Home    Bloggers    Messages    Resources
Tw  |  Fb  |  In  |  Rss
Rich Quinnell

32-Bit MCUs Invade 8-Bit Territory

Rich Quinnell
Oldest First   Newest First   Threaded View
Page 1 / 2   >   >>
northstar
northstar
3/28/2012 6:43:51 AM
User Rank
Program Manager
IAR tools
As a first signal from tools vendors, IAR is already announced support for the upcoming Freescale (and ST) chips:

http://www.marketwatch.com/story/iar-systems-supports-arm-cortex-m0-and-freescales-new-kinetis-l-series-2012-03-27-93200

One thing is for sure: supporting a single type of architecture (e.g. ARM) is easier for tools vendors.


50%
50%
raimond
raimond
3/28/2012 7:29:46 AM
User Rank
Word wizard
pin compatible chips
Atmel did it some time ago. The first AVR chips were pin compatible with clasical 80C51/52 chips. The Atmega8515 still is. I used this feature once, myself. The 8515 was the perfect drop-in replacement for an 89C52.

If I will find a cortex-m chip to fit in my current ATmega8 designs for example, I will use it for sure. Of course, the designs are 5V, the chip is the "big" 0.8mm pitch QFP32 .... A lot to think about for the cortex chip manufacturers ...

The 32bitters does have some problems though, with the RAM. On the Cortex-m chips you need over 32 bytes for the stack just for starting to use the interrupts. And another 32 bytes if you have more than one level of interrupts, even if you use the two stacks system of the cortex. So I estimate that an 8K flash/1K ram atmega8 will be about 8K flash /1.5K ram cortex-m equivalent.

100%
0%
duanebenson
duanebenson
3/28/2012 2:24:31 PM
User Rank
Blogger
Not just performance
It's not just performance or cost that keeps the 8-bitters in. Packaging is a factor as well. When all other things are equal, or nearly so, form-factor comes in to play. I recently put a microcontroller into a cell-phone car charger. A six-pin MCU would have been perfect for the application. A 48-pin QFP would not have fit. Were the MCU in the original design, a tiny QFN would not have physically fit but would not have been suitable. It's an all thru-hole design (Amazing in this day but it still happens) so an eight or six pin DIP would be the order of the day. If there are any 32-bit processors in those anachronistic form factors, I'd love to hear about it and take a look at one or two.

50%
50%
Jim Turley
Jim Turley
3/28/2012 4:04:43 PM
User Rank
Blogger
Re: Not just performance
That's an interesting take. The die size (i.e. silicon area) of a 32-bit MCU is small enough but the package often isn't. I wonder what the solution is. Packing a 32-bit MCU into a 6- or 8-pin package obviously means all the RAM and ROM have to be on-chip, but also most of the peripherals. Power, ground, and a couple of serial interfaces is all you'd have room for. Still... an interesting idea.

100%
0%
duanebenson
duanebenson
3/28/2012 5:09:27 PM
User Rank
Blogger
Re: Not just performance
I've never seen a small pin-count 32 bit MCU, but I find it hard to believe that there aren't any. I'm assuming I just haven't found them. If there really aren't any, that may be a key element keeping 8-bits alive.

50%
50%
Nemos
Nemos
3/29/2012 8:06:53 PM
User Rank
System supervisor
Why ?
"The core takes aim at power-critical designs that would normally use 8- or 16-bit MCUs"

I have one question : in witch case we need to 'downgrade" our computing power from 32 bits to 8bits ?

50%
50%
vish2207
vish2207
3/30/2012 2:39:33 AM
User Rank
Program Manager
Re: Why ?
There are applications that needs to downgrade our computing power. One of them is mentioned by Duane.

 

Lets take and example of Lighting control. If I want to change the duty cycle of the Power switch according to commands received remotely, I hardly need one pin for PWM and one or two pin for communication considering it to be simplest communication. Why should I drain more power by using 32 bit microcontroller there?

50%
50%
sdijohn
sdijohn
3/30/2012 1:24:06 PM
User Rank
Bit twiddler
cortex packages and power
There are very small packages available.  One vendor has a 24-pin Cortex M0 in a QFN24 package which is 5x5mm (.200 x .200 in.) and another has one in a .217x.232mm (less than .100 x .100 in.)package.  DIP packages are also available starting at 28 pins from one vendor.  Another advantage of the M0 is current consumption - typically 120-150uA/MHz.  You can run at 8MHz and only use about 1ma of power!

50%
50%
tonyxia
tonyxia
3/30/2012 1:26:46 PM
User Rank
Bit twiddler
Re: Not just performance
NXP has the popular LPC1100L series in low-pin-count packages, available in SO20, TSSOP20/28 and even DIP28.  http://www.nxp.com/products/microcontrollers/cortex_m0/lpc1100_x_l/#products

50%
50%
Jim Turley
Jim Turley
3/30/2012 1:43:13 PM
User Rank
Blogger
Small package MCUs
Energy Micro, a Norwegian company, just announced a Cortex-M3 MCU in a little bitty package that measures 4 mm on a side and 1 mm thick. It's a ball-grid array, so it's not something would could solder by hand, but it is very small.

 

50%
50%
Page 1 / 2   >   >>
More Blogs from Rich Quinnell
In the latest issue of this ARM-centric magazine, topics ranging from Android apps to digital cows.
There is still an opportunity for you to share your knowledge of ARM-based design with the industry, by proposing a paper for the ARM TechCon.
We need to talk about setting up discussion groups on the site, so I've set up a live chat for Friday.
If you think MCC needs some traditional discussion groups, come help set them up.
Hot on the heels of the Beaglebone Black has come a book telling how to use it. A pretty good book, too.
flash poll
MC on twitter
like us on facebook
Microcontroller Central    About Us     Contact Us     Help     Register     Twitter     Facebook     RSS