Skip to Content

Register Login

Industry Online Support

Technical Forum

Navigation

Advanced - adaptive smoothing filter

Created by: Pandiani at: 9/6/2008 9:48 PM (11 Replies)

Rating (3)

Thanks 3

Result pages:: |; 1 | 2; |

12 Entries

Entries per page: 10 | all

9/6/2008 9:48 PM	Rate (3)
Pandiani Bronze Member Joined: 10/31/2005 Last visit: 9/4/2024 Posts: 366 Rating: (31)	Hello to all, earlier this year I have written a small SCL implementation of more advanced filter for smoothing analog values. A few people tested it and some liked it and some didn't. I'd like you to read these in the attached zip file, test it and tell me what you think. In any case, I think you'll find it useful for certain applications. It is more advanced and powerful than classical first order filter. In the .zip file there is a small article that describes what I've done in SCL code. There is also Scilab script. Scilab software is free alternative to more powerful Matlab and can be downloaded from www.scilab.org I'd like to read you feedback. Attachment Smoothing filter.zip (1170 Downloads)
	Last edited by: Pandiani at: 9/6/2008 10:54 PM
Suggestion To thank Quote Answer
This contribution was helpful to 3 thankful Users Andruxa Andrey_Afonin yxgxt

9/6/2008 10:29 PM	Rate (0)
Gebhardt Karl Platinum Member Joined: 7/16/2006 Last visit: 1/2/2024 Posts: 1502 Rating: (176)	Thanks I will testing
	Karl
Suggestion To thank Quote Answer

9/7/2008 9:29 PM	Rate (0)
Kampf-Pasman Posts: 75 Rating: (13)	I will test it also Thanks!!
Kampf-Pasman Posts: 75 Rating: (13)
Suggestion To thank Quote Answer

9/9/2008 7:05 AM	Rate (0)
Aret Posts: 2348 Rating: (264)	Hi, Pandiani! It's very nice, that you decided to share your code. Here are some suggestions: You can exclude low and high limit deviation. 63% shown in pdf file is (1-1/e). Suposeyouhaveinputsignalleapfrom0to1. Youcanmakefilterso,that"afterxsecondsfromtheleapoutputwouldbey%ofleapvalue". I use for these things this code Smooth_Rate :=EXP(LN(Smooth_percentage)Main_cycle_time/(Ts+Main_cycle_time)); result_value := result_value Smooth_Rate + temp_value * (1 - Smooth_Rate); Smooth_percentage is 0.01 (that means after TS seconds output would be 99% (because 1-0.01) of the leap". temp_value - current value from IO, result_value - value, that have been smoothed. Main_cycle_time is OB1 scan time, Ts - filtering time.
Aret Posts: 2348 Rating: (264)
Suggestion To thank Quote Answer

9/9/2008 8:48 PM	Rate (0)
Pandiani Bronze Member Joined: 10/31/2005 Last visit: 9/4/2024 Posts: 366 Rating: (31)	Aret, yes 63% is actually time constant of first order filter. It can be calculated like 1-1/e. I intentially place deviations (hi limit and low limit) and it can be adjusted at will. However, user must first observe how signal is changing in steady state when there is only process and measurement noise and set these limits accordingly. Classical first order filter has one drawback, it acts on the signal in the same way regardles of magnitude of a change. For example if steady value is fluctuating +-0.1 around value 2 (max 2.1, min 1.9), than in case of step change from 2 to 3, filter will not act because this is obviously process dynamics and not random noise. Classical smoothing filter would provide response with same time constant no matter how change is big. This can sometimes hide very fast changes in the signal's amplitude which is not always desirable.
	Last edited by: Pandiani at: 9/9/2008 8:54 PM
Suggestion To thank Quote Answer

9/11/2008 7:59 AM	Rate (0)
Aret Posts: 2348 Rating: (264)	Agree, butwithproperfilterparameters(ex.99%inmycase)systemreactsquite fastonsignalchange(ofcourseifTsisbigenoughitwon't-butbisTsasumesstrongsmoothing). Nevertheless your addon is also nice feature that can used for sure:^)
Aret Posts: 2348 Rating: (264)
Suggestion To thank Quote Answer

9/12/2008 12:58 PM	Rate (0)
Remark Bronze Member Joined: 9/1/2006 Last visit: 8/1/2024 Posts: 193 Rating: (13)	Thanks for sharing the procedure, i'll see if i can learn something from your code. grtz, Remark

Suggestion To thank Quote Answer

9/12/2008 3:03 PM	Rate (0)
Pandiani Bronze Member Joined: 10/31/2005 Last visit: 9/4/2024 Posts: 366 Rating: (31)	You're welcome, I'm actually surprised there were not many comments on my [url=/tf/WW/en/Posts/20991#top]other project:[/url] I'm not very familiar with WinCC or WinCC flex, but when I have time I'll try to place couple of trends so one can learn about PID by changing its parameters and observe process response on trends.It is also posssible to change dynamics of process by changing its pramameters (first order and second order systems can approximatley describe my technological processes). Of course there are people here whocould do it in no time using only left hand :) Regards, Pandiani
	Last edited by: Pandiani at: 9/12/2008 3:05 PM Last edited by: Pandiani at: 9/12/2008 3:04 PM
Suggestion To thank Quote Answer

9/12/2008 5:34 PM	Rate (0)
Abdelaal Posts: 76 Rating: (16)	dear Pandini, Great work , I liked it so much thanks regards Abdelaal
Abdelaal Posts: 76 Rating: (16)
Suggestion To thank Quote Answer

4/30/2010 8:51 AM	Rate (0)
Andruxa Posts: 23 Rating: (0)	Thanks, it that are necessary for me!
Andruxa Posts: 23 Rating: (0)
Suggestion To thank Quote Answer

|

Share this page:

Share this page on...

Follow us on