| T O P I C R E V I E W |
| tlab |
Posted - 12/21/2004 : 1:24:13 PM
Origin Version (Select Help-->About Origin): 7
Operating System: Win2k
From a 300mm sample I measured 50 points. Then I got 50 sets of (x, y, z). I can convert it into a matrix. Is it possible to interpolate this matrix so that I can get a 300x300 matrix and see feature of the whole sample?
|
| 9 L A T E S T R E P L I E S (Newest First) |
| easwar |
Posted - 01/08/2005 : 11:13:02 AM
Hi,
So what you are looking for is to restrict the interpolation to only the "data limits" and not to perform any interpolation "outside of the data". This feature is currently not available in Origin, and since the interpolation is done over the entire matrix, the contour lines and levels at the "data boundaries" can be different from other products that restrict the interpolation to be only within the data limits.
Easwar
OriginLab
|
| tlab |
Posted - 01/07/2005 : 5:46:43 PM
the above is what I got from Origin. I try to remove its four corners.
this is what I got from Igor Pro.
|
| tlab |
Posted - 01/07/2005 : 5:01:48 PM
Mike,
Here is my datasheet:
X Y Z
0 120 0.515
-65 90 0.516
-25 100 0.512
25 100 0.508
65 90 0.51
-90 65 0.515
-50 75 0.514
0 75 0.51
50 75 0.512
90 65 0.507
-115 40 0.515
-75 50 0.513
-25 50 0.512
25 50 0.508
75 50 0.511
115 40 0.506
-120 25 0.518
-75 25 0.515
-25 25 0.514
25 25 0.51
75 25 0.511
120 25 0.506
-125 0 0.519
-75 0 0.516
-25 0 0.515
25 0 0.509
75 0 0.512
125 0 0.508
-120 -25 0.518
-75 -25 0.517
-25 -25 0.514
25 -25 0.511
75 -25 0.513
120 -25 0.507
-115 -40 0.519
-75 -50 0.517
-25 -50 0.515
25 -50 0.511
75 -50 0.514
115 -50 0.509
-90 -65 0.518
-50 -75 0.517
0 -75 0.513
50 -75 0.515
90 -75 0.511
-65 -90 0.517
-25 -100 0.517
25 -100 0.512
65 -90 0.514
0 -120 0.513
After converting to matrix by Random XYZ, for example, choose "Correlation", set parameter as 301, 301, 1, 0.2. Then OK. Now you can see Origin creates a square matrix. Those four corners are not apparent.
I am not sure how the real sample looks like since I can not measure it by 1mm step. But current interpolation result by Origin looks different from the result produced by another software, Igor Pro. This Igor Pro can create a matrix by only interpolation ( that means only points between two of original data points). Especially on the edge part. So I wonder if the extrapolation algorithm to get those value at corner in Origin makes this difference.
Thanks.
tlab
|
| Mike |
Posted - 01/07/2005 : 4:46:43 PM
tlab, I responded too quickly to your question. As a test exercise, I created a grid in XY space, drew a circle on it, then created a data point at each grid intersection. When plotted as a 2D scatter plot, it looks like a circle. Then I added a third (Z) column to the worksheet and filled it with random numbers. I then converted the XYZ data to matrix data, using each of the five available methods, and plotted the data as a contour plot. In each case, the shape of the "circle" was lost; the distortion was most severe where an extreme value lay near the edge of the "circle." So it seems as though we need to come up with a way to show some sort of base value in areas that lie outside your measured data points.
Mike
OriginLab
Edited by - Mike on 01/07/2005 4:47:50 PM |
| Mike |
Posted - 01/07/2005 : 2:42:35 PM
tlab, are your XY pairs derived from a Cartesian or Polar coordinate system?
I would have expected that your "circle" would still be visible within the bounds of the rectangular matrix. Undefined values in the matrix "corners" should be apparent to you and represented by some base value (0?).
Is this not the case?
Mike
OriginLab |
| tlab |
Posted - 01/07/2005 : 2:18:18 PM
Some new problems...
Actually what I measured are 50 points distributed uniformly as a circle. By choosing Random XYZ, I got a square matrix, i.e. Origin also gave me the outside of the region which I measured. Do you think Origin got those data by extrapolation? But what I want is only interpolation between two data points. Is it possible to get interpolation only?
Thanks.
tlab
|
| tlab |
Posted - 12/22/2004 : 7:11:59 PM
Thanks guys. I just did it as you said.
tlab
|
| easwar |
Posted - 12/21/2004 : 1:50:26 PM
Hi,
Is your measured data regularly spaced in x,y? If yes, I presume you are converting to matrix using Edit->Convert to Matrix->Regular XYZ menu item. If this is the case, after the matrix is created, you can use the matrix menu item Matrix->Expand, which will use bilinear interpolation to add more points in between.
If your measured data is random, you can also pick a gridding method such as Correlation, Quick Correlation, or Weighted Average under Edit->Convert to Matrix->Random XYZ menu and then specify the dimension of your matrix in the dialog, and then the gridding process will create directly a matrix with finer step size in x,y.
Easwar
OriginLab
Edited by - easwar on 12/21/2004 1:53:17 PM |
| Mike |
Posted - 12/21/2004 : 1:48:30 PM
Hi tlab,
Open your Help file (Help:Origin) and look for Index keyword = Expanding > matrix. I think this will give you what you want.
Mike
OriginLab |
