Final design considerations for LM5143A-Q1 buck converter. Are these features needed?

Question

After a few days of learning how to make a buck converter using the LM5143A-Q1 TI IC, it's finally becoming a reality! This is the final design, highly inspired from page 30 in the datasheet, some values are not yet added. There are just a couple of issues I'm not sure how to handle in the final design:

1. Are more input/output capacitors better? In the datasheet, for input/output capacitance they mention >= than a value we calculate. In our case: Cin>=45uF, Cout>=1954uF (1.95mF). If that's the minimum, can we add more? If yes, how much more? 1.5X , 2X , 3X the minimum capacitance? I'm using Al-Polymer and Ceramic caps at the output since the switching frequency is 100kHz and only ceramics at the input.

2. Is it normal for the input capacitance of a buck converter to be lower than the output capacitance? I find it semi-logical, assuming the power supply doesn't have a high ripple voltage?

3. In the datasheet, there's this DITH (Spread Spectrum Frequency Modulation), from what I understood, it modulates the oscillator frequency by +-5%, which I don't see the purpose of? Do we need this DITH? In the datasheet, they always pull it down to GND with a 10nF capacitor, but I've asked a friend about this, and he said DITH can also NOT be connected to anything. Why is that? Also, calculating FMOD from the formula with CDITH=10nF, we get 10.5kHz, which is indeed under 20kHz. So should I just not bother and use a 10nF as well to pull it to GND?

4. As this is a double-phase buck regulator (since I_OUT = 31A), I was wondering, is making it work really as simple as connecting the VOUT1 to VOUT2 together, and then stacking all the output capacitors at the end? Is that all that needs to be done?

Answer 1

1. Adds cost, weight and board space but reduces ripple.
2. Yes, as you often have requirements on output ripple but with your buck being fed by some intermediate voltage, it's more your own problem how much ripple you can have there. Also, since you are dropping voltage with a buck, the input have higher voltage so a given capacitance will hold more energy on the input than the output of a buck.
3. Helps with EMI. Instead of one potentially nasty spike going above the limit line, you smear it out over a larger frequency range. Whether you need it or not depends on your application, rules, implementation and layout.
4. I didn't read the datasheet in depth, but should be yes.