards for the latest Apple phones/tablets of the time, then to continue production of a card whose market share was dropping by the year. Olympus who by that time, was the only company still making cameras that used xD cards, hobbled along with the stocks of the slower xD cards, and eventually switched to SD card varients like most other cameras.

Fabs are very expensive to run and they need to produce enough chips that people will buy to continue running. And since it is even more expensive to build a completely new fab than to retool an existing one, companies have motivation to close unprofitable lines.

My coffeemaker has one in it, I also found one in a printer once. I guess there are plenty of applications where a more simple chip is, well, too simple. A mega 2560 might just be enough to do the job, an ARM chip might be overkill. Also, if a company already has a line of products based on AVR’s, they may not want to switch. On the other hand, I’ve seen a company spend $100.000 on rewriting their entire software stack so they could use a micro-controller that was just a tiny bit cheaper (but with a planned production of a million units, those tiny bits do add up).

pbrouwer:
My coffeemaker has one in it, I also found one in a printer once. I guess there are plenty of applications where a more simple chip is, well, too simple. A mega 2560 might just be enough to do the job, an ARM chip might be overkill. Also, if a company already has a line of products based on AVR's, they may not want to switch. On the other hand, I've seen a company spend $100.000 on rewriting their entire software stack so they could use a micro-controller that was just a tiny bit cheaper (but with a planned production of a million units, those tiny bits do add up).

Your coffeemaker has a 2560 in it? What is it, a Keurig or a thousand-dollar expresso machine? Most coffeemakers cost so little these days I would expect them to be using a ATTiny type microcontroller at most, if they use a microcontroller at all.