Designing a front-filter for a power supply.

[David Collier]

I have a DC-DC non-isolated power supply deign using the Micrel MIC2198BML

it takes in say 12V, and outputs 5V.

It has a "fixed" switching frequency of roughly 500KHz.

the input current is about 1A, output obviously 2A or so.

I'd like to design a front-end filter specifically to stop the 500KHz
frequency being fed back out to the 12V supply line, on both the live
and ground wires..

Trouble is, I started life as a programmer, and filter design is a bit
beyond me.

I could happily use a low-pass filter, with a cut-off well under 500KHz,
or a notch filter at 500KHz... it seems to be both would work.

Can anyone help by suggesting a usable circuit? I've done some goggling,
and come up with a lot on audio filters, but precious little on ones
rated at a few amps!!!

David

 

[Klaus Vestergaard Kragelund]

I have a DC-DC non-isolated power supply deign using the Micrel MIC2198BML

it takes in say 12V, and outputs 5V.

It has a "fixed" switching frequency of roughly 500KHz.

the input current is about 1A, output obviously 2A or so.

I'd like to design a front-end filter specifically to stop the 500KHz
frequency being fed back out to the 12V supply line, on both the live
and ground wires..

Trouble is, I started life as a programmer, and filter design is a bit
beyond me.

I could happily use a low-pass filter, with a cut-off well under 500KHz,
or a notch filter at 500KHz... it seems to be both would work.

Can anyone help by suggesting a usable circuit? I've done some goggling,
and come up with a lot on audio filters, but precious little on ones
rated at a few amps!!!

David

Hi

Your noise will be mainly differential, so a LC filter with for example
the sdr0805 from Bourns may be sufficient

But your input current of 1A is not enough information. You need to find
the peak current to decide for the relevant inductor

Then you need to supply information of what standards you must uphold

Regards

Klaus

 

[Nico Coesel]

I have a DC-DC non-isolated power supply deign using the Micrel MIC2198BML

it takes in say 12V, and outputs 5V.

It has a "fixed" switching frequency of roughly 500KHz.

the input current is about 1A, output obviously 2A or so.

I'd like to design a front-end filter specifically to stop the 500KHz
frequency being fed back out to the 12V supply line, on both the live
and ground wires..

Trouble is, I started life as a programmer, and filter design is a bit
beyond me.

I could happily use a low-pass filter, with a cut-off well under 500KHz,
or a notch filter at 500KHz... it seems to be both would work.

Can anyone help by suggesting a usable circuit? I've done some goggling,
and come up with a lot on audio filters, but precious little on ones
rated at a few amps!!!

Use a common mode filter and a capacitor.

 

[[email protected]]

These 3 links should tell you exactly what you are looking for........

http://www.beta-dyne.com/pdf/dcdc_an02_rev02.pdf
http://www.beta-dyne.com/pdf/dcdc_an15_rev01.pdf
http://www.pliki.jm.pl/karty/dc_dc_converters.pdf

 

[Mac]

Sorry to reply here. I can't see the OP.

You can use an LC filter similar to what you have on the output of the
switcher.

I did once neglect to filter the input on a switcher, and noise did couple
in to other boards through the power supply. It was basically square wave
noise at the switching frequency. This was a 5V to 3.3V converter
operating at around 2 amps output and 750 kHz.

With switchers, when the load current is steady, the duty cycle is a
function only of the input and output Voltages. This means you can pretty
easily figure out what the input current waveform will look like
(approximate it with a square wave) and thus conceptually test out
different values of L and C (also include the ESR) in a simulator or
spreadsheet.

HTH!

--Mac

 

[Ken Smith]

I have a DC-DC non-isolated power supply deign using the Micrel MIC2198BML

it takes in say 12V, and outputs 5V.

It has a "fixed" switching frequency of roughly 500KHz.

the input current is about 1A, output obviously 2A or so.

I'd like to design a front-end filter specifically to stop the 500KHz
frequency being fed back out to the 12V supply line, on both the live
and ground wires..

Trouble is, I started life as a programmer, and filter design is a bit
beyond me.

I could happily use a low-pass filter, with a cut-off well under 500KHz,
or a notch filter at 500KHz... it seems to be both would work.

Is your +5V output's ground returned to the +12V ground anywhere but at
the common connection of the power supply? If so your life is slightly
more complicated.

L1 L2 ------- L4
+12V -((((--+---((((---+---! DC-DC !---+---((((--+--+5V load
! ! ------- ! !
--- --- ! --- ---
C1 --- C2 --- ! --- C3 --- C4
! ! ! ! !
! -------+------- !
---------+---((((-----------+-----------------+-- 5V return
L3


Notice that the heart of the DC-DC converter is only connected directly to
the outside world at one point. It the 5V load is isolated and inside a
metal enclosure, it is likely that L3 and L4 can be deleted without a
problem.

Now for the simplifications that make your life easier:

We can assume that the +12V input has zero impedance and that the ripple
current into it is the specification that matters.

The allowed ripple current on the +12 in many times less than the input
ripple current of the converter.

The impedance of L1 is many times the impedance of C1. This includes C1's
ESR and ESL.

L3's impedance is similarly much larger than C2's.

Your DC input current on the DC-DC converter is 1A so I'd estimate that
you have a 0.5A RMS ripple current there.

Your ratios of values are huge so you can ignore phase relationships and
say things like:

The ripple in L2 = 0.5 * ZC2 / (ZL2 + ZL3)

The ripple in L1 = (ZC1/ZL1) * 0.5 * ZC2 / (ZL2 + ZL3)

Chances are C2 = (C1 + 0.1u)

The 0.1u is a high frequency bypass to help eat the higher harmonics.

Now you need to decide if this is thru hole or surface mount and a bunch
of issues like that and get out your inductor and capacitor catalogs.

Try www.coilcraft.com and look for a 5022 series inductor that can handle
the average current. Assume this for L1 and L2 and plug in the ZL1 and
ZL2 and solve for (ZC1 = ZC2). Look for a capacitor with this or less
impedance.