| Author |
 Topic  |
|
|
klaus.joehnk
Germany
3 Posts |
 Posted - 01/21/2003 : 09:22:12 AM
|
Hi,
there seems to be a flaw in the gridding routine.
A data set i am trying to convert is given as
xyz-columns with x-range ( 0 , 1128.5 )
and y-range ( 0 , 1458 ). Applying the
Convert to Matrix -> Random XYZ -> Correlation
function using a matrix size of 100 x 100
(or any other) results in a matrix with a
dimension set to 1130 x 1460. First flaw: this
does not correspond to the original range. But even worse,
it seems that the gridding routine extended the range
internally to about 1200 x 1600 in contrast to the dimension set.
In this way the contour plot does not show the true data but a
scaled version, where the scaling is only known to Origin's
developers. At least I want to know this scaling number.
To make things mysterious: If I rescale the original xyz-data to
an (integer) range from 0 to 1000 everything is ok.
No range extension, no rounding in dimensions.
Might it be possible, that there are simple but serious
rounding errors in the calculation of grid spacing??
Klaus
|
|
|
greg
USA
1378 Posts |
Posted - 01/21/2003 : 2:31:44 PM
|
We do not set the XY mapping to precisely your minimum and maximum values, but this is hardly a flaw.
I don't know where you are getting the 1200 x 1600 numbers from and there is no scaling so I can't give you a scaling number.
This question should be sent to tech@originlab.com with some illustrative data and graphs ...
|
 |
|
|
edgar.kaiser
Switzerland
29 Posts |
Posted - 01/27/2003 : 03:58:15 AM
|
Hi Klaus,
I tried Origin's built-in functions (all available) for gridding random xyz-data and had many problems. Possibly because my data is rather noisy I didn't get a good representation.
Finally I switched to OriginC and the NAG-functions for 2D B-splines. This works fine for me. Also I prefer the B-spline approach because it performs a Least Square Fit on the data and not an interpolation as most built-in gridding does.
Regards,
Edgar |
|
|
| |
Topic |
|
|
|
Gridding randomly distributed xyz-data flawed