design EMI input filter for Dc-DC converter

Marc Sugrue

Joined Jan 19, 2018
203
How to design EMI input filter in dc-dc converter? any experience in this point?
The first thing is to define the peak load (in Ohms) and the frequencies your trying to filter. The peak load normally occurs at low input voltage & max load when the current is typically at its largest. The next step depends on whether you know how much attenuation you actually require and the type of noise your experiencing. A finger in the air approach would be to assume only differential mode noise is present and as a start point design a LC filter to give approx 40db Attenuation at the switching frequency. To do this you can choose a capacitor with an impedance of around 1/10 of the peak load impedance c = 1/(2*pi*f*(Zload/10)) and an inductor with an impedance of 10x the load at the switching frequency L =((10*ZLoad)/Ipk)/(2*pi*F). If you find you need more attenuation lower the filter frequency and you'll get more attenuation at the higher frequencies and to lower the attenuation you can increase the frequency

A more scientific approach would be possible if you have details of the test LISN (as this would help define the filter topology required) and actual data from a spectrum analyser relating to common mode and differential mode emissions (this would clarify the type of filter and how much attenuation is required for compliance).
 

Attachments

Last edited:

Thread Starter

mah

Joined Mar 15, 2010
393
The first thing is to define the peak load (in Ohms) and the frequencies your trying to filter. The peak load normally occurs at low
Would the emi occur at 30khz frequency? Is it a must to design EMI input filter? and how to measure the noise dB in the circuit? and
I couldn't get the same L from your formula L =1/ (2*pi*F*(10*ZLoad)) =1/(2*pi*100e3*10*6.5)=2.4e-8
 
Last edited:

Marc Sugrue

Joined Jan 19, 2018
203
9
Would the emi occur at 30khz frequency? Is it a must to design EMI input filter? and how to measure the noise dB in the circuit? and
I couldn't get the same L from your formula L =1/ (2*pi*F*(10*ZLoad)) =1/(2*pi*100e3*10*6.5)=2.4e-8
Would the emi occur at 30khz frequency? Is it a must to design EMI input filter? and how to measure the noise dB in the circuit? and
I couldn't get the same L from your formula L =1/ (2*pi*F*(10*ZLoad)) =1/(2*pi*100e3*10*6.5)=2.4e-8
Any frequency on a wire can be considered a conducted emission, whether or not you need to worry about it will depend on the type of product and the type of market its going into as this will define what the certification requirements will be. If high frequencies are on a long cable they can radiate also so best practice is to deal with the emissions early in the design lifecycle. As a side note whilst a filter is not always a must it can serve many other purposes such as protect the circuit from outside influence (such as voltage spikes and transients) not just prevent emissions from coming out from the product your designing.

In terms of formally measuring its typically done using a LISN and a spectrum analyser or through the use of LISN, a RF current probe and a spectrum analyser with the results normally provided in dBuV or dBuA. The method of getting the data is normally dictated by the market your looking to enter. You can do crude emissions measurement on an Oscilloscope with FFT transform capability too but the noise floor is normally factors from where it needs to be to do anything over 1MHz so if its bad on a scope its usually awful with an spectrum analyser. Without a LISN in place things can also become more problematic as noise from the equipment powering the product can also create false data so a good setup is advised.

As some examples of EMI, in military conducted emissions testing can go down to low Hz whilst a commercial applications can start at 150kHz. For 30khz you will generate harmonics every 30Khz (so 30, 60, 90, 120, 150khz) so at some point you will be generating EMI > 150khz, the question is how much are you generating?

0Sorry i made a mistake on the L equation it should have read

L =((10*VinMin)/Ipk)/(2*pi*F)

65/(2*pi*100000) = 103e-6 or 100uH
 
Last edited:

Thread Starter

mah

Joined Mar 15, 2010
393
Thank you , how to change the resonant frequency?is there a relation between resonant and lc values and noise db. is there a reference or application note for these equations to read in depth?. As i want to know more about them.
 
Last edited:

Marc Sugrue

Joined Jan 19, 2018
203
Thank you , is there a reference or application note for these equations to read in depth?. As i want to know more about them.
Not really, its more a impedance scaling of L & C by rule of thumb. If the impedance of the filter is too large you starve the DC/DC converter so the 10x factor is a common impedance scaling factor to avoid this whist giving decent attenuation. TI do some decent application notes around filtering of DC/DC converters so you may find more detail there.

https://www.ti.com/lit/an/snva538/s...28825&ref_url=https%3A%2F%2Fwww.google.com%2F

https://www.ti.com/lit/an/snva801/s...44942&ref_url=https%3A%2F%2Fwww.google.com%2F

https://www.ti.com/lit/an/snva489c/...xmAEAoAEBwAEB%26sclient%3Dmobile-gws-wiz-serp
 
Last edited:

Marc Sugrue

Joined Jan 19, 2018
203
how to measure noise db on my digital oscilloscope?
Only if you have a FFT function, you can nomally get a dBV measurement which you have to convert to dBuV to get an idea of where you are but really you need the LISN and Ground plane to be representative.
 

Marc Sugrue

Joined Jan 19, 2018
203
Thank you , how to change the resonant frequency?is there a relation between resonant and lc values and noise db. is there a reference or application note for these equations to read in depth?. As i want to know more about them.
Missed this, Change the Value of F to change the 40dB point. The resonant frequency will change also as the role off of a LC filter is 40dB/Decade. Try playing with the formulas and do a frequency plot to see the effect. Test it at full load and light load and you'll se the resonance lift
 

Ian0

Joined Aug 7, 2020
1,172
Two things to note about EMC filters:
1) For the purposes of the filter, the power is flowing in the opposite direction - from the switching transistor back towards the mains input.
2) At the higher frequencies you are trying to remove the parasitic capacitance of inductors and the parasitic inductance of capacitors are beginning to dominate. Check the datasheet of your inductor to find its self-resonant frequency. Above that frequency, it's effectively a capacitor.
 
Top