Intro to the Simulation study

In short, I have learned that to design a power circuit right, you have to use the SPICE tools of the PROs [i.e., Texas Instruments and Analog Devices]. This is because their tools will have the proper model for the components they sell and the supporting components they use to test their devices. So their SPICE simulations will be far better than the ones that average engineers can do by importing models or making pseudo models of the components they plan to use.

The following are the simulations “learning nuggets” I have come across with that out the way.

Issue Solved #1: External Reference too noisy

In the beginning, it seemed that hooking Vref_ext as a voltage divider and voltage buffer would be a good idea. However, it turns out that the noise of Vin and voltage buffer “noise” can get in Vout. So buffer would not be a good idea. So Vreg_ext had to be a linear regulator.

Issue Solved #2: LT1575 with “virtual ground”

During simulation, it was found that the gate driver is not directly connected to the ground; therefore, the gate will always be held above ~2.0 Volts. Unfortunately, this will not allow you to output a voltage below 1.2V even if you use the “TI’s external voltage reference” modification.

LT1575 virtual ground

So I will have to use a part, not from analog devices. However, the MIC5156 does have a gate driver capable of going down to 1 volt at the gate voltage. And even though Vout_low is not zero volts, there are plenty of MOSFETs with a gate threshold above 1 volt.

Issue Solved #3: Proper Capacitor

During the simulation, I discovered why the Tantalum capacitor is bad for power supplies. The reason is that these capacitors have an “equivalent series resistance” around a few ohms. So one can only draw a few amps with a significant voltage drop. That is why you should use Aluminum Electrolytic Capacitors since their ESR is in the milliohm range.

Issue Solved #4: Power filter

For the reference voltage, I wanted to pre-clean the ripples from stage 1 to ease the regulator’s job. Moreover, I found that active filter #1 was the best choice of the three proposed designs.