Getting started with ARM processors - recommendations

Yes.

I've been happy with NXP. Your mileage may vary, of course. One thing you may want to do is to look at what dev boards are available for the various chips, at the usual places. See what Digikey et al. carry as well as

(drill down to the specific vendors) and (drill down for specific cores).

Compilers, development environments, JTAG adapters. I've been using Imagecraft's ARM tools with the NoICE debugger. Rowley CrossWorks is another option, as well as the usual gcc releases and the compilers from The Big Boys.

I got turned off PICs after having to force world+dog through that one damned W register. Found life was much more pleasant over in the AVR world and downright cheerful in the ARM universe.

One thing that Microchip offers that few others still do is ROM-based vice flash-based processors. One-time programmable ROMs have a longer expectancy than flash and are reasonably well protected alteration.

--
Rich Webb     Norfolk, VA

I'd recommend getting one of the very affordable SM32 eval kits from ST. The upside is they have good affordable parts (and kits). The downside is good dev SW is kinda pricey but well worth the price if you're a pro. I use IAR EWB but I'm sure you can use less expensive tools maybe even some supposedly free tools.

JJS

I have only had very good experiences with Microchip support, but there may be an entirely different reason for ARM to consider. Once you have invested in the toolset (and by this I mean invested time _and_ money _and_ tools _and_ operating environment and everything else that in some fashion diverts needed energy away from the product focus ahead of you), you aren't nearly so tied to one manufacturer as you might be with single sources. Yes, the boards may need a redesign if you shift. Yes, the peripherals may be different. Yes, the initialization code will have to be changed, too. But you will do all that without (usually) having to relearn how to properly use the compiler tools and so on. So you get to preserve at least _some_ of your prior investment should it come to pass that you feel a strong need to once again choose a different manufacturer or product line.

It seems that way to me, anyway. So it's not a bad way to go.

Your definition of "mostly low power" may be different than mine. 1-10 milliWatt average is more like it. Does Actel make FPGAs anywhere near there?

You wrote that you've had "extremely poor technical support." I don't know why. It's not my experience. Poor support is a totality thing, though, and that can be caused by reality (the vendor really is poor at it) or by a customer's (you) attitude. But assuming you aren't a "customer from hell" for now and that it is the reality:

I consider good support pretty important. And if you have had very poor experiences with Microchip and have enough experience with them to know that you can't uncover the problem and make it get better, then you should definitely look elsewhere until you find that positive support from another company.

By the way, when you do decide to shift over, write me a letter if you have any inclination to sell your Microchip tools. I have good experiences and wouldn't mind getting more tools if offered at a very reasonable cost to me.

You should look at everything when considering a move. What compiler development tools will you use? (With ARM, you'd be wise using gcc but that doesn't mean not paying -- as you can secure excellent libraries, IDEs, support, and so on from vendors who use gcc but also deal with all the associated and related issues so that you can focus on the applications with confidence.) Will you be using JTAG debugging and, if so, which tool? Check out the datasheets and see if you feel they are complete and readable and make sense. (I've had the experience where a cpu I chose [there were no alternatives, so I was stuck anyway] had incomplete information on the DMA and it required them spending months tracking down the original designer [done 8 years beforehand] of that section to get the answers back.) Ask yourself how long you expect to support all this, as well, and whether or not the company you choose has the kind of past behavior that matches your need here well. That goes not only for the chip itself, but the compiler vendor and JTAG tool and whatever else you will add to the development infrastructure.

I'm sure others will add many more.

Good luck, Jon

Thank you for the comprehensive responses.

I see that there are some very nice options from all of the above mentioned ARM variant suppliers ( AVR, NXP, Freescale, ST). From what I see most of them offer the power, peripheral requirements that I would be looking for. I will have to look further to see what kind of of the shelf availablities are. We do a lot of R&D, making high quantity purchases = NULL.

One major thing that I have like about using microchip is the pin to pin compatibility from 16 bit devices MCU's, DSPs and 32 bit MCUs. As well the debuggers are cross compatible. The code libraries are immense and the forum support has been good. The only downfall being the separate purchase of the 32 bit compiler. The hardware has operated quite well.

