REXYGEN Community Forum

stepan.ozana

Member

Registered: 2016-10-06

Posts: 34

Initial values of trend

Hi,
I uploaded this project:   project_oza77.zip

I wonder why TRND.u1 does not start at y0=pi as predefined in SINT2.
It seems to be  more or less -pi for some reason.
By coincidence I found out that for  DER.n=2 and DER1.n=2  (by default in the project DER.n=3 and DER1.n=3) it seems to be working properly (starting from pi)
I understand that change of some parameters may have influence on time waveforms, but it is strange that it affects starting points of trends for different variable in the projects.
What am I doing wrong? Can LPBRK blocks be the reason? Can I somehow trace the variable in more details (such as adding to System Log or so)?
Thank you for possible suggestions.

Last edited by stepan.ozana (2016-12-16 05:23:37)

tomáš čechura

Administrator

Registered: 2016-02-22

Posts: 371

Re: Initial values of trend

Hi,

I tried to reproduce your issue but with no luck. Are you sure that the trend you are looking at contains samples from the very beginning?
Your task period is set to 5 ms, TRND.l = 1000 (number of samples in buffer) - it means that trend has only 5 s data history. The value you are looking at (TRND.u1) changes from +pi to approx. -pi in about 1 s. Therefore if you open trend in approx. more than 6 s from execution start you will see exactly the situation you described in your post.

This is trend screenshot from my simulation:

Please, can you assure that your results are different? (e.g. by enlarging the buffer size to have more time for capturing the valuable data).

As you can see during compilation - RexComp warns you that there are uninitialised feedbacks in your project. Add "LPBRK" blocks between System and Regulator on x1 and x3 connections.

Please let me know if you have any questions.

Cheers, Tomas

---

Raspberry Pi as a miniature industrial PC. Analog and digital I/O, RS-485 and 1-Wire bus.

stepan.ozana

Member

Registered: 2016-10-06

Posts: 34

Re: Initial values of trend

Hi, you are absolutely right about insufficient number of samples in trend - my fault, sorry.
Here are my results:

Waveforms still depend on DER.n and DER1.n, but this must be due to the nature of the system which is very complex, sensitive to the sampling period and also strongly nonlinear. I will have to dig in more into the problem and put LBBRKs into the scheme.
Thank you very much for help with this.

Last edited by stepan.ozana (2016-12-16 05:24:27)

tomáš čechura

Administrator

Registered: 2016-02-22

Posts: 371

Re: Initial values of trend

Hi,

I'm glad I could help. Good luck with your project.

Regards, Tomas

---

Raspberry Pi as a miniature industrial PC. Analog and digital I/O, RS-485 and 1-Wire bus.