Saturday, December 8, 2012

Reducing switching supply racket (RF Interference)

Note:

There is a follow-up to this posting in the August 18, 2014 blog entry - link - where there are details given to contain the switching supply noise even more!

Switching power supplies are ubiquitous these days - and for several good reasons:
Figure 1:
Typical laptop-type switching
power supply - the very unit
that was modified!
Click on the image for
a larger version
  • They are more efficient than plain old iron transformer power supply with a linear regulator.
  • They can be much smaller and lighter than their transformer/linear counterparts.
  • They are cheap by comparison since they use less material overall - particularly iron and copper - in the transformer.
 They do have several real drawbacks:
  • Most tend to be less reliable than their old heavy iron counterparts.  I've observed that the typical switching-type "wall wart" (plug-in power supply) seems to last just 2-4 years whereas the old-fashioned iron types would usually outlast the device to which they were connected.
  • They can generate some terrible radio interference!
On the first point, I often wonder if the amount of power they save due to their efficiency is outweighed by the fact that they often fail after just a few years, often causing it and the device it powered to end up in the trash because of the failure of less than $1 worth of components - but that's another topic of discussion!

Shown in figure 1, above, is a typical power supply of the sort used on a laptop computer.  As far as switching supplies go, this is one of the better-built units, now used to power a small form-factor PC that I have attached to my TV to watch digital/online media - and because of this, it's plugged in pretty much all of the time.

Note:  I have since plugged this supply into a "smart" power strip.  This strip has a sensing circuit that detects when the TV is turned on and only then are the "switched" outlets powered up, saving energy by powering down those devices that are never used when the power is off.
Figure 2:
Typical "Common Mode" AC line filter.   The capacitors force RF
to be "common mode" so that the bifilar inductor (in the middle) can
best do its work!

As it turns out I could hear some (admittedly weak) harmonics of a switching supply on my HF receivers, but I generally ignored them until I happened to tune across the AM band on my newly-repaired Marantz receiver (see the previous blog entry) and heard some very strong hum-laden carriers every 30-50 kHz across the broadcast band that blotted out most of the local stations.  Unplugging nearby mains-powered devices soon revealed that the source was (mostly) the power supply pictured above, located only a few feet away from the receiver.

Taking this as a challenge - and an excuse to take some pictures and do a write-up for this blog - I set about to make this power supply much less obnoxious, RF-wise, so I put the power supply on the bench and popped it apart.

Figure 2:
Inside the power supply - the AC input on the left side.
The original bifilar RFI filtering choke (upper-left)
has the green/yellow wire wound onto it.
Click on the image for a larger version.
The usual warnings about high voltages:
  • This power supply - and others like it - operate from potentially lethal line voltages.
  • DO NOT attempt to open or modify a power supply unless you are thoroughly familiar with the proper techniques and safety precautions when working with these voltages.
  • Figure 3:
    A close up of the RFI limiting components.  Below the bifilar
    choke (the device with the green/yellow winding) is the
    capacitor that forces RF energy to be "common mode."
    Click on the image for a larger version.
  • If done improperly, modifications to the power supply may make it unsafe to use and become a fire and/or shock hazard, so do not do this sort of work unless you know exactly what you are doing!
The main RFI suppressing components of the power supply may be seen in figure 4 with the AC input on the left - namely the black device with the green and yellow wire wound on it (a bifilar-wound inductor) and a capacitor - the black rectangular box (marked with "104") below it.

A switching power supply is really a powerful oscillator with the voltage being transferred to the load with a small transformer - the size reduction compared to the old-style "wall warts" being permitted because the power supply operates at a frequency much higher than that of the line voltage's 60 (or 50) Hz, and at several 10's of KHz, usually in the 30-60 kHz range for most of these types of power supplies.  This higher frequency of operation is also the reason why switching power supplies often cause interference issues to radio receivers:  It is the harmonics from this high-power oscillator that are more likely to be conducted to the outside world via the AC power connector and/or the DC output.