" You wrote that you've had "extremely poor technical support." I don't know why. It's not my experience. Poor support is a totality thing, though, and that can be caused by reality (the vendor really is poor at it) or by a customer's (you) attitude. But assuming you aren't a "customer from hell" for now and that it is the reality:"

Microchip touts "the best technical support in the industry", but I have definitely not received it. I have heard from some like yourself, that they also have received good support. I do not beleive that it is the vendor's inability to supply it and I hope that I am not the "customer from hell". I have asked a number of newbie questions, as I was quite new to embedded design and i was in process of getting up the learning curve, and as a company we don't buy mass quantities or parts( not that they would know this). As a company we have been using them for over three years. My last open case is weeks old with no repsonse, and previous was closed with a quick guess as to what the problem was. Most of my support tickets are resolved with a guess and then closed. Would someone please let them know I am really not the "cutomer from hell" :). Because of the proportionately degraging support with time I have been left wondering if there is something better out there, hardware, support.

"Your definition of "mostly low power" may be different than mine. 1-10 milliWatt average is more like it. Does Actel make FPGAs anywhere near there?"

The very impressive spec given in data sheet is 4uW is flash freeze( power down) on igloos devices, gate configuration and clock speed is the main driver of power consumption, so active power is dependent on final gate configuration. I was quoted that I should be able to get active current in the range of 100uW, which is very suitable for our battery operating needs.

The main items that I am left open ended in question are the tool set options / supplier. I have been using microchip's IDE with their ICE, ICD programming/debug tools so I am unfamiliar with the more universal solutions. I like the sounds of the approach you mention here in being able to migrate from mfg to mfg with minimal impact.

"Once you have invested in the toolset (and by this I mean invested time _and_ money _and_ tools _and_ operating environment and everything else that in some fashion diverts needed energy away from the product focus ahead of you), you aren't nearly so tied to one manufacturer as you might be with single sources."

1. So what would most univeral toolset solution be? Already mentioned has been: ImageCraft, Keil, I like the sounds of the mention GCC solutions. Who offers these solutions?

1. Unlike Microchip, are most other devices programmed debugged through jtag? What is one of the better jtag hardware devices? I would like something that works with FPGA as well, if possible.

Thank you immensely for the advice.

--------------------------------------- Posted through

mentioned

of

for.

availablities

the

purchase

they

vendor's

last

know

power

in

ICD

has

I had problems with Microchip too. Their PIC32MS460F516L were suposed to have a UART that can work up to 20Mbps. After designing a prototype to run some tests i was having some communication problem. After some time i found out that the problem was with the start-bit timing. It was 33% shorter than it should be, After that i could find this problem described on a hidden ERRATA. The problem would happen for baud rates higher than 5Mbps. And the fantastic workaround was: Do not use it at that boudrate! So far, after 7 months and 3 tickets, i cant get any kind of further information about this problem, if it has been solved, or if there is any new version with the problem solved. All the 3 tickets were closed with some stupid and random aswer as check the baud rate registers.

I still use microchip uC, but not for important tasks.

For ARM i would sugest you getting a STM32 b oard that costs 10 USD. You can hack as much you want with it, and it comes with a jtag and a ARM Cortex-M3

Cya

--------------------------------------- Posted through

For small-quantity work and R&D, I'd push the importance of readily available dev boards up a notch. You could use something like

(also available at

pretty much as-is on a breadboard or proto-board.

The ARM cores themselves will be (nearly) the same but how the cores interoperate with on-chip peripherals, what peripherals are present, and the ultimate pinouts vary widely. The price of choice, I guess.

Fortunately, they all speak JTAG.

Rowley has a commercially-supported gcc with their own libraries, relieving end-users of LGPL licensing concerns (insofar as there may be LGPL concerns...). YAGARTO

is the "standard" (AFAIK) fully GPL gcc environment for ARMs.

Some (many? most?) also support some flavor of bootloader over a serial port. NXP has an "LPC2000 Flash Utility" for example. As far as JTAG goes, the standard seems to be the J-Link from Segger. There are less expensive options (search for "JTAG Wiggler clone").

As far as JTAG interoperation with FPGAs, if you're lucky your programmer supports "skipping" uninterested devices on the JTAG chain. Programming isn't going to happen; they all seem to use different protocols.

--
Rich Webb     Norfolk, VA

All ARM core devices should speak JTAG, but not all hardware JTAG interfaces will talk with the debugger software on the host. So make sure that these match up when picking tools.

Yagarto is very popular, and totally free (AFAIK), but CodeSourcery has a better claim to being the "standard". They are the people that actually write the compiler and tools (or at least, they are the official maintainers of the ARM port of gcc, and very heavy contributors). They have options ranging from free through cheap and up to expensive, depending on your requirements for support, library features, etc.

