When Will Someone Make a Clean Switcher?

We all know that switching power supplies are wonderful things. They eliminate the bulky transformers used in linear supplies and provide efficient power conversion. They can step up or step down. Some are even smart enough to handle both functions if the input supply drops too low. And they've gotten easy to use: Pick the chip that has the right specs for input voltage, output voltage, and current; add a cheap inductor and maybe a discrete FET or two; lay it down on a board — done.

But have you ever paid attention to the noise that these things generate and radiate? Go get a portable radio with an AM detector and hold it near any plug-in wall charger or the last switcher you designed. Tune around the AM broadcast band, or if you can find a radio with some shortwave bands, listen there.

Older, cheap switchers running at 20kHz tend to produce raspy-sounding noise throughout the MF and HF spectrum, with peaks spaced 20kHz apart. And since the 20kHz switching clock is not all that stable, it drifts around with temperature, load, and pretty much any other stimulus. Higher-frequency switchers do much the same, with big signals at multiples of the switching frequency. And they also put out a lot of wideband noise between the harmonics of the switching frequency.

This noise radiates, using internal wiring of the power supply as the antenna, and it also gets conducted as a common-mode signal through the input and output leads which serve as nice antennas and radiate the noise energy everywhere. In some cases, when the CM noise gets to the wall socket, there's an imbalance somewhere nearby, and the common-mode signal becomes differential-mode, and is carried far away by the power-wiring/transmission line.

Good old dumb linear ac-dc supplies didn't do this.

I've noticed some power supplies, notably in PCs, are missing the filters on the line voltage input. I have heard that some manufacturers install the filters on the sample used to pass FCC and CE certification, then omit them in the production model. By the time someone complains and enforcement is initiated, the model is long out of production.

The noise that escapes the enclosure can often be knocked down by the application of appropriate ferrite material. That's the little doughnut-shaped thing on the cord of your laptop's power supply.

I also heard about a trick to pass the FCC specs that involved deliberately modulating the switching frequency such that the noise was smeared out over a wider band so that the noise power in any narrow measurement band was lower. I call that meeting the letter of the law, but not the spirit.

So why is this a problem? It is pollution on a grand scale. Most urban and suburban environments now exhibit noise in the MF and HF spectrum that interferes with various services, and as the problem gets worse, more services will be impacted. If we can start to take steps to reduce this pollution source, like a river with a polluting factory shut down, things will be clean again.

But whose responsibility is it? Is it the chip makers? Can the problem be fixed with either better designs or better app notes on how to eliminate radiated and conducted noise? How about the equipment manufacturers taking some responsibility for better use of the chips, and honesty in leaving the filters in the BOM for mass production? Is it more enforcement (and if so, who will fund it)?

What has been your experience in dealing with switcher noise? How did you fix the problem if the design wasn't your own? Did you get satisfaction from the responsible party?

Related posts:

18 comments on “When Will Someone Make a Clean Switcher?

  1. Vishal Prajapati
    September 25, 2013

    I never have to deal with such noise issue or probably I have been ignorant about it. But I think there is enough documentation and design notes available from standard switcher companies. Most of the time designer consideres all that things and designs a circuit. But at the end it is up to manufacturer to decide whether to include the EMI filter on the input stage or not. May be for the sake of more profit, they eliminate the EMI filter from BOM for mass production.


    So, untill the manufacturer doesn't understand it as his responsibility, which is less likely to happen, or some regulatory body which imposes compulsion on manufacturer it is not going to happen.

  2. Netcrawl
    September 25, 2013

    For economic reason manufacturer might not include an EMI filter, its will only add some cost to their current production. Its sounds like an extra works for their engineers and for their design. They want to get their products faster, adding extra stuff like EMI would not help them.  

  3. jkvasan
    September 25, 2013

    “The noise that escapes the enclosure can often be knocked down by the application of appropriate ferrite material. That's the little doughnut-shaped thing on the cord of your laptop's power supply.”


    We have used a similar technique by using those split ferrite modules on the cables that interconnect PCBs/modules. That paid rich dividends in passing the tests. We even used it on a 16 wire cable going to the LCD. 