In Figure 2 is the diagram of a typical "common mode" AC line filter.  Looking similar to a transformer is the bifilar choke that is doing most of the work of filtering the high frequency components of the switching supply plus it also can do a pretty good job of actually isolating the power line at these higher frequencies so that not only are those spectral components generated inside the power line contained therein, but also that the supply itself won't supply a path to conduct RF energy from whatever it is that is being powered by the supply (a computer, set-top box, modem, etc.) into the power line itself.

The way that this works is that any RF energy on one side of the choke will get coupled to the other side of the choke equally.  Since this bifilar choke is a choke, its inductance will form a series impedance to block higher frequencies from passing through, the effectiveness being related to the inductance of the winding itself.

Key to this working properly is that any RF energy on one side of the bifilar choke must be exactly equal to the other side or else the imbalance can actually cause more interference as unequal RF energy from one side would be induced on the other side.  To force the RF energy to be equal is the job of the two capacitors shown - one on the input, and the other on the output.

This particular power supply had only a capacitor on the load side of the power supply - where the noise was being generated.  While this will do most of the work, it does help to have a capacitor on both sides, but this is often not done as a cost-saving measure.

Figure 4:
Inductance of the original coil.
With only 268 uH per side:
That's not much filtering at
AM or the lower HF bands!
Click on the image
for a larger version.
Since the yellow-green wired inductor didn't seem to be adequate, I removed it from the power supply and measured it (see figure 5).  Noting that the inductance is a mere 268 uH, I thought that I could do better with some other line-filtering inductors that I happened to have in my junk box - this one (figure 6) measuring about 4.594 millihenries (4594 uH) which  is about 17 times as much inductance which also means that it will, ideally, offer 17 times as much impedance to RF energy that might escape from the power supply via the AC power line.

Since the original choke was 268 uH, let's find out how much equivalent series resistance that amount of inductance offers at, say, 1 MHz - in the middle of the AM broadcast band.  The formula for inductive reactance is:

Z = 2 * Pi * F * L

Where:
   F = Frequency in Hz
   L = Inductance in Henries
   Z = Inductive reactance in Ohms

So, plugging 268 uH at 1 MHz into the above we get 1683 ohms - not too bad, actually.  By replacing this choke with the 4.594 millihenry version our impedance scales up proportionally to 28.850k ohms at 1 MHz!  In addition to the bifilar action of the choke, this significant amount of inductive reactance will go a long way toward both keeping the RF energy from the switching supply off the power line, but it will also keep the power supply itself from acting as a pathway to couple potential interference from the devices connected to it to/from the power line.

Comment:

It is common to attempt the use of ferrite beads to suppress RF Interference of this sort, but it's very unlikely that it will help much - particularly at lower frequencies (e.g. lower HF bands such as 160 and 80 meters, not to mention the AM broadcast band) because these devices simply cannot add enough inductance to add a significant amount of impedance:  At these frequencies (say, below 10 MHz) it takes multiple turns on a chunk of ferrite to add enough reactance to make even a small dent in the amount of conducted interference!

Cramming this much larger component into the same space as the original bifilar choke was a bit of a challenge, but laying it on its side and using "flying leads" to connect the inductor to the circuit board made it possible to fit it inside the case.

For good measure I also added another capacitor (a 0.047 uF device) to the "other" side of the inductor (the side opposite the black capacitor mentioned above) to better-equalize any RF currents that might occur across it (the small green capacitor in figure 7).  Just to be safe, I also put some polyimide (a.k.a. Kaptontm) tape on the aluminum heat sink (visible in figure 9) to make sure that the windings of the coil could not touch (and electrify) the heat sink itself or other nearby components.

Figure 5:
Inductance of the new coil.
With 17 times the original
coil's inductance, it's likely
to provide better filtering
at lower frequencies!
Click on the image
for a larger version.
Having installed this new bifilar inductor I still had the original bifilar device (the one with the green and yellow wire) on hand so I decided to put it on the DC output to further contain any RF energy emitted by the power supply - and why not, since it was "free"!  Using its color coded windings, I connected it as shown in figures 7 and 8 with heat shrink tubing to insulate the soldered connections on the power supply's DC output cord.

