Advice on switching microntrollers

Ok, flamewar time I guess.
I am looking to explore new microcontrollers because the ones I have under my belt are coming up short for what I need. I have a lot of
experience with the 16 series of PICs from Microchip, I have played with basic Stamps (I, II, SX), some experience with Motorola's HC11, HC12 and HC16 series. Currently I use Atmel's AVR (Atmega128) at work. But in trying to branch out and find a better or at least alternate choice I turn to the advice of those who already have experience with something else.
My requirements are: *External interrupts *PWM outputs (everything I've used so far has done these with hardware timers. seems nice) *ADC inputs 8 bit resolution minimum, 10 or 12 bits nice. *I2C or whatever that manufacturer likes to call it. Atmel calls it TWI so they don't have to pay license fees to Philips or something. *UART(s) *hardware SPI *JTAG nice but not required *CAN support would be VERY nice but not absolutely required *In circuit serially programmable *Support for a secondary external crystal (32KHz for timekeeping) *C friendly (and no, PICs do NOT fit this...) As a side note, GCC compatible is a BIG plus. Paying out the nose for 3rd party compilers is a killer. *SMALL!!! I'd like to keep it smaller than an Atmega128 which is 64TQFP at 16x16mm. *Low power (sleep modes are good). We're trying to stay below 10-20 mA for full speed execution and <1mA for a sleep or low power mode.
My coworker would also like a cheap/free set of programming and debug IDEs for it. I have experience with MPLab from Microchip and AVRStudio from Atmel so anything that good or better would be fine. Just something graphical to step through code, set breakpoints, simulate, etc.
So anything not made of unobtanium that you guys would recommend? Thanks! -Will
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
snipped-for-privacy@aol.com wrote:

I don't see what you have against unobtainium. It is a wonderful material. It's just a shame it's so expensive. ;>
Have you looked at the MSP430 series from TI? Quite a bit of your requirements list covered (although I think not all). 14 bit ADC's on some processors in the family 12 bit on others.
--
********************************************************************
Paul E. Bennett ....................<email:// snipped-for-privacy@amleth.demon.co.uk>
  Click to see the full signature.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
snipped-for-privacy@aol.com wrote:

Since you already have a reasonable range of 8 bit uC's, there would seem little point adding another one. just to get a learnng curve....
I'd look at the ARM families, Philips presently have the cheapest/smallest start at $1.50, and Atmel have ones with Ethernet. Pretty much everyone makes some ARMxx variant, including Freescale and TI.
-jg
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
On Sat, 22 Oct 2005 09:28:57 +1300, Jim Granville wrote:

...
Or, work on new approaches to task sharing. What ever happened to The Connection Machine? I remember reading a thing in some mag, where some guy said, (I paraphrase) "Well, the only problem we have now is how to program the thing!" "Ah, but we have an endless supply of programmer flesh - undergrads!"
But nothing ever seems to have come of it.
Thanks, Rich
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Danny Hillis is having more fun building crazy 10000 year clocks, and big ole robots, out of Glendale CA, at a place called Applied Minds.
wrote:

Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Rich Grise, Plainclothes Hippie wrote:

Aahhhhhhhh!!!!, dreams of hypercubes....
There have been some interesting approaches to programming meshed and close coupled processor cores. Some of them have been reported in The Journal for Forth Applications Research. Forth was apparently the bring-up language for one of the MPP's.
--
********************************************************************
Paul E. Bennett ....................<email:// snipped-for-privacy@amleth.demon.co.uk>
  Click to see the full signature.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Rich Grise, Plainclothes Hippie wrote:

I think you hit the nail on the head. Programming and programmers. As I recall, the original CM-1 was a large array of 32,000 or so 1-bit processors, which I think no one could effectively program. Then, around 1987, they changed over to the CM-2 with the 1-bit processors being replaced with a bunch of Natl Semi 32032 [a dead horse] 32-bit chips on plugin boards. Intel and maybe TI and some others also had various hypercube architectures, but I think they all died out since their target market was the same people buying large NEC and IBM supercomputers. Couldn't compete on doing things the old way.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Hello Will,

Nah. Just different opinions.
I'd side with Paul. The MSP430 series is really nice and their ALU runs at 16 bits. Almost feels like getting into that new V8, including the financial pain that comes with that. Don't expect them to be much under a Dollar for the bar bones versions, no matter what TI marketeers claim. Once you want ADC and some other niceties expect to pay over $2. So, right now they can't play in the Atmel price range.

All uC I have seen so far do it in hardware timers, or not at all. The MSP is geared nicely here but again it uses the timer. For truly dedicated PWM you need to look at DSP.

