I've built a power-stabilized load for batteries (tester)
It switches ~16V using powerful NFET's at 3.9khz, ~30A (yes, over 400W)
The problem is, that averaging all the hall-effect current sensor samples does not give perfect result.
And so I'd like to smooth out this load.
one way would be to know what L&C I need, the other way, would be to build it using analog electronics
(like very powerful transistors, and have proper heat dissipation for them - keeping them as a resistive load.)
The amount of power you have to filter has no effect on the calculations you have to make.
The values of inductor and capacitor determine the roll off frequency, that is the frequency where the noise (AC signal) drops by half.
The formula is Fr = 1/ Sqr(LC)
Where Fr is the roll off frequency you want L is the inductor value in Henries and C the capacitance in Farads.
The filter is a second order filter and falls of at a rate of 3dBs per octave. To make it drop of faster you need a higher order filter.
Grumpy_Mike:
I was quoting power roll off not voltage roll off,
Sorry Mike but I hope I'm not making you even more grumpy!
Roll off of second order LP filter: If you look at it simplistically the impedance of the inductor doubles for each doubling of frequency and the impedance of the capacitor halves so the voltage attenuation is 4 times and the power attenuation is 16 times. 4 x voltage is 12 dB as is 16 x power.
AndreK:
I've built a power-stabilized load for batteries (tester)
It switches ~16V using powerful NFET's at 3.9khz, ~30A (yes, over 400W)
The problem is, that averaging all the hall-effect current sensor samples does not give perfect result.
And so I'd like to smooth out this load.
What load? You've not explained what circuit you have.
one way would be to know what L&C I need, the other way, would be to build it using analog electronics
If you just want to low-pass filter the output of hall sensors then an RC anti-aliasing filter
will probably do - thereafter you can filter in the digital domain.
(like very powerful transistors, and have proper heat dissipation for them - keeping them as a resistive load.)
The output ciccuit was 7 paralelled 2 Ohm 100W resistors fedby two IRFP064N , driven by ir2125.
yes, it is really targeted for much more than 400W.
I say was, because I never were satsfied with the ripple.
So I switched to a linear resistive load, using a smoothed out ~31khz PWM, controlling an 2N3055 , which controls 4x MJ11015 and a similar resistive load.
Now the darligtons do most of the dissipation, and so far I had luck keeping them (water)cooled.
The N-FET solution was much less complicated, and had no chance of a thermal runaway.
When I say thermal runaway, I don't really mean current going out of control, because the current sensor feedback would make MCU reduce the base current - I am a little conserned about a scenario where the hottest transistor may get too much load on itself, burning itself.
Adding significant emitter resistors is no good for the max current, so I am aiming for a even cooling and tight mounting which should make those 4 always be in the same temperature range.
You would have needed large and expensive components to filter the 400W line
itself at that switching frequency. Something more like 20kHz would have helped
bring the cost down.
Personally I would have tried using MOSFETs in source-follower mode for linear
power regulator, one load resistor per MOSFET - no thermal runaway issues and
simpler to drive.