Putting everything back in the case I carefully re-checked the clearances and insulation to make certain that not only would everything fit, but also that nothing could short out - especially when everything was smashed together when the cover was put back on.  While I could have glued the two halves of the cover back together, I decided to use some of the same polyimide tape mentioned above as it has a very strong adhesive - and I would be able to easily take the power supply apart should there have be a problem.  After reassembly, I then re-checked the DC polarity of the output connector to make sure that I didn't accidentally reverse it when connecting the output choke.

The result?

While I can still "hear" the harmonics radiated from this power supply on the AM radio that's just a few feet away, they were now weaker that most AM stations instead of being "extremely loud" and clobbering much of the AM dial - this fact indicating a reasonable amount of success.  While the intent was not to attempt to completely "clean up" the power supply's spurious radiation, the radical difference indicated that all spurious radiation from this particular power supply was likely to be very much reduced.  Elsewhere in the spectrum, I can no longer hear even a hint of this power supply on any HF band!

Figure 6:
"New" inductor with added 0.047 uF capacitor.  It is
connected with "flying leads" to provide connections
into the circuit.  Not seen in this picture is additional
insulation added between the body of the
new choke and the heat sink.
Click on the image for a larger version.
Since these types of power supplies are seemingly everywhere, it should come as no surprise that there are several of these in my ham shack and I've applied the above techniques to those other power supplies that were found to cause interference on the HF frequencies.  Depending on the power supply and the amount of extra room inside the case, one may (or may not) be able to add as many additional inductors and capacitors as was needed to quash the RFI emitted by the power supply, so in several instances I've added filtering outside the case,  typically inserting capacitors and a bifilar inductor on the DC lead (but close to the power supply) where it would be safe to do so.


Figure 7:
The original bifilar inductor, now connected on the DC
output to provide additional filtering.  Heat-shrink
tubing was used to insulate the output DC connections.
Click on the image for a larger version.
Ideally, one would put such filtering (e.g. inductors) on both the AC and DC leads, but it's worth remembering that these power supplies pollute the RF environment largely by conducting the harmonics of the switching frequencies through the input and output leads:  If one blocks the RF energy from being conducted on just one lead or the other (e.g. the AC input or the DC output) the circulating currents carrying this energy through the power supply (e.g. in on the AC side and out on the DC side) are significantly reduced and adding such blocking can considerably reduce emitted RFI.  Also worth mentioning is the fact that many switching-type DC supplies - particularly "wall-wart" types - have minimal or no common-mode filtering (e.g. using a bifilar choke or two separate series chokes) on their DC output, probably because it's a bit more expensive to do it this way.

I've noticed upon opening the case that some switching power supplies - perhaps of dubious origin and quality - are completely missing the RFI filtering components.  In these same power supplies it is often apparent that there is a position on the circuit board for these components, but they are either empty (in the case of missing capacitors) or jumpered over (in the case of missing inductors) - clearly a cost-saving measure and probably illegal in some countries.  For these power supplies the addition of any RFI suppressing components will likely have a significant effect on reducing interference that they may generate!  I've also observed that many of these same supplies of unknown pedigree often use the cheapest-possible components and it may well be that they will not prove to have a long lifespan!
Figure 8:
The modified power supply with the reconfigured filtering
and placed in the bottom half of the original case.
Click on the image for a larger version

Where does one get these bifilar inductors?  Most computer-type power supplies have these on their inputs and they may be found in most reasonably-quality switching supplies.  Remember how I mentioned that these switching supplies often die after just a couple of years?  These dead supplies may be a ready source of components to better RFI-proof the supply that may be causing interference to you!

