Choosing a Microcontroller Brand

Joey,

most of what you say has an implicit correctness that depends on what are your global goals. I think you can't compare ARM classes with Atmel AVR / PIC and similar. These are two kind of different things. The first are really micro controller, i.e. near-to-computer devices with too limited features to make them a real computer as we intend today. The biggest characterising difference is the absence of an operating system. You should know that ARM processors, especially the CPU used for embedded applications, also very very vertical ones, has the advantage that all bases their behaviour on a operating system; most of the cases (almost all) are Linux OS. Exists small and super-reduced linux distributions just though to be set to do a single job.

Just some example I can mention, based on my personal experience:

• Thermal printers based on ESC/POS protocol
• Thermal printers from aps-printers.com that has their own Epson-like printing protocol
• Some nintendo accessories for WII (in particular the add-on system to encode the games DVD over the home networks saving them on home NAS or networked HD)
• All the router I know
• Most of the AAA+ freezers

In my personal opinion what you can  - and probably should - do is a reasonable choice having a reduced set of options choosing the better one that fits your needs depending on the project you are developing. Let me suggest a couple of guidelines that has sense conditioning this choice

AVR has generally less power than the PIC microcontrollers (comparing the same devices). It has the advantage together the low price to be really robust micro controllers. The reason is the same that an old 486 is more and more robust - if still exists one outside of some museum - than a i586 quad-core: slower and less delicate.

The advantage of the microcontrollers like AVR and PIC (just to cite two)

Arm is an entire world, But you should focus on them as a computer / OS-based.

What I can definitely suggest you is to make a choice at least forked in two branches: one is choose one or two micro controller: all the small projects, you can do them with AVR, PIC, BBB or whatever you like. You can do all by yourself including the schematics, PCB, design etc.

Then, when the project comes more complex, instead of coming to be an acrobat of the small power MCU you can easily move on a SBC. I can suggest first of all the Raspberry, not because it is better than all the other (anyway it is better than many other, including the YUN) but because there is a wide community of people continuously working on it, solving problems and crating new things. I don't know in depth but I think that also BBB is something similar.

When the Raspi is not sufficient, you can - also easily - interface your favourite microcontroller with the SBC and your architecture grows without serious difficulties (don't worry, difficulties and issues arrive soon anyway, for free!).

What I discourage you instead is to study the assembler of the processors while there are very good C/C++ languages able to compile binaries with better optimisations than what a programmer can do.

Enrico

There are so many different ARM microcontrollers that it's hard to know what to do.  In general, the core itself is pretty much the same so your real question is what on-chip peripherals do you need.  Most have the usual SPI and I2C, but if you want something like touch pads or LCD screen you have fewer options, which simplifies life.  Some cores have significantly higher performance -- for example, the ARM Cortex-M4 has DSP instructions and usually floating-point, and the Cortex-M7 has cache IIRC.

I wouldn't worry about ARM assembly language.  They are almost always programmed in C, and you usually don't get enough performance improvement to justify going to ASM.  OTOH, a generic C compiler won't know about special ARM instructions for DSP, etc.  In that case using ASM or linking in an ASM library can buy you a lot.

My general approach to these things is to keep using what's working and only switch "when you can't go on doing what you're doing now".  There's always a learning curve for any new processor and its software, so if your current AVRs are doing the job it's probably not worth climbing the new learning curve unless you have another reason to do so, like curiosity or wanting to add to your "bag of tricks".

JMO/YMMV

Enrico Miglino wrote:

 

ARM is an entire world, But you should focus on them as a computer / OS-based.

Enrico, what you say about ARM being OS-oriented (especially Linux) is true of the high-end ARM Cortex-A and Cortex-R processors.  However, the ARM Cortex-M processors are very much intended for non-OS microcontroller applications, and various vendors (NXP, ST, Freescale, Cypress, SiLabs, TI, and others) have low-cost, low-power Cortex-M chips for the same applications as AVR, PIC, and MSP430.  The difference is that since ARM is a 32-bit machine with multiply instructions, it's a lot easier to write high-level programs so you can concentrate on the application instead of the limitations of an 8- or 16-bit architecture.  The Cortex-M use 16/32-bit Thumb2 instruction coding, so program size isn't worse than an 8-bit processor.

The low-end microcontrollers usually have simple peripherals and initialization.  You don't have to read a huge manual like a Cortex-A SoC, and it's easy to deal with device registers directly instead of having to face Linux device drivers.

JMO/YMMV

When requirements and price don't point to a clear winner, I just select my favorite one.

Then I at least know that I don't have to add the price/time of a learning cycle to the project.

Hello John,

first of all, what's JMO / YMMV and OTOH ??? I am a Spanish-living Italian ahah

To be honest, it's true, but as far as what Joey wrote - for some unknown reason - I though to ARM a some higher level alternative. Instead there is a problem, I see, adopting ARM micro controllers as the main horse: the relative difficult to find already developed decent boards (e.g. Arduino instead the more sophisticated and in some way limited ChipKit PI and so on) based on these ARM micro. My two experiences of non-OS based ARM processors was developing a really great ARM from Nordic. It was a BLE device to be controlled with Android and iOS, to power and manage the speeds based on several variable curves in PWM. It was a dedicated one all-in-one. Really nice (with its IDE, lot of libraries and well documented). This was a client work but the cost of the development kit was about 650$ plus some other essential accessories, and it was almost impossible to manage it (5x3 mm, 48 pin if I remember) in any other way. I can imagine to develop a project with this kind os - exceptional - hardware.

The other example was for my experiments and was the first NXP Arm modular micro controller a couple of years ago. I pre-ordered it at a super low price from Farnell, wait six months, then was so small and I found nothing to do than buy from NXP their dev kit. I stopped this line just for the cost of the development platforms.

Enrico.

Both Texas Instruments (Hercules LaunchPad Series) and Silicon Labs (Gecko series) have very very cheap ARM dev boards, with free development tool licenses.

I use dev boards and software stack of both companies. They're great.

These two Cypress PSoC boards are very nice and very cheap.  The programmer part can be snapped off, leaving a DIP board.

PSoC4 with ARM Cortex-M0, US$4: http://www.cypress.com/?rID=92146

PSoC5LP with ARM Cortex-M3, US$10: http://www.cypress.com/?rID=108038

These are amazing chips, with programmable digital and analog peripherals.  Unfortunately, there's no open-source software and you have to run on Windows.  I wish silicon companies would stick to silicon

True, I have not yet had to opportunity to do something over these devices.

Enrico

Just a note,

I see that these devices costs from 4$ up to 10$. But I see also that the programmer is sold at 87 Euro on Farnell site. Is this mandatory or there are other ways to program these devices?

Enrico

The programmer comes with it, it snaps off, weird.