Devlog #23

The trick for the MOS capacitor worked. Making the electrons react faster put the slow wiggles at a speed the computer can handle, and the answers came out right. That FINALLY finished the AC phase. I also added noise, a tiny random jitter in current that every device has.

Next I added time. Now its possible to flip a diode from on to off and watch it recover. When you switch a diode off, it keeps conducting for a moment because of all the charge stored inside it. I compared that against DEVSIM, and our results were within 0.03% of each other which is quite frankly something im really proud off.

Then I added a feature where the simulator shows how inaccurate it may be. It now checks its own answer against a finer mesh and puts a little error bar on every point in the plots. It can also improve its own mesh by adding points only where they’re needed. On 5 of my 8 test devices it hit the target accuracy with fewer points than my hand-made meshes. I was aiming for around 6, so that’s still doable.

The newest addition is a bipolar transistor, where a small current into the middle layer controls a much bigger current through the whole device. I built a simple 1D version and a more realistic 2D one. The 1D version matches the textbook formula to within 0.03%. The best part is that its gain drops off at both very low and very high current, the way real transistors do, and I never told it to. It figured that out from the physics.