Accuracy of rt_get_time_ns()?

John F. Wesseling jwesseling at unav.es
Fri Aug 4 18:21:49 CEST 2006


Thank you Paolo and Calin,

I have gotten it working now.

I could calibrate for a better accuracy with 
calibrate -c, but even then, rt_get_time_ns() 
seemed to have a precise and repeatable 
inaccuracy of ~1:10E4; I do have confidence in my 
measurement devices.  To the best of my 
knowledge, I am not using power saving features or CPU frequency scaling.

Perhaps I am doing something else wrong, but, no 
matter, the "drift" seems to have been fixed by 
adjusting - by hand - the rtai_cpufreq_arg that I 
pass to the hal module.  (I switched to RTAI 3.2 
with linux kernel 2.6.10 because the calibration 
tests weren't as easy to use with RTAI 3.1 - for a novice, at least).

This fix seems to be stable, accurate, and 
precise down to at most 100 microseconds over 100 
seconds, which is what I need.

Thanks again,
John


>John F. Wesseling writes:
>>What is the timing accuracy of rt_get_time_ns() supposed to be?
>>Also, what is the preferred way to keep track of timing?
>>In my hands - details below - the timer is losing 1 ms every 3 s (1:3000).
>>I only need 100 microsecond precision, but I do need much better accuracy,
>>i.e. something like 100 microseconds/100 s (1:1E6).
>>Details:  kernel: 2.6.8.1 patched with rtai 3.1 patch hal7-2.6.8.2.patch
>>         processor:  333MHz Pentium 2
>>         lxrt (I think)
>>         one shot mode (I think)
>>Apologies if this question is obvious:  I am new to Linux and RTAI.
>
>In oneshot rt_get_time_ns gives the CPU time 
>stamp in nanos. The calibration is done against 
>the 8254 freq (1182180 Hz?). On my PIII if I 
>calibrate with the CPU freq in the range of 10 
>nanos uncertainty I can see that the default 
>Linux calibration is in error by 2.5/10000, 
>which is roughly what you get too. So, if no 
>other check source is implied in your 
>verification, you should calibrate for a better 
>ratio using the RTAI calibration tool, letting 
>it run till only the very last figure is 
>changing by a few units. It can take a few minutes.
>Paolo.
>Paolo.



John F. Wesseling, Ph.D.
Departamento de Neurociencias
Centro de Investigación Médica Aplicada (CIMA)
Pio XII, 55
31008 Pamplona
Spain
Phone +34 948 19 47 00 ext. 2005
(if you don't speak Spanish, when the recording comes on,
just enter the extension and wait)
Fax     +34 948 19 47 15








More information about the Rtai mailing list