Yagarto, Rowley, Code Red, etc., all get their basic compiler from CodeSourcery, and provide their own additional features - support, libraries, bundling and integration with an IDE, and so on.

I think that before you choose your toolset, you should pick your main chip vendor. The different vendors will have more or less support for different toolchain suppliers. For example, if you pick TI's Stellaris devices, all their application notes, libraries, demos, etc., have direct support for IAR, Keil, Code Red and CodeSourcery. This doesn't mean that you can't then use e.g., ImageCraft - but it might be easier to get started using one of the vendor-supported toolchains.

The best toolchain for /you/ will depend on your experience, your budget, your needs, your particular project, your personal preferences, and the needs of other projects (for consistency between different processors). Ignore anyone who tells you you have to spend vast amounts of money on IAR or Keil for professional work, or that those tools will save you money in the long run. They /may/ be the best choice for you, but equally they may be far more costly in time and effort than cheaper or free tools would have been - cost price is never a guarantee of quality or time efficiency. Look at the web sites of the vendors and try out evaluation versions of tools that are within your budget and see what suits.

JTAG devices come in two basic forms. You get simple USB devices, which are very often based on an FTDI usb chip - these are typically cheap, and work reasonably well but may be slow if you are working with very large programs. The other main type are "intelligent" devices, that typically have an Ethernet interface to the host PC. These will be a lot more powerful - they will handle things like watchpoints much more efficiently, downloads are faster, they may support tracing, etc. But they cost a lot more.

As for the FPGA, I'd recommend buying a USB Blaster for Altera, or the equivalent for other FPGA manufacturers, and using that. It is typically /possible/ to program ARMs and FPGAs on the same JTAG chain, but it is almost certainly going to be more effort than it's worth. It's a lot easier just to put two separate jtag connectors on the board.

Thank you all for the very informed advice. This will greatly benefit me in the transition. I will likely start working with some of these newer devices until comfortable and then fully transition away from the microchip product.

By the way I just got off the phone with sourcery. They highly recommended the TI stellaris cortex M3 famil as they offer very wide support and a very discounted dicounted G++ package directed for its use. Interestingly the Stellaris has not been aforementioned. It looks like a nice product with TI support. I will continue to do further research and try and narrow down a supplier.

--------------------------------------- Posted through

Don't forget Atmel, TI and there are some other, newer players.

People tout the advantages of the ARM CPU popularity as allowing you to switch vendors easily, but that is over rated in my book. It is still a ton of work to switch because of all the differences, pinout, peripherals, init code, clocking, etc... That is one really good reason to pick a vendor with a wide portfolio of CPUs, they will be much more compatible within the line.

I can't say anything about Microchip as I have not worked with their parts. With so many alternatives, there has to be good vendors in the ARM market... or they don't last long in the ARM market. TI is making a VERY strong push in the ARM market. In addition to the CM3 line from Luminary Micro, they have higher end devices and will be coming out with a CM4, I believe.

Don't confused the Igloo parts with the SmartFusion parts. Igloo has a few chips which will support the soft core CM1. I haven't seem much to indicate these are selling well. With the CM1 taking up nearly the entire die in the smallest member of this short line, I expect there are few designs that are enhanced by this approach. The SmartFusion chip has FPGA, CM3 and analog all on one chip... one $40 chip. These parts are all much cheaper if bought separately. They will take up more room on the board, but your cost will still be lower, significantly.

If you want to see low power, check out Silicon Blue parts. They have very low static current and the dynamic current scales with frequency. Even at the same rates, they are lower power than any other FPGA.

WHOA!!! Don't think the impact is "minimal". You may not need to buy a new code tools, but literally everything else will change, possibly even the download/debug cable. Embedded designs often require the software to be tightly coupled to the hardware and everything other than the CMx will be different with different mfgs.

Oh, I almost forgot. Freescale is coming out with a CM4 device which should be pretty much a hot rod. The Kinetis K10 looks like a really nice chip. It even has good analog on board. Packages range from 5x5 mm 32QFN to 144QFP or BGA. I expect this chip, if priced right, to find sockets in my designs in the future.

Rick

[%X]

Try

--
********************************************************************
Paul E. Bennett...............
Forth based HIDECS Consultancy
Mob: +44 (0)7811-639972
Tel: +44 (0)1235-510979
Going Forth Safely ..... EBA. www.electric-boat-association.org.uk..
********************************************************************

We use Stellaris chips for a few designs - so far, they've been quite good. Their evaluation kits at very cheap and easy to work with, and the CodeSourcery support is solid. As you say, CodeSourcery have a low-price version that only supports Stellaris, or you can get the full ARM toolchains that support the full range of ARM devices.

