Well, building a 100amp low ripple switcher is for sure a
project. It's one on my list for my solar panels (where I actually need about twice that). The inductor(s) required between the switching and the caps are pretty daunting however. It's no small thing to make a decent inductor that size that won't saturate or be super-lossy (skin resistance in the wire) so no one does that. As in good luck finding a big enough ferrite core (kilograms) and making your own litz (multi conductor for less skin effect loss) wire of the required size. I wrote a perl script to calculate the desired value for a buck switcher, based on that power supply book JohnF recommended (and I do too -- it's the straight skinny on these). That may or may not be enough C - at these kind of currents, it's more the ripple current rating of the caps than the uF -- the uF being a lot easier to get than to get it with low parasitic R and the resultant heating in the capacitors.
You obviously cannot hook a 20khz sq wave to a capacitor -- instant smoke someplace -- you know that, right?
Every example of that around here uses multiple sets of fets and inductors to get to those kind of current levels, not just one switch block/L/C -- you just can't get the parts to make that in a single anyplace. You can't use IGBTs reasonably with their very high saturation voltage drop -- well over a volt, which would be wasting 100w per device -- they are only used in high voltage designs where that drop isn't so significant as a fraction of total power. The much lower effective on resistance of FETs at low voltages make them the clear winner in a design where they fit.
It's only at mains voltages and above that the IGBTs show any advantage -- at 12v FETs are perhaps 10-100 times less lossy -- not a small factor. But at the high currents, it's going to be the required inductance that will be very very difficult to assemble.
Which is why a few people here, who all really know their stuff, say "Don't bother trying that". It's not that it can't be done. It's simply not a very good way, and will consume all too much time and resources when there are much easier ways -- so you can get on with whatever the actual experiment is quicker. Capacitors of multiple farads are now available for super car audio systems that draw kilowatts at 12v and exceed what the car electrical system can do on the peaks. I had one such system in a car, which required those, they work fine.
- Code: Select all
#!/usr/bin/perl
#use Tk;
use strict;
#Compute L for a buck switcher, for min current 1/10 max, note no input error checks
# can't use warnings as we do something a little slungy to have default inputs.
my $frequency = 30000;
my $Vin = 120 * 1.414; # set some defaults
my $Iout = 10;
my $T = 1/$frequency;
my $Vout = 24;
my $L;
my $temp;
my $Mh;
print ("Input Voltage? ($Vin) >");
chomp ($temp = <STDIN>);
if ($temp + 0 > 0) {$Vin = $temp;}
print ("Output Voltage? ($Vout) >");
chomp ($temp = <STDIN>);
if ($temp + 0 > 0) {$Vout = $temp;}
print ("Current? ($Iout) >");
chomp ($temp = <STDIN>);
if ($temp + 0 > 0) {$Iout = $temp;}
print ("Frequency? ($frequency) >");
chomp ($temp = <STDIN>);
if ($temp + 0 > 0) {
$frequency = $temp;
$T = 1/$frequency;
}
print ("Vin is now $Vin\n");
print ("Vout is now $Vout\n");
print ("Iout is now $Iout\n");
print ("Frequency is now $frequency\n");
print ("Time is now $T\n");
$L = (5 * ($Vin - $Vout) * $Vout * $T) / ($Vin * $Iout);
print ("\nL is $L Henry.\n");
$Mh = $L * 1000;
print ("L is $Mh millihenry\n");
Posting as just me, not as the forum owner. Everything I say is "in my opinion" and YMMV -- which should go for everyone without saying.