Bug in CEL?

Lance · June 29, 2009, 08:27

Hi,
I have a CEL expression that looks like:

Code:

80[Pa]+30[Pa]*(1+sin(5[Hz]*pi*t-pi/2))

and if I plot the expression during 20 s it looks like:

which isnt really what it should look like...rather, the waves should be equal in size.

Changing the time to 10 s gives the following plot:

It seems like there is something strange with CFX when it plot the expression. Anyone else found the same thing?

Rui · June 29, 2009, 11:47

Increase the number of points in the plot

ckleanth · June 29, 2009, 16:09

Quote:

Originally Posted by Rui

Increase the number of points in the plot

good one mate, didn't think of that one

Lance · June 30, 2009, 05:25

Quote:

Originally Posted by Rui

Increase the number of points in the plot

d'oh, didnt think of that

Thanks

Rui · June 30, 2009, 07:01

It was evident from your 2nd plot (the vertices in the curve are well visible).

The frequency of your expression is 2.5 Hz, or T = 0.4s. CFX uses, by default, 100 points to make the plot.
In the 1st plot (20 s), the points will be spaced by 20 s/ 99 intervals = 0.202 s. So, the plot shows approximately only 2 points for each period of the sinusoidal expression.

ghorrocks · June 30, 2009, 08:52

Hi,

Yes, this is classic frequency aliasing. Remember the Nyquist critereon (http://en.wikipedia.org/wiki/Nyquist...ling_theorem)? In short to resolve a given frequency you need to sample it at least at twice the frequency you wish to resolve. Frequencies you do not have enough speed to resolve will result in aliased frequencies at lower frequencies which are spurious and not real. It usually manifests itself as beats in ampiltude of a high frequncy wave - see Lance's plots for a classic example.

Glenn Horrocks

June 29, 2009, 08:27	Bug in CEL?	#1
Lance Senior Member Lance Join Date: Mar 2009 Posts: 669 Rep Power: 22	Hi, I have a CEL expression that looks like: Code: 80[Pa]+30[Pa](1+sin(5[Hz]pi*t-pi/2)) and if I plot the expression during 20 s it looks like: which isnt really what it should look like...rather, the waves should be equal in size. Changing the time to 10 s gives the following plot: It seems like there is something strange with CFX when it plot the expression. Anyone else found the same thing?

June 29, 2009, 11:47		#2
Rui Member Rui Igreja Join Date: Mar 2009 Location: Aveiro Posts: 68 Blog Entries: 1 Rep Power: 17	Increase the number of points in the plot __________________ Rui

June 30, 2009, 07:01		#5
Rui Member Rui Igreja Join Date: Mar 2009 Location: Aveiro Posts: 68 Blog Entries: 1 Rep Power: 17	It was evident from your 2nd plot (the vertices in the curve are well visible). The frequency of your expression is 2.5 Hz, or T = 0.4s. CFX uses, by default, 100 points to make the plot. In the 1st plot (20 s), the points will be spaced by 20 s/ 99 intervals = 0.202 s. So, the plot shows approximately only 2 points for each period of the sinusoidal expression. __________________ Rui

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Angle variable under CEL.	Gloria Gaynor	CFX	11	November 26, 2012 07:29
junction box routine and CEL function	bornspur	CFX	2	February 3, 2009 03:24
Rotate node via cel	Elian81	CFX	2	September 25, 2007 06:31
Bug reports	Mattijs Janssens (Mattijs)	OpenFOAM	0	January 10, 2005 11:05
Forum y2k Bug	Jonas Larsson	Main CFD Forum	1	January 5, 2000 11:22

June 30, 2009, 08:52		#6
ghorrocks Super Moderator Glenn Horrocks Join Date: Mar 2009 Location: Sydney, Australia Posts: 17,872 Rep Power: 144	Hi, Yes, this is classic frequency aliasing. Remember the Nyquist critereon (http://en.wikipedia.org/wiki/Nyquist...ling_theorem)? In short to resolve a given frequency you need to sample it at least at twice the frequency you wish to resolve. Frequencies you do not have enough speed to resolve will result in aliased frequencies at lower frequencies which are spurious and not real. It usually manifests itself as beats in ampiltude of a high frequncy wave - see Lance's plots for a classic example. Glenn Horrocks