## Spectra Comparisons

02-Analytical capabilities
mwoods
Posts: 12
Joined: 10 Dec 2010, 15:21

### Spectra Comparisons

In the Spectra Comparisons tab, what algorithm is used to compute "Absolute" and "Relative" differences between record spectra and the target spectrum? My target spectrum has only 20, unequally spaced ordinates, compared to the many more, equally spaced (0.02 sec) ordinates that SeismoSpect computes for my individual records. But even at the matching ordinates, I can't figure out how SeismoSpect is computing the differences. For example, here are the results of comparing Sa for a particular record ("GLFE1FN") to my target Sa at 0.01 second period:

Code: Select all

``````period  GLFE1FN  Target          Dif1         Dif2    Dif3
0.01    1.251    1.14747464      0.118792725  0.0804  0.0687
``````
Dif1 = abs(Target - GLFE1FN)
Dif2 = "absolute difference" calculated by SeismoSpect
Dif3 = "relative difference" calculated by SeismoSpect

I've tried several different ideas for "Dif1" but I cannot reproduce either Dif2 or Dif3.

seismosoft
Posts: 1211
Joined: 06 Jul 2007, 04:55

### Re: Spectra Comparisons

Dear mwoods,

We have checked things, and the programs does seem to be making the right calculations, which are:
- Absolute differences: ABS (target_spectrum - computed_spectrum)
- Relative differences: ABS (target_spectrum - computed_spectrum)/target_spectrum

Since your target spectrum has unequally spaced ordinates, the program is interpolating such values in order to carry out the comparisons with the computed spectrum. Please check that the interpolated values correspond to what you were expecting (you can obtain (and copy) them in the 'Target Spectrum' dialogue box) and use such values to carry out your manual calculations.

Seismosoft Support
mwoods
Posts: 12
Joined: 10 Dec 2010, 15:21

### Re: Spectra Comparisons

quote:Please check that the interpolated values correspond to what you were expecting (you can obtain (and copy) them in the 'Target Spectrum' dialogue box) and use such values to carry out your manual calculations.

The example I provided is at a non-interpolated ordinate excerpted from a spreadsheet I used to make my own direct calculations. So, @0.01 sec, the absolute difference between my target (1.14747) and the calculated spectrum (1.251) is 0.11879, yet the absolute difference reported by SeismoSpect is 0.0804. I've made a point here of using a non-interpolated case. It still seems to me that my problem remains. Perhaps I should send you my time history and my target spectrum?