DC Ripple Attenuation Products or Techniques

Thread Starter


Joined Apr 9, 2018
I am looking at buying a AC-DC power supply but I need it to have less than or equal to 3mV p-p ripple on the output. The supply I am looking at is a 5V/20A with 100mV p-p ripple. So if I am going to go with that supply, I am going to need to drop that ripple about -15dB. Does anyone know of a black box solution or product available on the market that can accomplish this? I have looked at the Ripple Attenuator Modules from Vicor but they don't seem to make them anymore. Is there any other company doing this? Alternatively does any on know if there any other power supply companies that offer low noise AC-DC power with my requirements?


Joined Mar 30, 2015
What is the input voltage (AC or DC?), output voltage, and current requirement?

Out of curiosity, what type of device/circuit/thingy requires the supply voltage to be so accurate?


Joined Feb 8, 2018
I'm assuming this is a switcher.

Typically you would use a PI filter with ceramic capacitors and probably toroidal inductors on powdered iron cores, though there are several other inductor types that would likely work well enough. This might be supplemented with lossy ferrite suppression "beads" (some pretty big chunks of ferrite get called beads) which can take care of some of the very high frequency noise. The ceramic caps might be supplemented with electrolytic capacitors, especially if there is residual lower frequency ripple, but sometimes just to help avoid high Q filters that have a sharp gain peak that can sometimes be troublesome.

The inductors need to be selected carefully. Ideally they would have a single layer or "progressive" winding (two steps forward one step back sort of thing) to minimize capacitance between the input and output ends. Because they are outside of the control loop, the DC resistance must be low to avoid excessive impairment of regulation. The "saturation" current rating must be high enough to avoid excessive loss of inductance due to the DC bias. Depending on the core material you can safely operate well beyond the rated saturation current if you can tolerate some more reduction in inductance. It is common to find saturation specified for only 10% reduction in inductance, but some core types are fine at three times that.

What you require in terms of values depends very much on the frequencies that need reduction. There will probably be a little residual ripple at the switching frequency and some trash at much higher frequencies. It is fairly rare to find a commercial switcher at that power level operating below 50 kHz these days, though some drop the frequency back at light load for better efficiency. Those that don't drop the frequency can actually produce higher ripple at very light load.

Sometimes you can do better by using two PI stages each with lower inductance and possibly less capacitance. This can be more compact overall and ease issues with the inductors.

You may or may not want inductors in both the +5 V and return lines to the supply.

Sometimes just addition of a moderately large amount of good high-frequency capacitance can help substantially. X5R or X7R surface mount ceramic caps in quite high values are fairly inexpensive, though a bit of a nuisance as add-ons. There are also some remarkably good very low ESR aluminum-polymer and tantalum-polymer capacitors available, with lots of choices at 5 volts.

I recommend have a look at the websites of some of the notable manufacturers of switchers such as TDK-Lambda. They may publish ap notes on supplementary ripple and noise filters.


Joined Aug 21, 2008
The source of the noise is an important consideration. What is the source and nature of the noise you want to get rid of?

One good approach when the situation warrants the expense is to use the power supply as a pre-regulator and follow it with an analog regulator.

In a broadcast television camera video board where every db of signal-to-noise ratio counted I used "ripple clippers" which were emitter followers in which the base was connected to the i+12 volt bus through an RC filter and the collect connected directly to the +12 volts. The emitter had some bypass capacitors on it as well. This isolated the video circuitry from noise on the +12 volts. For the current you are talking about you would need more than an emitter follower.