DPRG
DPRG List  



[DPRG] tiny-ARM from New Micros

Subject: [DPRG] tiny-ARM from New Micros
From: Kipton Moravec kip at kdream.com
Date: Tue Feb 22 07:28:15 CST 2005

On Mon, 2005-02-21 at 16:01 -0600, Randy M. Dumse wrote:
> > It looks like it has all of the power you might
> > need in a little robot. It will make the
> > calculations for the balancing robots really fly.
> 
> We've found the 60MHz ARM and the 80MHz DSP processor seem to
> run about neck and neck, best we can tell. We've got a balancing
> robot Mike Keesling did on a 'Pod, and there's one at MIT doing
> another balancer.
> 
> I feel the DSP is better for motion control apps, while the ARM
> is better for data processing apps. Actually, I think the ideal
> would be one of each working together. There would be an awesome
> bot.

While I agree with you in theory, in practice until good low-cost or
free C compilers are available for the DSP will be out of reach for most
hobbyists.  That is not to say anything against your great efforts with
Forth on the DSP. But it is a learning curve that will be a problem for
many people. 

I first started with the 8051 on the NMI-0031 $39.00 special and the
free small C compiler. The compiler was not very good, but I was able to
make what I needed work, even though it did not have floating-point. I
got so involved with the 8051 on projects that I finally spent the $1500
for the Keil compiler. The difference was night and day. Fortunately
there is a GCC compiler for the Atmel AVR and for the ARM.  I have heard
there is one for one of the TI DSPs but I have not looked into it.  As a
result, I do not think I will purchase a compiler again for my projects.
(If a customer pays then that is different.)

> 
> > Why did you go with the Philips over the Atmel? ... What
> > were the main differences that made you go to the Philips?
> 
> Well, that Philips LPC2106 was really well packed out with a
> very small chip format. Seemed a natural for our line of parts.
> So this was an easier entry point for us to start with.
> 
> If you look at the Atmel, most of their package sizes are much
> larger. However, I wouldn't assume too much about one over the
> other, because we also have developments in works now. We are
> building an Atmel processor board up right now. We also plan to
> get around to Motorola's ARM's sometime in the future. We really
> see ARM as being a processor of the future, and we will have
> several products supporting it.
> 

After seeing one of the Tiny-AVRs at the contest, I started reading up
on it this weekend. I was looking at your stuff on the New Micros
website, and the Atmel and Philips website.

I have a waveform generator that I have been building for a few years
using the 89C51RD2 (a 8051 variant) and A Xilinx Spartin FPGA. The first
prototype used the NMI-0031. :) 

I am looking at building the next generation version. I was going to use
the Atmega 128 but after this weekend I am beginning to lean towards the
ARM with a simpler FPGA running at 60 MHz. I will put more functions in
the SW than before. The 89C51RD2 was running about 2MIPS. The Atmega 128
would be more like 14 MIPS while the ARM would be closer to 54 MIPS (and
they would be 32-bit MIPS versus 8-bit MIPS which helps with math and
moving data. I can move data 4x as fast per instruction.)  Also in my
old version I saved the waves in 32K x 8 memory (remember the 64K
limits?) I want to use 4M x 16 memory in the new version and was working
out a way to bank address it. Being able to deal with 32-bits at a time
makes getting at the large memory a lot easier.

I am seriously thinking of getting the Tiny-ARM for the prototype. If it
is not classified, which Atmel ARM are you looking at? The one with USB
looks most interesting to me for this project.

Kip

-- 
Kipton Moravec <kip at kdream.com>


More information about the DPRG mailing list