5

I have been developing embedded stm applications bare metal along with ST Hal for a few years and would still consider my self a beginner as the process of learning in a bare metal environment can be tedious. My aim is to be able to develop hard real time systems like flight controllers. I have known about mbed for a while but have not tried it as my impression is that it is not suitable for hard real time systems because of the overhead that comes with it. It seems that mbed is for IOT devices only.

I would like to go with a more standardized development ecosystem with good community support like mbed. Does it make sense to switch over to mbed or to just keep on the track I am currently which is using freertos or QP with bare metal?

more info: mbed OS claims to have an rtos in it, however it is not clear if it is suitable for real hard time systems.

Hadi Jaber
  • 61
  • 3
  • An RTOS stands for real-time operating system, so an RTOS, by definition, should be suitable for real time systems. – DKNguyen Mar 03 '19 at 22:24
  • https://www.electronicshub.org/real-time-operating-system-rtos/ I understand however not all rtos are able to do hard real time vs soft real time. – Hadi Jaber Mar 03 '19 at 22:28
  • 1
    This will probably have to be closed as opinion based. But it is worth considering that the strength of, er, "assisted" environments is typically early in the cycle with their trade-off costs and oddities showing up later in a project. So it's probably not worth transitioning something that *already* works. But keep in mind that "mbed" is several different pieces with complex relationships and a degree of potential independence, too. – Chris Stratton Mar 03 '19 at 22:29
  • I write hard real-time O/S all the time. It's just a component for the measurement system. Equipment I work on requires extremely low sampling jitter and FFTs and other types of analysis are often done with output measurements (and PID control is strongly impacted by sampling jitter, anyway) so I usually require no more than about \$2\:\mu\text{s}\$ as the worst case in ADC-in sampling to DAC-out variability. All processing must be guaranteed performed, though it depends on the application. I start with this as input and design everything (+O/S) around that. Here, I don't use COTS O/S. – jonk Mar 03 '19 at 23:04

0 Answers0