FreeRTOS a good choice for timed tasks?

misterdsp wrote on Thursday, April 26, 2007:

I’m looking at coding a project that has two kinds of timed tasks.

One kind of task - regularly timed, such as data from a converter. The uP needs to generate the timing using a uP hardware timer e.g. sample audio at 44100Hz rate. The uP I’m using has hardware dma and I need the task to use that. (Microchip dsPIC33FJxxxGPyyy)

Second kind of task - irregular timing such as an interrupt from a zero cross detect circuit. The software will need to use a uP hardware timer to time the irregular events.

Is FreeRTOS a good choice? I’m just starting to look at the tutorials and any extra “real world” advice is most appreciated.

nobody wrote on Thursday, April 26, 2007:

44.1K is a fast sample rate if you need software to run on each sample.  If most of this is done using the DMA (as you hint) then the software overhead will be a lot smaller.

When using any RTOS, not just FreeRTOS, then you have to consider the timing jitter and response times.  When processing data within a task response times can be extended due to the need to save and restore task contexts.  Jitter can be introduced as in any mutitasking system it is sometimes required to disable interrupts to perform critical operations.

The advantage with in both of these cases is that you get the source code and it is very flexible.  With the source code you can see exactly what the RTOS is doing, and configure you system as you require.  For example, if you need a very fast very accurate (w.r.t time) interrupt then you can make the interrupt priority greater than the RTOS priroity, and make critical sections not disable your fast interrupt.  That way your fast operations work above the RTOS operation and are not interfered with.  How achievable this is depends on the processor, I know it is possible on ARM7 and M3 devices but I am not familiar with the dsPIC.