Design of input capacitors for switched-mode power supplies?

You did not provide any information that you did the tinyiest effort on your own. Please do


I did a quick internet search. About every semiconductor manufacturer provides such application notes.
Usually they are all good, I expect them to give similar results.

and thousands of other hits.

Also using simple capacitor equations (wikipedia and many many other sources) and Ohm´s law gives ripple values. No rocket science.

Additionally there are free circuit simulators .. and I´m sure semiconductor manufacturers will provide dedicated SMPS simulation tools.

Klaus

Suggest to answer the question yourself by setting up a simple simulation:
- grid voltage with some impedance, e.g. 50 uH
- bridge rectifier
- dc-link capacitor of variable capacitance
- DC load, either resistive, constant I or constant P, whatever fits your application best

Observe how capacitance affects Irms, harmonic current, dc-link ripple

You'll notice that it's not possible to achieve 3.3 kW at 16 A Irms due to limited power factor of non-PFC rectifier, except with very small dc-link capacitance.

FvM said:

You'll notice that it's not possible to achieve 3.3 kW at 16 A Irms due to limited power factor of non-PFC rectifier, except with very small dc-link capacitance.

Click to expand...

I know, but traders selling Chinese welding inverters in the EU don't know this.
I'm looking for an inverter welder and because they write so much crazy technical crap that it hurts, I choose based on the photos of the electronics inside.
Now I'm looking at an IGBT inverter without PFC at the input, there are two 680uF 450V capacitors behind the diode bridge. Switching frequency about 40kHz.
The ripple current of a quality capacitor of this size is under 3A, I would use at least 4 pieces, maybe even 6 pcs.
I thought of converting it to 3x230V 16A.
use 3 pcs of 1200V 50A diode bridges for three phases, this would fill C significantly faster thanks to the phase shift and could improve the shortcomings of the cheap Chinese solution.
There is no room for adding C, and then quality capacitors are expensive, diode bridges are small and cheap.

asrock70 said:

they write so much crazy technical crap that it hurts

Click to expand...

who wrote post#1?

Klaus

I'd begin with the idea that your upstream source
could be "totally inductive", and so input filter C
must be large and good enough Q (not too good,
mind you) that your "VIN" toggling will be a few
percent, "on" vs "off" switch. The filter has to
provide all the current during "on". Q=CV,
dQ=CdV, Q=imax*ton, check it out.

You can play with economics once you get good
behavior (like, you don't bump into UVLO at
max load for input sag during pulse from weak
input reservoir).