    Another important question : All these LCD Modules seem vulnerable to EMI. Is there a manufacturer who designs them for EMI environment? Or it is left to us, designers?

  4. Vishal Prajapati
    September 25, 2013

    I am taking a case of LED Drivers. The design notes of LED Drivers are always showing Common Mode Chock at the input stage before Rectifier. But whatever drivers available here in India below 20W doesn't use any type of EMI Filter in the input stage. It is to reduce the cost of entire module. And surprisingly no one cares about EMI feeding to Mains Grid line.

  5. Vishal Prajapati
    September 25, 2013

    I think the most harsh environment LCDs I have seen is HMIs. They even comes with Capacitive touch display. So, I think for that harsh industrial environment they perform pretty well with no errors seen on the screen. So, I think you should look at it which modules they are using or how they are protecting them.

  6. jkvasan
    September 25, 2013


    I agree that these modules are available. My point is, they are protected by those EMI rejection panels and other components employed by the product designer. The LCD modules, as I understand, do not have inherent rejection capabilities.

  7. Bill_Jaffa
    September 25, 2013

    My HP Officejet 4110 printer is an excellent EMI source, I can pick it up in nearby AM radios even 5-10, even 15 feet away. The noise goes up to about 800 kHz–wipes out station below that frequency, much less above that. SO to listen to some sports games, I have to turn off the printer. 

  8. Scott Elder
    September 25, 2013

    @ Doug – <<...a trick to pass the FCC specs...modulating the switching frequency such that the noise was smeared out over a wider band ...meeting the letter of the law, but not the spirit.>>

    Aren't you being a bit harsh, Doug?  Consider that CDMA for cell phones depends upon the same principle and they intentionally radiate 0.25W.  That is how MORE users are added on to a slice of the RF spectrum rather than interfering.

  9. Doug Grant
    September 25, 2013

    No, I don't think I'm being harsh. Spread-spectrum switchers are designed to take advantage of the measurement method, not to minimize the total amount of noise power. There's no other benefit to the end-user. CDMA phones (all flavors) spread the signal so more users can share a chunk of spectrum, not to cover up their spurious emissions. And the theoretical processing gain of a CDMA phone allows it to transmit less total power for a given BER at the receiver so there's a user benefit to the DS spreading.

  10. Doug Grant
    September 25, 2013

    Yes, it saves cost, but it is illegal to submit a sample with an EMI iflter for certification then change the circuit, especially if it results in non-compliance. There is just not enough funding for strict enforcement. Sometimes a filter is not needed, and that is a good thing.

  11. Scott Elder
    September 25, 2013

    @Doug – Good point about cell phones.  So let me try again.  

    What about US CFR 47 Part 15 toy radios?  Certain sections of Part 15 allow emissions at almost ANY frequency provided the MEASURED emission is below a rated template.  In other words, nobody cares below a certain power level.  While I'm not an expert in the area of compliance testing, my understanding is that the bandwidths are pretty narrow for these measurements even though modern measurement systems could easily be set to wider bandwidths.

    So isn't this the same situation applied to emissions from a non intentional radiator like a PC power supply?  Below a template level, the regulatory bodies don't care.  I believe that impulse radar operates on the same principle.  Huge radiated emissions if mathematically integrated, but rendered small by spreading over a large band (cheap spread spectrum).

    To lighten things up a bit, I recall many years ago buying a Class A computing device and using it at home without being arrested.:-).  This is the same situation again although a bit more bizarre, especially given that the computer was shipped to a home address.

    P.S.  If it isn't clear by now, I actually like the idea of spread spectrum power supplies.:-) Also eases issues with PCB cross talk.  i think the FCC understands that these “tricks” are played which is why they also state that even if you meet the regulations, they can still shut you down.


  12. RedDerek
    September 25, 2013

    If needed, shielding the supply to keep all the noise within the box is one approach. Then is it left to taking care of what is coming out the input and output lines. Filtering on both is the next approach. In all cases, it does cost additional money in components and assembly to fix the issue. Tough when cost is a key factor in manufacturing.

    I have one mW supply design that needed to clean up its noise going back out to the supply source. A common-mode choke what a good solution to help quench that problem, but it was not 100%.

