Wednesday, September 4, 2013

Quieting high-current switching power supplies used in the ham shack.

Over the years I have acquired several switching supplies that I use in the shack and in portable operation such as field day.  These power supplies (two Samlex model 1223's and a Radio Shack #22-510) were designed to be "RF Quiet" compared to more typical switching supplies that might be used for computer or industrial applications.


It's all in the filtering...

What separates a typical, industrial power supply (and most computer-type supplies) from one that is intended to be "quiet" (RF-wise) is largely filtering, as what is contained within the box comprising the switching supply is essentially a high-power transmitter!  It's pretty easy for low-level harmonics of, say, a 300 watt switching power supply (oscillator!) to leak out, and since even a few billionths of a watt at the input of a receiver make a signal that is annoyingly strong, one can appreciate the need for proper containment!

Fortunately, most switching power supplies used in this application operate in the 30-60 kHz range which means that, by virtue of the large frequency difference, the harmonics - those buzzy, raspy things that often appear every 30-60 kHz across the HF bands - are already weakened considerably, but one needs to do more to submerge them below the noise floor!

It should be no surprise that it's largely the AC input and DC output leads that conduct this energy out of the box so some fairly good filters are required.

AC Line filtering:

Take as an example the circuit depicted in Figure 1, below:

Figure 1:  2-stage "brute-force" line filter using bifilar inductors.  The AC power comes in on the left and is delivered to the "guts" of the switching supply ("load") on the right.

This is a typical filter found on the AC power line of better-quality power supplies that are designed to be "RF quiet" and what does the most work are the two bifilar inductors.  How this works is that at capacitor Cd, where there are strong RF components of the switching energy, the two sides of the AC power line are "shorted" at RF frequencies (e.g. made to be "common-mode") so that when equal amounts of RF pass through bifilar inductor, they get canceled out and "choked" by the inductance.  The first of these inductors (the one on the right) isn't able to do all of the work, so another stage of filtering consisting of capacitor Cc and another bifilar inductor is applied.

For an article describing what is meant by a "common-mode" signal, read here:  http://en.wikipedia.org/wiki/Common-mode_signal

Finally, at the power line we have capacitors Ca and Cb and these not only help reinforce the "common mode-ness" and the effects of the bifilar inductor, but shunt the remaining small amount of RF from the switcher to the metal box containing the switching power supply so that it does not escape to the power line.  Typically the values of Cc and Cd are in the range of 0.01uF to 1 uF and the higher the capacitance, the better - but at the cost of the component itself (larger capacitors are more expensive) and the fact that as you increase the capacitance, it will draw more and more current from the AC power line on its own due to capacitive reactance.  (This leakage current will not generate heat since it is very reactive - but that's another discussion altogether.)

The values of Ca and Cb can vary, but it's common to find anything from 0.001uF to 0.47uF, but some safety laws limit the values owing to the fact that at this midpoint (ground) there will be approximately 1/2 the AC mains voltage (with respect to either side of the AC mains) should the ground be disconnected:  The value of these capacitors and their reactance will dictate how much of a shock hazard (current flow) that this might present should accidental contact occur.

These capacitors must also be appropriately safety-rated since failure could put the full mains voltage on the safety ground and pose a lethal shock or fire hazard.  Typically, these capacitors are blue - sometimes yellow - and have imprinted on them their specific AC rated voltages and have an "X2" marking on them as well as having symbols indicating the various safety and regulatory bodies by which their use is approved.

The inductors are the most expensive components in this filter since they use fairly heavy copper wire for to handle the multi-hundred watt load as well as fairly pricey ferrite material.  They are fairly large and heavy so it is not too surprising to find an off-brand or counterfeit power supply where all of these filtering components (inductors and capacitors alike) are omitted to cut costs:  Such power supplies radiate lots of noise and do not meet regulatory (or even safety) requirements in most countries!


