Rabbit Alternative

We currently use an RCM3200 in our main product line.  It's a great
chip for what we're doing with it, except that it's soon going to be
not fast enough at 40MHz.

Can anyone recommend a suitable, faster, alternative to it?

The key features that we need are:

At least 2, preferably 3 or more serial ports.
At least 10 general purpose I/O pins, 26 or more would be the bee's
knees.
At least 1 pwm output, preferably 2 or more programmable pwm like the
rabbit.
Ethernet port with a UDP/IP stack.
500KB data store (flash, ram, it's not stored between sessions so I
don't care)
Enough program flash space for the UDP/IP stack and a simple tftp
application.

The programming environment is less important to me.  It does, however,
need to be usable for a 2-man development team (I'm software, my boss
makes the hardware).  Price per unit, anything under $200 is fine --
more than that and it would just take some evaluation to determine the
advantage of the product to see if it's worth it.

The key that we're needing is SPEED.

Thank you for any suggestions you've got.

Alex McHale

On 14 Feb 2006 09:14:17 -0800 in comp.arch.embedded,
"alexmchale@gmail.com" <alexmchale@gmail.com> wrote:

>We currently use an RCM3200 in our main product line.  It's a great
>chip for what we're doing with it, except that it's soon going to be
>not fast enough at 40MHz.
>
>Can anyone recommend a suitable, faster, alternative to it?

You might look at Netburner's stuff.  I used the MOD5282 at a PPOE.
60-some MHZ Coldfire, gcc environment, pretty good support.  Probably
not as cheap as Rabbit.  I needed CAN bus, some of their other modules
might be cheaper and fit your needs better.

Good luck,
                                        -=Dave

-- 
Change is inevitable, progress is not.

alexmchale@gmail.com wrote:

> We currently use an RCM3200 in our main product line.  It's a great
> chip for what we're doing with it, except that it's soon going to be
> not fast enough at 40MHz.
> 
> Can anyone recommend a suitable, faster, alternative to it?
> 
> The key features that we need are:
> 
> At least 2, preferably 3 or more serial ports.
> At least 10 general purpose I/O pins, 26 or more would be the bee's
> knees.
> At least 1 pwm output, preferably 2 or more programmable pwm like the
> rabbit.
> Ethernet port with a UDP/IP stack.
> 500KB data store (flash, ram, it's not stored between sessions so I
> don't care)
> Enough program flash space for the UDP/IP stack and a simple tftp
> application.
> 
> The programming environment is less important to me.  It does, however,
> need to be usable for a 2-man development team (I'm software, my boss
> makes the hardware).  Price per unit, anything under $200 is fine --
> more than that and it would just take some evaluation to determine the
> advantage of the product to see if it's worth it.
> 
> The key that we're needing is SPEED.
> 
> Thank you for any suggestions you've got.
> 
> Alex McHale
> 
The TMS320F2812 was doing fine there until you got to the Ethernet port 
-- could you do that externally?  There may be other TI DSP chips (or 
perhaps an Analog Devices BlackFin) that would do.

-- 

Tim Wescott
Wescott Design Services
http://www.wescottdesign.com

Posting from Google?  See http://cfaj.freeshell.org/google/

Our  MOD5270 meets all of your requirements 

The list prices is $59.00 in thousands.

See:http://www.netburner.com/products/core_modules/mod5270.html

We have had a lot of customers switching from Rabbit and they seem happy with the result.

Paul Breed
CTO NetBurner.

alexmchale@gmail.com wrote:

> We currently use an RCM3200 in our main product line.  It's a great
> chip for what we're doing with it, except that it's soon going to be
> not fast enough at 40MHz.
> 
> Can anyone recommend a suitable, faster, alternative to it?
> 
> The key features that we need are:
> 
> At least 2, preferably 3 or more serial ports.
> At least 10 general purpose I/O pins, 26 or more would be the bee's
> knees.
> At least 1 pwm output, preferably 2 or more programmable pwm like the
> rabbit.
> Ethernet port with a UDP/IP stack.
> 500KB data store (flash, ram, it's not stored between sessions so I
> don't care)
> Enough program flash space for the UDP/IP stack and a simple tftp
> application.
> 
> The programming environment is less important to me.  It does, however,
> need to be usable for a 2-man development team (I'm software, my boss
> makes the hardware).  Price per unit, anything under $200 is fine --
> more than that and it would just take some evaluation to determine the
> advantage of the product to see if it's worth it.
> 
> The key that we're needing is SPEED.

You could ask them about the Rabbit4000, as a first step ?
That would be the easiest upgrade path.

If they don't have a comming module fast enough, then you could look at
the ROMLESS ARM9's : the Philips LPC3180, and Atmel AT91SAM9260
are going to give a module that is enough of a step up to bother with.

-jg

alexmchale@gmail.com wrote:

> We currently use an RCM3200 in our main product line.  It's a great
> chip for what we're doing with it, except that it's soon going to be
> not fast enough at 40MHz.
> 
> Can anyone recommend a suitable, faster, alternative to it?
> 
> The key features that we need are:
> 
> At least 2, preferably 3 or more serial ports.
> At least 10 general purpose I/O pins, 26 or more would be the bee's
> knees.
> At least 1 pwm output, preferably 2 or more programmable pwm like the
> rabbit.
> Ethernet port with a UDP/IP stack.
> 500KB data store (flash, ram, it's not stored between sessions so I
> don't care)
> Enough program flash space for the UDP/IP stack and a simple tftp
> application.
> 

You should probably consider moving to an ARM.

Ian

Jim Granville wrote:
> You could ask them about the Rabbit4000, as a first step ?
> That would be the easiest upgrade path.

Personally, after watching Rabbit's track record with the 2000 and their 
early core modules and compiler I'm happy to wait a while before jumping 
on the new one.

Kelly

Just wondered what exactly your application is. I'd have thought that
the RCM3200 would run a simple TFTP application without even having to
get out of bed.

If you want lots of I/O why not just add on whats needed using I2C?

If speed is the issue, then try to determine whereabouts in your system
any bottlenecks can occur. Are your files stored in some slow flash
using an inefficient system?

Maybe write your own file system instead of using the ones supplied by
Rabbit. We did that, essentially using a linked list for storage. Very
simple with low code overheads.

Do you store any data in the userblock area? This can be slow,
particularly as it is mirrored. Any data writes take twice the amount
of time that they should do. Convert the userblock type from 4 to 2
which gets rid of the mirrored system. Any writes will now take half
the time as before.

Have another look at your existing product and see what you can do.
Perhaps it is better than moving to new hardware, new software and new
bugs.

Hope that helps :)

alexmchale@gmail.com wrote:

> We currently use an RCM3200 in our main product line.  It's a great
> chip for what we're doing with it, except that it's soon going to be
> not fast enough at 40MHz.


Alex,

What operation is pushing you over the top with your current platform?

To Rob's point, take a look at your major consumers of resources and see 
if there's a better approach - what are the major suspects?  (I'd think 
long and hard before uprooting a working system; optimizing is usually 
much more appealing.)

Richard