Filtering ripples in power supply

[engr_joni_ee]

Hi, I am wondering which additional circuit can be used at the output of isolated DC-DC converter having 28 V output that can support up to 2 A but having output ripples of 200 mV pp. Our requirement for powering SiPM is to have ripples within 10 mV pp. Any idea which LC circuit can used to have stable output ? The peak current will be needed only for short time for example 10 u second, otherwise for other times the current requirement will be less then 100 mA.

 

[barry]

i= C*dv/dt

 

[KlausST]

Hi,

The power supply is 28V/2A?
And the load current usually is 100mA?
Just for about 10us there is current up to 2A?

For calculating the power (current) demand you should say how often these "10us" current peaks can happen (frequency).

If this is at a rather low rate, let´s say 10 times per second, then the average current is just 100mA + 1.9A * 10us * 10Hz = 100.19mA. (just 0.19mA added by the pulse)

I calculate this, because the 10us pulse easily can be supplied by a capacitor. So you may use a 28V/150mA power supply instead. The benefit is, that a 150mA power supply is likely to operation in continous switching, while a 2A power supply may go into burst mode operation. Burst mode will extremely increase the output ripple and to make it worse at a rather low ripple frequency. This ripple is hard to filter.

Filtering a continously switching power supply is way more effective.

***

Another aproach is to divide the path into a "clean" one and a "dirty" one. Where the clean one will be filtered or even analog post regulated.

Klaus

 

[dick_freebird]

You might prefer a LDO regulator with decent HF PSRR.
These are a thing, now. Their whole purpose is to clean
up switchers for fussy loads.

For light loads you might LC-filter, but anything you add
in series will degrade load regulation and load-step may
be compromised - you'd have to ensure that your close-in
filtering can supply the 200mA for the 10us at an acceptable
voltage sag across the pulse duration. If I called it 1V sag,
that would be a C=2A*10E-6s/1V or 20uF? That's not
terrible. An inductor with a sub-ohm DC resistance ought
to not bother DC level "too terribly".

You'd have to check out the box between fSW of the switcher
(which you must reject to the tune of 95%, so filter tau at least
2X, maybe 3X switcher period) and the rep rate of the pulses,
which can't be allowed to "staircase" the local supply downward,
meaning filter tau must be similarly a 1/2 - 1/3 fraction of that
rate. If you've got a 100kHz switcher and a 1kHz pulse rep rate,
you might just be able to fit an acceptable corner freq in there
at 10kHz or so. 100kHz switcher and 10kHz rep rate, that's
liable to bother one end or the other.

 

[FvM]

Should use a damped LC filter, C parallel RC. Optimal parameters depending on ripple frequency, pulse rate and maximal average current which haven't yet been specified.

 

[Calm down]

Firstly, you need a CLC filter circuit composed of polymer solid-state capacitors to filter out larger ripples, and then use a CLC circuit composed of common mode inductors to filter out noise. Then, use a linear regulator to improve current stability. If it is a high-frequency power supply, it is also necessary to reduce the IPK current to improve the linearity of the output current.