SMPS noise reduction

Thread Starter

witssq

Joined Mar 29, 2009
48
The commercial SMPS 5V/1A is purchased. It have 140mV ripple, and then
I reduced ripple 40mV using pi LC filter as attached schematic.

But as you see on the scope view 2.jpg attached, about 330KHz transient(impulse) voltage exists. I would like to eliminate transient 40mV to make to 10mV ripple.

Could you anybody help me?




PS.
1.jpg - No filter component
2.jpg - After filter applied
 

Attachments

Thread Starter

witssq

Joined Mar 29, 2009
48
As I may know, the ferrit bead is used for high frequency above 1Mhz.
If it is true, It may vbe useless. How do you think about?
 

bountyhunter

Joined Sep 7, 2009
2,512
This is the world of EMI, I spent about 25 years fighting it. There are no magic fixes. It radiates so easily that the wires inside a power supply always pick some of it up. You have to do the mechanical design to contain the E MI radiating components and use filtering into th "clean areas". One solution:

Build a steel box with banana plugs on one end to plug into the front of your P/S and jacks on the front. Inside the box, put some good bypass caps and series L filters. You can reduce the ripple based on how good you build the filter.
 

Thread Starter

witssq

Joined Mar 29, 2009
48
Thanks for your response.

Please advise me whether or not Twin-t notch filer could eliminate the above transiend transient.
If possible, target frequecy become 330KHz because that the transiend interval is 3.03ms. is this right?

The calculator/schematic is linked as http://sim.okawa-denshi.jp/en/TwinTCRkeisan.htm

Regards
SunSung Hwang
 
Last edited:

t06afre

Joined May 11, 2009
5,934
Thanks for your response.

Please advise me whether or not Twin-t notch filer could eliminate the above transiend transient.
If possible, target frequecy become 330KHz because that the transiend interval is 3.03ms. is this right?

The calculator/schematic is linked as http://sim.okawa-denshi.jp/en/TwinTCRkeisan.htm

Regards
SunSung Hwang
No the target frequency is the frequency in the transient itself . From you picture. I would say it has period around .2 usec roughly estimated or about 5 MHz. It is important to use both caps and inductors that are suited for such high frequency
 

bountyhunter

Joined Sep 7, 2009
2,512
The frequency at which the noise repeats is not what you have the problem with: look at the noise waveform. The sharply rising edges have components that go into the hundreds of Mega Hertz. That's whay they are so hard to kill, the radiate into just about any short wire they can find.

Use a low pass filter. As stated, you must use caps that are good at mega hertz ranges like ceramic caps and possibly polyester film caps. Aluminum electrolytics and tantalums will have very little effect since the noise frequencies are above their resonant frequency.
 

kubeek

Joined Sep 20, 2005
5,793
If possible, target frequecy become 330KHz because that the transiend interval is 3.03ms. is this right?
Just from looking on the picture I would say that the fundamental of the noise is about 10-20MHz. Just try the ferrite beads and see if it helps. Also try measuring the noise with faster timebase and higher bandwidth probes.
 

Thread Starter

witssq

Joined Mar 29, 2009
48
It seems to be about 15-20Mhz. I applied ferrit bead, Noise is attenuated
from 120mV pk to pk, 80mV pkpk. but I need 5-10mV. I seems more to
do.

I will test with another smps, then return here.


Thank you all.
 

crutschow

Joined Mar 14, 2008
34,201
The capacitor and ground connections must be as short as possible.

You might try a feed-through type capacitor or filter. They minimize the effect of wire inductance which is a problem with high frequency noise.
 
Top