Hello,
The buck converter equations typically take a
maximum output current value. Then, the calculated inductor value (as a rule of thumb) is typically good down to about 10% min. current.
You're correct in saying that if we expect, on average, a far lower output current, then the inductance should be raised. Given the same current expectations across rails, this is even more true for higher output voltages (e.g. 3V3 vs 1V0 for the same output current -> the 3V3 buck should use a larger inductance value). For BoM consolidation however, I ended up choosing the same value for all.
For your system: If you are expecting lower currents on average, then yes - increase the value of inductance. Keep in mind that - generally speaking - inductors with a greater inductance, given the same package, will have a higher DCR and a lower max./saturation current.
Given the real-world performance of the ZettBrett system, I haven't run into any issues using the lower inductance value.
However, this is due to the system drawing a fair amount of power (tested at 5V Vin) and current per rail. For example, idle: ~1W, Ethernet test: ~2.5W, and more for running PetaLinux.
There can be an increased output voltage ripple, of course due to the lower value inductors (can be combatted to some effect by using large value output capacitors, as is done in the ZettBrett). The max. ripple I'm measuring across all rails is about 35mV peak-to-peak.
Lastly, I'd suggest also checking out the schematic and power sections of a few other 'reference designs', e.g. from Digilent, Numato, etc. They use buck converters set to the same frequency (1MHz) and use similarly-sized inductors (Digilent also indicate current requirements per rail on the schematic):
Styx Z7
https://numato.com/help/wp-content/u...1/Styx_Sch.pdfBest,
Phil