Figure 9 shows, in the highlighted portions, the bifilar inductors - and some associated capacitors - found in some typical junked power supplies.  On the left is a typical PC power supply  where one can see what looks like a small transformer next to the AC power line fuse.  On the right is a power supply from a junked VCR with the bifilar inductor also very near the AC power line fuse.

Note:  If you raid junked power supplies for components, make sure that they are unplugged (obviously!) and that the large, high-voltage capacitors filter have been safely dischargedIf you are unsure about how to do this, please seek advice and help from someone who does know before engaging in a project dealing with potentially deadly AC power voltages!

Figure 9: 
Examples of RF filtering components found in junked
switching power supplies.  On the left is a PC (computer)
power supply while on the right is a power supply from
a VCR.
Click on the image for a larger version.
These two bifilar chokes - while somewhat different in style - are split in two with one side of the AC power line on one side, and the other side of the AC power line on the other.  Being wound on a common core, their winding are very tightly AC-coupled (at radio frequencies, at least!) which is how they function to prevent conduction of this energy onto the AC power line.  Before removing them from the board, verify with an ohmmeter from the original AC power connection that the inductors you spot are, in fact, in series with the power line - with one half on one side, and the the other half on the other side!

You'll also notice that these two power supplies have something in common:  There are capacitors very near the bifilar inductor.  In the case of the PC power supply (on the left) there is a large, yellow rectangular capacitor on the AC input of the power supply and on the opposite side, there are two blue disk-ceramic capacitors (one of them covered with heat-shrink tubing).  In the case of the VCR power supply (on the right) you'll see even more filtering:  There are several blue capacitors sprinkled throughout, but also the orange-red capacitors next to the bifilar inductor itself.

It is quite typical for there to be blue capacitors on the inputs of power supplies for filtering - these being "safety components" that are specifically designed for both filtering, and for reliability so that their failure won't inadvertently cause the case of the device to be connected to the dangerous AC line voltage!  The other capacitors - the big yellow one on the PC supply and the two orange-red ones on the VCR supply - actually do much of the filtering.  The one thing that all of these capacitors (blue, yellow and orange-red) have in common is that they are specifically rated to withstand the AC line voltage!  Careful inspection of these components will reveal not only their capacitance value, but also their voltage rating.

If one is reasonably careful, discarded switching power supplies can offer a ready source of components - both inductors and capacitors - to help reduce their conduction of switching energy and the interference that it may cause.


In severe cases I have found it necessary to enclose the entire switching supply in a larger box such as that of a discarded PC supply, using RFI filtering on the AC line as well as the DC output:  This was finally covered in the August 18, 2014 blog entry - link.

Links to other articles about power supply noise reduction found at ka7oei.blogspot.com:


[End]

This page stolen from ka7oei.blogspot.com

10 comments:

  1. Excellent article!

    Thank you

    73

    VE2ZFP

    ReplyDelete
  2. By the way
    I'm building a 5 amp power supply for my QRP
    rig, I want to put it in a metal box for further improvement. Should I ground the box to the primary, the secondary or leave it floating?

    Thanks

    ReplyDelete
    Replies
    1. If you have a case "ground" available, the filter design changes slightly - take a look at this article:

      http://www.arrl.org/files/file/Technology/tis/info/pdf/8407016.pdf

      At the end there is a "Brute Force Line Filter" that looks much like the filter in the blog post, above with the addition of the case ground between a pair of capacitors. These capacitors do double-duty:

      - The two in series act as one capacitor to force the RFI to common mode so that it can be quashed by the bifilar line choke.

      - Any RF on the lines is shunted to "ground" - in this case, the case!

      The hazard with this approach is in the two capacitors with the ground in the middle: These *MUST* be rated for the AC line voltage, so if you don't see printed on them a voltage rating that indicates that it would be OK to connect them to AC mains voltage (e.g. at least 1000 volts DC, or an AC voltage rating safely above the mains voltage) then *DON'T* use it. If one of these capacitors were to fail, it's possible that the full line voltage could appear on the case ground - a very bad thing, indeed!

