<div> <br></div><blockquote style="padding-left:5px;margin-left:5px;border-left:#a0c6e5 2px solid;margin-right:0px"><br><div class="gmail_quote">On Tue, Sep 20, 2011 at 11:17, Gong Ding <span dir="ltr"><<a href="mailto:gdiso@ustc.edu">gdiso@ustc.edu</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex;">
<div>Thanks. Any recommends for solving this type of problems? <br>It's our bottleneck now.<br>Direct solver works, but not efficient for large 3D problem.<br>Krylov space based solver seems not rubost enough. <br>A more exact ILU preconditioner may help? </div>
</blockquote></div><br><div>You can use domain decomposition with the jump aligned to subdomain boundaries. There are a variety of schemes with convergence rate independent of the jump, from both the overlapping Schwarz and Neumann-Neumann families.</div>
<div><br></div><div>Can you describe the 3D problem in more detail?<br><br>The problem arising from a CMOS inverter, or any CMOS gate.<br>For the inverter, there are PMOS and NMOS connected together.<br>When the PMOS and NMOS are both closed -- means nearly no current pass though the inverter.<br>the connection metal will see both huge resistance at VCC and GND end. <br> <br> VCC connection metal GND<br> ____ PGate _____________________ NGate _____<br> | ____________ | | ____________ |<br> __|________________|__ __|________________|__ <br> P | | PSource NDrain | | N<br> ----- ----- ----- -----<br> NSub PSub<br><br>In fact, I had already add some leakage current to the inverter -- can not be more due to accurate limit.<br>The conductance of metal still 1e6~1e10 times larger than "effective" conductance in the NMOS and PMOS.<br><br>Can anyone give some detailed suggestion?<br><br>Thanks.<br><br></div>
</blockquote><span ></span>