DC output filtering:

The other place where RFI might escape is the DC output.  Take the example of Figure 2, below:
Figure 2:
Simplified diagram of the DC output of a switching power supply.
Capacitors Ca and Cb are typically large electrolytic units that remove high-frequency ripple of the switching supply from the power supply's output voltage.  See the notes below regarding capacitor Cc.
Here, we have the switcher's output circuit:  A high-power oscillator running in the 30-60 kHz range feeding a transformer that converts the voltage from the 150-300 volts of rectified and filtered AC line input, down to the 13.8 volts while also isolating the power mains from the DC output.  This is typically a center-tapped transformer with a full-wave rectifier followed by bank (usually 2 or more) of good-quality (hopefully!) filter capacitors represented by "Ca."  Even with the best capacitors there is still residual switching energy, so inductor "L" is typically used to filter it further followed Cb which consists of another capacitor or two in parallel to knock it down even more.

If the circuit board has been laid out properly properly and good-quality components have been used, the "V+ Out" line will be pretty clean - but notice something else:  The "ground" to which the transformer center-tap, Ca and Cb are connected is different from that of the chassis (case) ground in that that they aren't even connected directly to each other!

There are several reasons for this.  First of all, it is often desired that the case ground - which is usually connected to the AC mains safety ground, as well, be isolated (DC-wise) from the DC output of the power supply, this being done to prevent "ground loops" - that is, power finding its way along more than one lead and back to the same place.  In extreme cases this can cause hum or, in the case of faulty mains wiring, put a shock hazard on the metal case of the gear being powered.  The use of a capacitor such as Cc "connects" the two at RF, but not at DC or at mains frequencies.

In the cases depicted below, plastic capacitors rated for at least 250 volts are used which is adequate for 120 volt mains.  This seemed to provide adequate bypassing - even at fairly low radio frequencies - and the value used still presents reasonably high reactance at AC mains frequencies (>800 ohms at 60 Hz, >960 ohms at 50 Hz for 3.3uF) to afford a reasonable degree of safety, minimize circulating currents (hum) at those frequencies while eliminating the possibility of any DC ground loops.

Common problems with "noisy" power supplies:

First, some warnings:
  • Do not perform any modification described here unless you are familiar with the techniques involved in high voltage and high current circuits.  Accidental contact with mains voltages can be lethal!
  • You must make absolutely certain that all components that you use are rated for the voltage/current involved.  In particular, any capacitors that bypass from the AC (mains) input to the chassis must have the appropriate voltage and safety ratings to prevent the accidental imposition of potentially lethal mains voltages on the chassis/ground of the power supply and connected equipment!
  • In this article, some of the filtering components are depicted as being added prior to line fusing.  In all cases, such components must be appropriately safety-rated for the voltage and current.  In some areas (such as the EU) it may be permitted that such components are connected only after line fusing - and that line fusing is required on both leads of the AC power connection.  In any case, take sensible safety precautions!
  • Be certain that any inductors used are rated for the current involved and that their insulation is capable of withstanding the voltage applied.  For the bifilar chokes on the AC input, these must be rated for at least 4-5 amps while the output choke ("L" in Figure 2) should be heavy enough to handle 23-25 amps with minimal voltage drop.
  • Some of the techniques described in this article may not meet safety regulations in certain countries.  Examples might include:  The placement of RFI/EMI components before the fuse, the types of capacitors, the values of capacitors and the amount of leakage current that they would consume and/or place on the chassis ground, etc.  Please be aware of these issues and address them in a manner appropriate.  I thought that I'd mention that twice...
  • There are likely other things not mentioned here.  You have been warned!

For this discussion we are assuming several things about our power supply:
  • It is contained in metal case.  The case doesn't provide "shielding" as much as it provides a common, low-impedance point to which all filtering that helps remove switching energy can be connected.  Doing this prevents the formation of differential currents between the AC input and DC output leads which could impose low-level switching supply energy onto those leads!
  • It includes at least some of the above features to filter out switching components.  If this power supply was for, say,

Case study #1:  An older style model Samlex 1223

The first example is an older Samlex model 1223 power supply, a 23 amp, 13.8 volt unit that I bought in the late 1990's.  It was intended for use in, among other places, amateur radio stations, and is an inexpensive, yet fairly well-designed power supply.  Despite this, I noticed that it produced some low-level - yet annoying - spurious emissions on the lower HF amateur bands (160-40 meters) that were weakly audible at even higher frequencies.