      (Note: Most RFI filtering "safety" capacitors are a light blue color and are typically found in the AC input filtering sections of scrapped PC power supplies and the like. These will usually have printed on them their AC voltage ratings.)

      New and surplus, ready-built power line RFI filter modules are also available and many of these have built-in IEC (computer-type) power cord receptacles on the outside and solder lug terminals on the inside. These work very well and provide everything you need for the AC input filter in one handy package.

      * * *

      Another escape route for the RFI is on the DC side and it is best if both the positive and negative lead have series inductors to block RFI and a minimum recommended value for these chokes at RF frequencies is 10uH - although the higher, the better generally speaking. Finding a choke rated for the maximum current and with a low enough DC resistance to offer an acceptable voltage drop under full load is a bit trickier, but these chokes can often be found in scrapped PC power supplies, too.

      In addition to a capacitor across the + and - of both the power supply and output side of these series chokes (say, a 0.1uF) on the output side a 0.01uF capacitor should be connected between each DC output lead and ground.

      Inside the box it is recommended that the bypass capacitors be connected to the case ground with as short as leads as practical rather than have a long "case ground" lead inside the box to connect it. Also, it is best if the ground of these capacitors on the input (AC) side and the output (DC) side be connected near-ish where their respective wires exit the box.

      Best of luck!

      Delete
  3. Finally, a nice display of the RFI problem and repair. TNX.!

    Can you do more info/blog on keeping RFI from getting into the AC lines- and making an "outboard" brute force filter for the SMPS input? I find that is where most re-radiated hash comes from. Also, how to best handle wall-warts with no ground pin on the AC plug and the use of plastic boxes- and radiation from them- would be helpful.
    Thanks

    PS- My and other friends' story :
    SMPS hash as been the last straw and taken me off the air (I cant work them if you can't hear them) and made AM BC listening impossible from our rural home. Same at my Girlfriends place.

    When national cable TV providers supply their equipment with noisy "wall warts" to power their digital boxes, (and their modems /wifi too) there is no way I can fix a whole neighbourhood full of them. The next door neighbour is LOUD and cumulative effect from the whole neighbourhood kills the full RF spectrum.

    Plus there are so many noisy SMPS, just in my home. I have to kill the breaker box "mains" and run off battery to have a usable receiver noise floor. To fix all the SMPS's seems a daunting task. But your info has inspired me to tackle the worst offenders. (Replacing them all with old style or one large regulated supply is an option- but again, doesn't solve the neighbourhood RF smog problem

    Sad thing is; the AM BC'ers don't seem to be doing much with complaints to regulating agencies. A very old established Toronto AM station's solution was to apply for an FM assignment, sighting higher AM new "device-noise" levels as the reason. I feel especially sorry for Shortwave listeners who have no protection in the regulations, like a licensed Ham..

    Your finding of missing parts when they were used for certification or importing approval is maddening and, again, lack of fines or charges just make the whole priceless RF spectrum resource destined to be lost and wasted.

    Thanks for your info!
    G

    ReplyDelete
    Replies
    1. Hello G,

      The noise radiated by switching power supplies does not really get radiated by the power supply itself, but from three main places:

      - Differentially, on the AC input
      - Differentially, on the DC output
      - Differentially, between the DC output and the AC input

      The first two can be significantly reduced by the addition of decent filtering (capacitors, inductors) as described in the article above, but the last one is more difficult to address.

      To be sure, it can be significantly reduced with the use of a good-quality common-mode choke/capacitor filter as that will increase the reactance at high frequencies and reduce the amount of energy that can be conducted onto the AC and/or DC lines and thus onto the "air" and cause QRM: The higher the inductance and the better the filtering, less energy is likely to be conducted and cause a problem - but practically speaking, if the power supply is not contained in a metal case to provide a common reference for both the AC input and DC output to which circulating RF currents can be shunted, one can go only so far in reducing the problem.