  13. bcarso
    September 25, 2013

    I chatted with Bruce Hofer some years ago when Audio Precision was presenting a new line of analyzers, which use switching supplies.  Hofer was candid in admitting his skepticism as to whether the switchers would be adequate for the extremely demanding application.  He said they evaluated a good many before finding a worthy product, that it was also more expensive due to a number of things inclusing the amount of shielding, and that in the end, the crucial component was a very high performance common-mode choke, constructed with great care.

    He went on to say that, in contrast to the older insturments using standard toroids and rectifiers/filters, for the first time he could not see any line-related spuriae in the DFTs.

    So, clean switchers exist, but they are rare birds indeed.  I share Doug's distaste for the spread spectrum approach as well, although my understanding is that makers of EMC test equipment are developing the capability of tracking the variable frequencies, so that you won't be able to get away with this forever.

  14. Chuck Sampson
    September 26, 2013

    “Pick the chip that has the right specs for input voltage, output voltage, and current; add a cheap inductor and maybe a discrete FET or two; lay it down on a board — done.”

    The sentence above relates the source of the entire problem. Most engineers think designing a switcher is that simple. “It's only five parts, how hard can it be?” 

    Also, most product design teams pay little attention to the power. It's usually the last thing they think of. With less than a week to go they finaly get around to asking “How are we going to power it?”

    A switching power supply, is a class D audio amplifier that operates as a time-varying and nonlinear system. Unlike an op amp or other linear amplifiers, where the bias is kept constant and the input signal varies, the switcher is the exact opposite. Its input signal, which is the reference voltage, stays constant, and its bias voltage varies from the minimum to the maximum input voltage it is designed for. Try doing that with an op amp. Sometimes the ratio between min and max is 4 to 1 and even higher.

    If you are using some excel spread sheet based web design tool to come up with a dc-dc for powering a PA or a pll, you are going to have noise problems. 

    In the first place, most load impedances are not resistors and that is usually what most novice ps designers, or managers or digital hardware guys who think they are just as good as power electronics engineers, assume.

    Most often the load is some sort of complex impedance, or negative impedance and this needs to be taken into consideration. Why? because the load can resonate with and amplify the switching noise. Or it can cause low magnitude sub harmonic oscillations, or it can cause large signal oscillations, or it can blow up. 

    As far as linear power supplies go, if incorrectly designed, they are noise nightmares. To maintain a constant output voltage, the output capacitor is charged repeatedly if the load varies repeatedly. So if you are powering an mcu that has to accept a repeating interrupt signal , or a pulsing GSM radio, the linear output cap sees a current spike repeatedly and you can get noise ripples just as bad as a switcher or worse.

    Sometimes clever RF engineers use linears to filter out the high frequency switching noise coming from dc-dcs. The problem they run into is the ripple rejection of most low drop out linears is less than 20 dB past 100 kHz. So if your switcher is operating at 600 kHz, linears are no help at all, they are just burning power. 

    I have designed many low noise dc-dc ps. None of them for the low-end commercial market. It's a mattery of economics. No one is going to have the power supply of their product be more than say 5% of the total product cost. So if you have a $50.00 product, you get $2.50 to make the dc-dc. If you want low noise, forget about it! 


  15. Brad_Albing
    September 27, 2013

    @VP – Well, no one cares now . I expect that in a few years, there may be enough electronic pollution (conducted over the power mains and radiated) that the government may institute new regulations.

  16. Brad_Albing
    September 27, 2013

    @Bill – so, no CKLW or WABC for you while you're printing, I guess. But you can still listen to WBZ and whatever WCFL is now.

  17. jkvasan
    September 28, 2013


    Shielding and filtering on both ends can suppress most of the noise. It also pays to route the cables in a proper fashion to avoid any criss-cross between the noisy and the supposed-to-be-quite lines.

    It also can pay in the longer run to route high speed lines and important signals through separate connectors and cables and provide sufficient filtering. 

  18. amrutah
    February 1, 2014

    Switchers are supposed to switch when in regulation mode, but the snubber circuits are important to keep the noise levels low.  Being a part of it I have seen snubber circuits being considered while designing the switchers, but it is upto the customer to specify it and use it in their application.  Check this for snubber circuit design.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.