A particularly nice feature of the Stellaris chips is the very flexible IO - most of the IO functions (UARTs, timers, etc.) can be connected to several different pins.

Another manufacturer no one seems to have mentioned is Energy Micros. They make very low power Cortex devices. I haven't used them or looked at them in much detail, but if low power is important then they are worth a look.

hase

This can be an important detail, so you should at least get a PIC32 starter system and try it.

The ARM ecosystem is large, and broad, with M4 parts soon from NXP and Freescale, and M0 parts from NXP and Nuvoton.

- as well as M3 from many suppliers.

Overall, the Starter systems are quite cheap, so buy more than one, and install the code and test it.

A new entrant that appeals to me, is the Nuvoton M0 parts, because they are 5V operation, and they have some 8051 flavoured (pin/peripheral) variants (M05x), that are quite cheap (some under $1, and a low cost eval/debug board ~$20) In stock at NuHorizons.

Cypress PSoC5 could be nice, if it ever fully releases.. at sensible prices.

-jg

Yes, They are almost as low power as the AVR32 based AT32UC3L (0,29 mW/MHz), but not quite ;-) The AVR32 is more in the Cortex-M4 class, than the M3.

A toolchain which might be worth looking at for an Atmel SAM3 (Cortex-M3) is Atollic TrueStudio. This is a free of charge Eclipse/gcc environment. There is also a "Pro" version which supports a lot of fancy features, but cost some money. The SAM3S will support FPGA communication over a parallel bus even with a 100 pin part.

If you start to read in on the soft ARM cores from Actel, you will find that the small FPGAs will fit the core, but nothing else. Not even on-chip debugging. The newer variants with CM3 hardcore are quite expensive.

An MCU + FPGA is likely the best alternative, unless you have real space constraints.

--
Best Regards
Ulf Samuelsson
These are my own personal opinions, which may
or may not be shared by my employer Atmel Nordic AB

How does the AVR32 compare to the MIPS processors ??

hamilton

It compares with difficulty.

MIPS is a family of licensable cores covering a huge spread. The microMIPS cores will be probably be smaller and lower power than an AVR32 (though it will depend on the implementation), while the biggest cores are orders of magnitude faster. "MIPS" is more like "ARM" in this respect.

The most power-efficient MIPS core is probably the M14Kc at 90nm, and if you accept the raw figures from MIPS website it has a core power of 0.08 mW/mHz, runs at 200 MHz and does 1.5 DMIPS/MHz.

If you look at the list of MIPS licensees, you will not have heard of most of the companies. And of those you have heard of, you know most of them as end-product manufacturers (such as Cisco). There are familiar names on the list, such as Texas Instruments, but you won't find a mention of MIPS on their website - they have it available as a core for their ASIC customers. MIPS processors are one of the most widely used architectures (I'd love to see figures comparing the numbers of MIPS, ARM and x86 processors in use). But they are also mostly invisible unless you are in the market for at least hundreds of thousands of devices at a time, or are involved in particular market segments (high-end networking and set-top boxes are a couple of examples).

The only MIPS devices that are easily available in the mass market are the Microchip PIC32 devices. AFAIK, they use the M4K - a somewhat larger and older core than the above-mentioned M14Kc. How it compares in power to the AVR32 devices, I don't know.

Thank you for this explanation.

I will look at PIC32, FMI .

hamilton

They have just announced that they are licensing the M14K.

Leon

I guess you can't knock the MIPS parts for being a poor step child like the AVR32. It may not be available from more than one maker as a standard part, it does have a huge base of users, etc. So it is not likely to go away... at least not as a core. Has Microchip been selling these chips in high enough volume to assure that they will continue to make them for years to come? That would be my biggest concern at this point.

The other 32 bit chips with a sole source core may have commitment from their makers. Renesas for example, will not likely drop their offerings. But Toshiba has at least put one foot on the ARM bandwagon. I can't imagine they will continue to develop and promote both product lines indefinitely, likewise for Atmel, regardless of what some will tell you.

So if the PIC32 parts don't sell well, how much pain is Microchip willing to take before they drop the line??? I guess if they are licensing the M14K core, they are at least going to continue with this part for another five years.

I'm sure they are working to keep up with anything the ARM camp throws at them. If they are going in ASICs, I expect they aren't just being selected because of the existing code base. They also need to be technically competitive.

Rick