      To fully address that last point - eliminate the differential RF currents between the DC output and AC input - the errant power supply must be put in a metal enclosure, and for smaller supplies, discarded PC power supply cases work nicely for this as they are well-ventilated and they already have an AC attachment for a power cord. What is needed to go inside these are:

      - Appropriate AC line filtering to keep the RF energy from getting back out onto the mains.
      - Filtering on the DC output, referenced to the metal case at RF, to contain the RF of the power supply within the case.
      - All of this would have to be done with safety in mind, considering wiring/fire/electrical safety regulations!

      It is often possible to put several small switching "wall warts" within one of these cases and completely eliminate (as in make undetectable) the "crud" that they produce - although it is sometimes the case that the devices that they power (e.g. a DSL modem) produce a bit of crud on their own!

      As noted on the article above, it is hoped that this will be the topic for a future blog entry - if/when I get time, of course!

      73

      Delete
  4. If you wanted to make your own common mode choke, what would you recommend for a core mix? I intend to knock out the noise of a DC-DC switcher so it can be used to stabilize the voltage of a battery pack. 73, AF2RF

    ReplyDelete
    Replies
    1. You didn't specify a frequency range, but mixes 73/77 are very high permeability and good for the very low end of HF.

      Mixes 31 and 43 are more appropriate for higher bands - but have lower permeability and may not work for lower frequency as well (e.g. AM broadcast, 160/80 meters.)

      I've often used cores salvaged from PC supplies and these often tend to be toward the "77/73" end of the spectrum: As long as good construction techniques and proper components are applied, even those will work nicely into VHF.

      Let me know how it works out!

      Delete
  5. I inclined to disagree with the following comment of yours,
    "It is common to attempt the use of ferrite beads to suppress RF Interference of this sort, but it's very unlikely that it will help much - particularly at lower frequencies (e.g. lower HF bands such as 160 and 80 meters, not to mention the AM broadcast band) because these devices simply cannot add enough inductance to add a significant amount of impedance: At these frequencies (say, below 10 MHz) it takes multiple turns on a chunk of ferrite to add enough reactance to make even a small dent in the amount of conducted interference!"

    The philosophy of using ferrite beads is that they become very lossy at HF, thereby converting RF energy into heat and thus removing RFI, on unlike an inductance which offers high impedance to RF.
    I'd love to hear your opinion.

    ReplyDelete
    Replies
    1. From a technical stand point, a review of the technical data provided by the manufacturer of such devices will reveal their specified series resistance at a given frequency: At HF frequencies, this is often in the 10s of ohms (maybe low 100s) per device. Considering that several k-ohms ESR would be required to effectively choke the noise of an offending device by the (often) needed *10s* of dB, these devices simply cannot offer the needed "loss". I have wound/measured inductances of known devices that I have on hand and found that they track, by inductive reactance at a given frequency, the published charts pretty well (within 20%).

      For the ferrite devices typically used by the amateur radio operator (Mix 77, 31, 43) the series impedance values closely agree with the inductive reactance obtained by the specified number of turns for a given frequency - so for these devices, at least, the attenuation of RF via power dissipation (e.g. heat) is not the primary means of choking: This fact is further demonstrated by the fact that when these devices are used as intended (for choking impedance) that they do not get hot - even when choking 10s of watts of common-mode RF, or even when used at the kilowatt power level as an impedance transformation device.

      To be sure, there are devices specifically intended for common-mode suppression on power supplies - typically of very high permeability - that are not well-characterized into the MHz range that do seem to provide useful amounts of attenuation - but these, too, offer minimal effect when used singly/in series (e.g. simply slipped over a conductor).

      As one might expect, the exponential effects of multiple turns and that of other means of bypassing (e.g. capacitors) can offer far better attenuation of conducted energy, but it is often the case that it is impractical to do so.

      Delete





PLEASE NOTE:


While I DO appreciate comments, those comments that are just vehicles to other web sites without substantial content in their own right WILL NOT be posted!

If you include a link in your comment that simply points to advertisements or a commercial web page, it WILL be rejected as SPAM!