Time management for real-time applications

Question

How to generate precise pulses with python?

from threading import Thread
import time

def runA():
    while True:
        print "a"
        sendSignal(1)
        time.sleep(0.0001)
        sendSignal(0)
        time.sleep(0.005)


if __name__ == "__main__":
    t1 = Thread(target = runA)
    t1.setDaemon(True)
    t1.start()
    while True:
        sleep(10)

time management in python is not precise. specially to create multi-threaded PWM generators (assuming each thread to send data at specific rate) . is there any better way to generate pulses? precise ones!

[The solution is language-independent]

`This code does not print "a" 10 times every second` - What does it do? — thefourtheye, May 30 '15 at 15:00
it prints "a" on random intervals, but it should be printing a 10 times each second — Rami Dabain, May 30 '15 at 15:01
well, it would print it about 10 times per second of that thread's execution time. If the CPU is really busy you might end up with it being less regular than that — Eric Renouf, May 30 '15 at 15:09
running the function `runA()` directly (not within a thread) does not have any problems. CPU load is 0% (running linux without a UI - just command line + SSH) — Rami Dabain, May 30 '15 at 15:16
The CPU shouldn't be zero, since your main thread appears to run a continuous loop and should therefore be getting much more CPU time than the sleeping sub-thread. — holdenweb, May 30 '15 at 15:24
I ran the code exactly as is, unchanged, and it works exactly the same under python3 and python2-- python2 behaves exactly the same as python3. — snapshoe, May 31 '15 at 01:50
You can use time.sleep() to sleep a thread. So you might try a `sleep(0.1)` instead of `pass` in your main thread's loop — holdenweb, Jun 05 '15 at 16:22
@holdenweb I understand this frees up some time for other code to run in the processor, but still it will not be precise right? i need it to be precise to 1ms for PWM generation. and this won't fix it. ... i guess i will change my question. — Rami Dabain, Jun 05 '15 at 16:35
I suspect you may be struggling to achieve that level of precision without using a specialized OS (such as one for embedded systems). I would also point out the 0.0001sec is one-tenth of a millisecond, not that I understand your issues fully. — holdenweb, Jun 05 '15 at 21:11
@holdenweb with a single loop in a single thread i tried a 5 millisecond delays and it worked, but had +- 1 millisecond from time to time. whemever there's another thread running the error margin increases. what specialised OS can achive that? will itnrun python? — Rami Dabain, Jun 06 '15 at 09:51
I'm not an embedded systems expert, I'm afraid. You should be looking at QNX and similar systems. [This question](http://stackoverflow.com/questions/1402933/python-on-an-real-time-operation-system-rtos) might shed more light on the subject for you. — holdenweb, Jun 06 '15 at 10:03
Sounds exactly what i need. can you please post this as an answer? — Rami Dabain, Jun 10 '15 at 12:02

score 0 · Accepted Answer · answered Jun 16 '15 at 17:24

My Solution was to find a real-time OS Actually raspberry pi have a real-time patch for the raspbian OS

For more information : https://www.raspberrypi.org/forums/viewtopic.php?t=39951

Was a little of a challenge to enable it. but it is working and the error rate dropped dramatically.

Thank you @holdenweb for the tip with the real-time OS

score -1 · Answer 2 · answered May 30 '15 at 15:02

-1

Your output is probably buffered so you'll get a big block of them printed all at once. You could try flushing stdout after each print like

sys.stdout.flush()

answered May 30 '15 at 15:02

Eric Renouf

13,950
3
45
67

It is not about the `stdout`. the `print "a"` is just an example of whatever function I need. the original is a function to send data through a socket ... another function is PWM generation – Rami Dabain May 30 '15 at 15:04

Time management for real-time applications

2 Answers2