Stability of time ticks on the XDK
응답
15. 8. 27 오전 9:01
Dear all,

I have a question about the time ticks on the XDK (I use the function xTaskGetTickCount() in file SDK/lib/FreeRTOS/source/tasks.c to get the ticks).
Are they generated by a stable intern clock, or are they increased by software? Has one tick always the same duration, independent of the workload on the XDK?
How it is possible if I set an interval and the difference between two messages is less than these interval (in ticks)?
(E.g. 10 ticks sending interval and difference between two messages is 4 ticks)

Thanks and Regards
Frank
0 (0 투표)
Stability of time ticks on the XDK
응답
15. 8. 28 오전 4:13 as a reply to Frank Slomian.
Hi Frank,

I can't confirm your suspicion yet, but we had problems with the tickCount before in the DataLogger Demo where we've seen some irregularities before.
We're currently investigating here to see what's wrong here.

In the meantime, I try to get some information for you to see from where the xTaskGetTicketCount() takes the actual tick Count.

- Florian
0 (0 투표)
Stability of time ticks on the XDK
응답
15. 9. 8 오전 2:07 as a reply to Florian Harr.
Hi Frank, 

I've forwarded your request on August 28th to our second level support, but haven't heard back yet.

I'll make sure to get a answer within the next 7 days. Sorry I can't help more right now.

- Florian
0 (0 투표)
Stability of time ticks on the XDK
응답
15. 9. 21 오후 12:25 as a reply to Florian Harr.
Dear all,

I think, we have also a accurateness problem on the timestamps of the data logger demo application.
After switching off all sensors except bmi150 and magnetometer, we see a logging interval of 6 ms (difference between 2 timestamps). We measured a running cycle time of 600 ms (100 points).
An simultaneously camera based measurement with a camera picture (frame) frequency of 250 Hz (4 ms interval), the same running cycle times are mesared of about 720 ms (+ 20 %).
We have found, that the camera frequency is korrekt.
Is it possibel, that the inaccuracy of the timebase of the XDK110 reachs up too 20 % ?
kind regards
  Manfred
 
0 (0 투표)
Stability of time ticks on the XDK
응답
15. 9. 24 오전 4:03 as a reply to Manfred Holzapfel.
Hi Manfred,

we've seen quite some issues with the DataLogger Demo and the internal clock as well.
I would love to provide you with a solution to these problems as of right now, but we haven't quite completed the work on these issues just yet. 

I'm still waiting for a response from our second level on the internal clock an other questions.

What I couldn't fully understand is how you were conducting these tests? Could you might try to explain these tests/experiments again in some greater detail?

Regards,
Florian
0 (0 투표)
Stability of time ticks on the XDK
응답
15. 9. 24 오전 10:58 as a reply to Florian Harr.
Hi Florian,

I had now run the data logger Demo Applicatein two times for exact 10 minutes.
Firtst with the sensors bmi160 and bmm150 switched on (Sample time showes 6 ms). ==> Timestamp of last log: 470040 ms, Error -129960 ms or -21,66 %

Second with the sensors bmi160 switched on only.  (Sample time showes 4 ms). ==> Timestamp of last log: 618868 ms, Error +18868 ms or +3,14 %.
So on the second test, the timebase accuracy seems to be mutch better

Kind regards
  Manfred






 
0 (0 투표)
Stability of time ticks on the XDK
응답
15. 9. 24 오후 10:19 as a reply to Manfred Holzapfel.
Manfred,

this is exactly what we've seen before and what I documented here in another experiment as well.

We're working hard on getting a fix for this.
Once we have something we can share, we'll let everyone know.

For now, my apologise for the inconvenience caused.

- Florian
0 (0 투표)
Stability of time ticks on the XDK
응답
15. 11. 11 오후 1:07 as a reply to Florian Harr.
Hi Florian,
we now have a solution for the timimg problem.

Kind Regards
  Manfred
 
0 (0 투표)
Stability of time ticks on the XDK
응답
15. 11. 12 오후 6:44 as a reply to Manfred Holzapfel.
Great to hear Manfred!

Thanks for letting us know
0 (0 투표)
Stability of time ticks on the XDK
응답
15. 11. 13 오전 10:14 as a reply to Florian Harr.
Dear Manfred,
Also for me the timing issue is very relevant. So I'm very intersted in your solution. Is there anything you can share?

Thanks,
Andre
0 (0 투표)
Stability of time ticks on the XDK
응답
15. 11. 13 오전 11:00 as a reply to Andre Schwarzmeier.
Dear Manfred,

I also would be interested in your solution. I encounter the same problem of wrong time intervals.
Maybe the overhead for the task switching should be considered when calculating the time for the next wakeup of the task?

Kind regards,

Kai Liebich
0 (0 투표)
Stability of time ticks on the XDK
응답
15. 11. 27 오후 12:38 as a reply to Kai Clemens Liebich.
Kai,

If Manfred won't respond till Monday, I'll write him an email asking him to share his solution.

Hope this helps.

- Florian
0 (0 투표)
Stability of time ticks on the XDK
응답
16. 2. 23 오후 1:56 as a reply to Florian Harr.
Hi!
As the stability of time ticks is a central topic within many XDK applications I would like to bring up this topic again. If there is any explanation or solution you can share, it would be very helpful!

Thank you very much in advance.

Andre
0 (0 투표)
RE: Stability of time ticks on the XDK
응답
16. 6. 3 오후 12:00 as a reply to Andre Schwarzmeier.

Hello guys,

sorry for the long delay. Fortunately I can bring some light in the situation.

The freeRTOS time is based on a low frequency external crystal oscillator with 32.768kHz. One tick period is 1 ms. There only occurs a 1-2% tolerance between two ticks. Depending on this the timer ticks of the XDK are measured stable.

Therefore the described issues can have another reason depending on the used tasks. The issue with the SD card example occurs with the additional runtime of the application. This parameter can also be configured in the logfile of the SD card.  Therefore both runtime of the application and sample rate of the sensor conclude to the occuring timestamp of the measurement. This implicates the discrepancy of the timestamp accuracy.

For more information about that issue you can find the proper explanation here.

I hope this brings some clarity.

Kind regards,
Franjo

+1 (1 투표)