A semiconductor optical amplifier configuration for reducing the output noise in an all-optical space-switching structure
✍ Scribed by M. Marcos Garcia; D. Uttamchandani
- Publisher
- John Wiley and Sons
- Year
- 1997
- Tongue
- English
- Weight
- 234 KB
- Volume
- 16
- Category
- Article
- ISSN
- 0895-2477
No coin nor oath required. For personal study only.
✦ Synopsis
dense grid for the case of surface length L s 30. The CPU advantages get better for long surfaces. This is because the dense grid is still used for distance separations less than r , f and r s 4 is substantial compared with the surface length f Ž . of L s 30 see Fig. 4 . To show the computational advantage for longer surfaces, we show in Figure 5 the results of a bistatic scattering coefficient of a surface length of L s 120. For the single grid, we use the coarse grid of the usual n s 10 points per wavelength, and for the PBTG, we use the d two grids of n s 10 and n s 30. The single coarse grid is dc d g not accurate, as shown in the results in Figure 5. In Table 2, we show the comparisons of CPU. We note that the ratio of CPU is 9545.4r6586.6 s 1.45. This means that the PBTG provides accuracy with only a 45% increase in CPU over the inaccurate single coarse grid. On the other hand, the estimated CPU for a single dense grid of n s 30 corresponding d to 7200 surface unknowns is about nine times more. This single dense grid of n s 30 is not computed because of the d large memory requirements. On the other hand, the PBTG of n s 10 and n s 30 can be performed as indicated be-
cause the PBTG also has a substantial savings in memory.
VII. CONCLUSIONS
In this letter, we have developed and applied the PBTG method to address the dense grid requirement for dielectric surfaces with a large lossy dielectric constant. The method saves both CPU and memory, and provides the required accuracy. We are presently combining it with the canonical w x grid method 5᎐7 and extending the approach to threedimensional problems.