Compare with SAVA

[vkazei]

compare seismograms and snapshots

• turning off the boundaries in homogeneous orthorhombic medium
• in two halfspaces
• in complex model - e.g. ortho time-lapse SEAM II

[vkazei]

[vkazei]

Isotropic case works

[vkazei]

[vkazei]

free surface...

[vkazei]

I tried to reduce the number of multiplications in the time loop, but the effect seems to be negligible.
SAVA does it this way:
/* save inverse density values /
for (i=nx1;i<=nx2;i++){
for (j=ny1;j<=ny2;j++){
for (k=nz1;k<=nz2;k++){
c1111[i][j][k] = DTc1111[i][j][k];
c1122[i][j][k] = DTc1122[i][j][k];
c1133[i][j][k] = DTc1133[i][j][k];
c1123[i][j][k] = DTc1123[i][j][k];
c1113[i][j][k] = DTc1113[i][j][k];
c1112[i][j][k] = DTc1112[i][j][k];
c2222[i][j][k] = DTc2222[i][j][k];
c2233[i][j][k] = DTc2233[i][j][k];
c2223[i][j][k] = DTc2223[i][j][k];
c1322[i][j][k] = DTc1322[i][j][k];
c1222[i][j][k] = DTc1222[i][j][k];
c3333[i][j][k] = DTc3333[i][j][k];
c2333[i][j][k] = DTc2333[i][j][k];
c1333[i][j][k] = DTc1333[i][j][k];
c1233[i][j][k] = DTc1233[i][j][k];
c1123h[i][j][k] = DTc1123h[i][j][k];
c2223h[i][j][k] = DTc2223h[i][j][k];
c2333h[i][j][k] = DTc2333h[i][j][k];
c2323h[i][j][k] = DTc2323h[i][j][k];
c1323ha[i][j][k] = DTc1323ha[i][j][k];
c1223ha[i][j][k] = DTc1223ha[i][j][k];
c1113h[i][j][k] = DTc1113h[i][j][k];
c1322h[i][j][k] = DTc1322h[i][j][k];
c1333h[i][j][k] = DTc1333h[i][j][k];
c1323h[i][j][k] = DTc1323h[i][j][k];
c1313h[i][j][k] = DTc1313h[i][j][k];
c1213ha[i][j][k] = DTc1213ha[i][j][k];
c1112h[i][j][k] = DTc1112h[i][j][k];
c1222h[i][j][k] = DTc1222h[i][j][k];
c1233h[i][j][k] = DTc1233h[i][j][k];
c1223h[i][j][k] = DTc1223h[i][j][k];
c1213h[i][j][k] = DTc1213h[i][j][k];
c1212h[i][j][k] = DTc1212h[i][j][k];

			if (rhoijpkp[i][j][k] != 0.0)
				rhoijpkp[i][j][k] = DT/rhoijpkp[i][j][k];
			if (rhoipjkp[i][j][k] != 0.0)
				rhoipjkp[i][j][k] = DT/rhoipjkp[i][j][k];
			if (rhoipjpk[i][j][k] != 0.0)
				rhoipjpk[i][j][k] = DT/rhoipjpk[i][j][k];
		}
	}
}

[vkazei]

the effect of removing DT multiplication from time-stepping proved to be negligible

[vkazei]

Wrap in a test?

[dmitry-kabanov]

I found it difficult to reproduce the figures above, as the config files in the Dropbox folder apparently are for other simulations.
I ended up with the test with the default SAVA parameters (one layer, ortho case) (it is test # 9).