This is where the MSP430 truly excels. I don't know if any other uC can go that low in power. They even have a mode where you can keep a RTC running yet it sips only a couple uA or so.
Regards, Joerg
http://www.analogconsultants.com
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

NEC V850 series. Unfortunately the tools I used were Greenhills. Expensive crap, but V850 might be supported by GCC now.
John
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Somewhere on usenet Fri, 21 Oct 2005 22:41:51 +0100, John B wrote:

a list of current targets for gnu/gcc http://gcc.gnu.org/install/specific.html
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

There is a east MSP430 missing msp430gcc is available too. AFAIK there is a HOWTO out with eclipse as IDE. Feels like CCE from TI ;-)
Marte
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
On Fri, 21 Oct 2005 11:55:01 -0700, larkmore wrote:

[shopping list snipped]
Get the biggest FPGA or CPLD you can, and write the one of your dreams. ;-P
Personally, I rather like microprogramming, albeit I'm kind of a solution in search of a problem in that respect. For example, my first real computer was a "Scelbi 8-H", which had an intel 8008 and 256 bytes of RAM. Eight toggle switches and three pushbuttons on the front panel, and three rows of 8 LEDs: H, L, and M. The buttons were, "Interrupt", "Step", and "Run". They weren't debounced, which was a very powerful learning experience! ;-)
I guess the point is, the 8008 "assembly language" mapped almost character for character onto the machine instructions, which, when I saw the diagram of the architecture of the 8008 <googles> Ah, rats. Most of the google hits are about history and crap - I was hoping for a diagram. But it's so freaking simple that the program is the equivalent of a microprogram. Like, the instructions are in octal, two bits of op code, and three bits each for usually the destination register (A,B,C,D,E,H,L,M) and three for the source register (see above.) "3" was the opcode for "move". And when you look at the architecture diagram of the 8008, you can see that when the instruction enters the black box, it's split up into opcode, destination, and source - which go directly to the data selectors on the inputs or outputs, as designated! Oh, of course, there are other opcodes, but they had sub-codes, which still only needed one more level of decoding - my point being, that the program for an 8008 was, essentially, a microprogram. :-)
Cheers! Richh
Cheers! Rich
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
Rich Grise, Plainclothes Hippie wrote:

For more direct-mapped nostalgia, compare the instruction set of Data General Nova and the Texas Instruments SN74181 ALU chip.
Another pretty easily microcoded example is the Digital PDP-11.
--

Tauno Voipio
tauno voipio (at) iki fi
  Click to see the full signature.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
On Sat, 22 Oct 2005 06:42:37 +0000, Tauno Voipio wrote:

I had an opportunity to use an editor on a PDP-11 and cross-assembler to write some test stuff for an IMP-16: http://www.bitsavers.org/pdf/national/imp16/4200036A_IMP16P_Descr_1974.pdf The company had the naked boards in stock, to use in another thing; they didn't have that box. There's more stuff: http://www.bitsavers.org/pdf/national/imp16 / And, I believe I've seen the 74181s, or maybe AM2903s, on a PDP-11's CPU board. I also once worked on an all-TTL "microprogrammable processor", but it was 32 bits, and was used to test disk drives. That one had even TTL RAM. The instruction cycle was 168 ns, which was unheard of at the time. And, the thing that really makes the microprogram stand out as a microprogram is the "next instruction" field in the instruction. The G-15's machine language was like that too, but that was for timing, since all of the memory was circulating on a magnetic drum.
Ah, I nostalge. <sigh> Sorry.
Cheers! Rich
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
snipped-for-privacy@aol.com wrote:

You do of coursde realise that I2C and SPI just need a couple of port pins.
There's nothing special about the hardware. Simply write your own device handler.
Graham
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
On Sat, 22 Oct 2005 00:08:36 +0100, Pooh Bear

If one need to bit-bang these ports, then one should just note the I/O speed of the MCU. On many of the new ARM MCUs, the possible I/O toggling speed is a lot lower for a specific system clock speed than for the 8-bit MCUs.
Regards Anton Erasmus
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

How is that so? I mean, is there some kind of long propagation delay or what? I have played around with PICs for years, but I have been gaining interest in the ARM micros. This sounds like a serious issue to me.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

ARMs typically use the AMBA bus where you connect the CPU to the 32 bit Advanced System Bus (ASB) and then there is a bus bridge to an Advanced Peripheral Bus (APB). The Bus Bridge will add delays so you find bit banging will take 5-6 clocks at best due to delays. The Philips LPC was so bad at this, that some new parts have the PIO on the ASB bus instead of on the APB bus. The AT91SAM7 is limited by the I/O buffers but H/W can shuffle bits at 25-30 Mbps.
--
Best Regards,
Ulf Samuelsson
  Click to see the full signature.
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
On 21 Oct 2005 11:55:01 -0700, snipped-for-privacy@aol.com wrote:

The MLF package is a LOT smaller (9 x 9 mm), and not that difficult to work with. (Much easier than BGA).

Look at the various new generation of ARM based MCUs. There are quite a lot available from manufacturers such as Philips, Atmel, ST, Analog Devices, TI and Freescale.
Regards Anton Erasmus
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload
snipped-for-privacy@aol.com wrote:

http://www.cypress.com
Click on PSoC Mixed-Signal Controllers.
The in-circuit emulator is not free.
Cheers, John
Add pictures here
<% if( /^image/.test(type) ){ %>
<% } %>
<%-name%>
Add image file
Upload

Polytechforum.com is a website by engineers for engineers. It is not affiliated with any of manufacturers or vendors discussed here. All logos and trade names are the property of their respective owners.