choosing a processor for hard real-time

[shaiko]

In your opinon, what should be the most important considerations when choosing a processor for a hard real-time control application?

For example: short interrupt latency, fast context switching etc...
*Power consumption isn't an issue.

 

[mtwieg]

A short and consistent interrupt latency, and a sleep mode that wakes up in a couple clock cycles are key for me. Beyond that it depends exactly what you inputs and outputs are, and how tight your requirements are.

 

[keith1200rs]

Predictable instruction timing can be an issue with pipelined processors.

Keith

 

[doraemon]

Hello!

In your opinon, what should be the most important considerations when choosing a
processor for a hard real-time control application?

Everything depends on what you call "real-time". Real-time means everything and therefore
nothing.

Beyond that it depends exactly what you inputs and outputs are, and how tight your requirements are.

Exactly!
Example: if you have a swimming pool and want to measure pH 4 times a day adjust
it to a specific target value, then use a 4-bit micro controller (if you can still find one)
and your system will be real-time.
Now if you want to do real-time face detection and react within 1 ms, then almost any
existing processor will not make it and your system will not be "real-time".

Dora.

 

[shaiko]

keith1200rs,

I don't know of any new processor that isn't pipelined - do you ?

 

[keith1200rs]

It depends on what you mean by new and how the pipelining works and its effect. I had some issues with a PIC32 giving me erratic pulse widths (which were software controlled) which work fine on the PIC24E. I tracked it down to the PIC32 pipeline effect. I don't think the PIC24E uses pipelining but the PIC32 definitely does.

Keith

 

[xueyanxu]

Hi Dora,

As a newbee in the embedded world, I am trying to implement "real time" FFT in a project for MSP430F5529.

I found your old post about FFT algorithm in 2011 very helpful. I am just wondering wheter you have any updated suggestions on this. I found a c code fix_fft.c online. It should work for us. Our sampling frequency is 500 Hz, and we don't need very high accuracy. As you suggested, fixed point radix style FFT should be good for us.

Thanks.

Sherry
Sherry