Archived

This topic is now archived and is closed to further replies.

pilarsor

CFD - Cell Size h

Recommended Posts

pilarsor    130
I wonder, if reducing or increasing the cell size h in a cartesian grid would cause any changes in the numerical behaviour of my fluid simulation? (Keeping in mind, that the number of cells doesn''t change.) Cheers, Oliver

Share this post


Link to post
Share on other sites
pickups    122
ad far as i know, the simulation would be more accurate, since you sample the equations more often, but your final output will be smaler, or youll have to add cells.

Share this post


Link to post
Share on other sites
pilarsor    130
Thx, for your response.

But, the sample rate is still the same, because I assume, that the number of cells don't change. However, you are right, that my output gets tighter if I change h=1.0 to h=0.5, for example.

To be honest, I actually don't think, that the numerical behaviour will be affected, when changing the cell size h.

[edited by - pilarsor on August 7, 2003 3:55:20 AM]

Share this post


Link to post
Share on other sites
pickups    122
No, it would.

think you have a 1D function graph, and you want to compute its slope at a given point, and you dont how to analiticaly do that.
then you will discretize the function, sample it every h, and then just use:
slope = (y2-y1)/(x2-x1) to compute the slope, if your h is small, the derivative would be more precise.

Share this post


Link to post
Share on other sites
Guest Anonymous Poster   
Guest Anonymous Poster
There is a limited numerical range in a floating point number. If you shift your cell size, then you shift the point where your numbers go denormal, or go to infinite, compared to your cell size.

Here''s a good brain teaser:

What is the exact result of adding two 32-bit IEEE floating point numbers, where one has the value 1e10, and the other has the value 1e-10 ?

Share this post


Link to post
Share on other sites