[ArmDeveloper]

I am starting a new machine controller project. It involves lots of PWM generation, some A2D conversion, some CAN communciations, and we are going to try to connect a PC to the controller via USB. We allso want to log a bunch of data to an SD card on the controller as a way of collecting operational data and error logging.

I used an LPC2148 on a past (smaller) project and it worked well.

For this project I am trying to decide between an LPC2939 (ARM968-ES) and the STM32F205 or the STM32F405. I love the I/O on these processors, but I'm wondering if they have enough processing muscle to do everything.

How do these processors (ARM7, ARM9 and Cortex M3) compare in speed for general purpose controller work ?

I am not sure that we would need floating point for anything, but there is control loop that we might tackle digitally and I'm wondering if FP operations might be very handy for it. Comments ?

If these controllers don't have enough grunt, the next processor we would consider is the Freescale MPC55xx line. They look like a decent processor, but overkill for our project. And I am not taking a liking to the dedicated timing engine with its own language approach to things. I also question the public knowledge base available for these processors.

Is there anything else (ARM, preferably) that I should be looking at ?

Any and all advice welcome.

Thanks !

[42Bastian]

Hi
in a simple benchmark using our RTOS I found that a Cortex-M3@50Mhz comes equal to an MPC5554@132MHz. So for the step towards PowerPC, I'd first use a quicker ARM.
Regarding ARM9<->Cortex-M, I would go in a new design for the new (not so new anymore) architecture.
AFAIK, you can get Cortex-M up to 120MHz. Or Cortex-R4F (TI Hercules) up to 160MHz.
Also TI offers an optimized Fixpoint library, which might ease the decision FPU or not.

Just my 2€-cents.

[ArmDeveloper]

42Bastian, on 30 January 2012 - 07:29 AM, said:

Hi
in a simple benchmark using our RTOS I found that a Cortex-M3@50Mhz comes equal to an MPC5554@132MHz. So for the step towards PowerPC, I'd first use a quicker ARM.

I'm shocked to hear that. Thanks for sharing it. I thought for sure that the MPC55xx devices would be a lot faster, with the PPC core and all. I guess processors have advanced since they were introduced.

Quote

Regarding ARM9<->Cortex-M, I would go in a new design for the new (not so new anymore) architecture.
AFAIK, you can get Cortex-M up to 120MHz.

Actually 168 MHZ in the M4 in the STM32F4xx series, which (apparently) equates to 210 DMIPS.

Quote

Or Cortex-R4F (TI Hercules) up to 160MHz.
Also TI offers an optimized Fixpoint library, which might ease the decision FPU or not.

The M4 has a single precision floating point library. I think I found my solution.

Thanks for your quick reply.