In disassembling the unit I noticed immediately that it had just one AC input line filter.  Fortunately, there was enough room to wedge into it another bifilar choke (scavenged from a junked power supply) and the necessary bypass capacitors.

Figure 3:
Added bifilar choke on the AC input side of the old style Samlex 1223 power supply having been attached to the rear wall using RTV ("Silicone") adhesive.  In the foreground (lower-right) a pair of capacitors were added to the power line on the power cord receptacle, represented by "Ca" and "Cb" in Figure 1, above.  Because this power supply can
draw 300-400 watts, be certain that the added choke is rated for the expected current.
Click on the image for a larger version.
In Figure 3 you can see this modification with the added bifilar choke attached, using RTV, to the back wall of the power supply with the added capacitors (see "Ca" and "Cb" in Figure 1, above) soldered directly onto the IEC power cord receptacle.

I also put an oscilloscope across the DC output terminals and noticed that even though the DC output itself was quite clean - just a few millivolts of residual switcher energy - I saw few hundred millivolts of switcher energy when I measured between the chassis of the power supply and either of the DC outputs:  See Figure 4, below.

Figure 4:
The waveform present between either DC output and the chassis of the unmodified power supply.  The frequency/time noted in the box in the lower left is measured between the two purple vertical lines.
Click on the image for a larger version.

The magnitude of the "square" portions of the waveform are on the order of 130 millivolts with the extents of the high-frequency spikes going out to at least 268 millivolts:  It is this energy that is going to cause us the most grief!  This waveform looked the same whether I measured between the V- terminal and ground or the V+ terminal and ground, but this was not surprising since I already knew that from measuring across V- and V+, the waveform was quite clean.

At this point I had a choice:  Should I simply short the V- to the chassis ground and risk a ground loop, or install a capacitor?  Preferring to NOT subject myself to the possibility of a ground loop and the possibility of induced AC hum in the future, I rummaged around in my capacitor collection and found a large, 3.3uF plastic capacitor with a 200+ volt rating - probably something scrapped from an old switching supply or computer monitor.  When I connected this between the V- lead and the chassis of the power supply, I got the waveform in Figure 5, below:

Figure 5:
The output of the power supply after adding the capacitors to the output with the same vertical/horizontal scale as the plot in Figure 4.  Notice that only a fraction of the original "grunge" remains!
Click on the image for a larger version.

As you can see, there is a significant improvement!  The narrow spikes are much lower in amplitude (about 66 millivolts peak-peak rather than 268 millivolts) and, although it is a bit difficult to see in the above trace, the pulses are also much slower in their rise/fall time.  This last point (pun intended!) is important since it is the rate of change (dV/dT) of these pulses that dictate how much harmonic content they have, so between their reduction in amplitude and their being "slowed" considerably, this power supply was now VERY much "quieter."

Figure 6, above, shows the modifications made to the power supply and here are the steps:
  • I found a small piece of glass-epoxy circuit board material and cut it to fit the empty space above the DC output terminals.
  • Flipping the power supply upside-down, I drilled a hole for a 6-32 machine screw through the case and piece of circuit board material.  Flipping the case upside-down ensured that metal cuttings would not fall into the power supply.
  • After de-burring the holes with a drill bit (also done with the power supply upside-down) I bolted the piece of circuit board material to the case using some "star" washers to ensure a solid, electrical connection.
  • Between the piece of circuit board - which is now connected to the metal chassis ground of the power supply's box - I soldered a 3.3uF plastic capacitor between it and the V- lead.  Any value of 0.47uF and up would be fine, but 2.2uF-4.7uF is better.
  • I also soldered a 2200uF, 25 volt low-ESR (switching supply-type) electrolytic capacitor between the V- and V+ terminals using short pieces of heavy (#12 AWG or larger) wire.  This wasn't really necessary, but it did knock down those small "spikes" in Figure 5 a bit more.
Figure 6:
The modifications of the DC output of the (older) Samlex 1223 switching power supply showing the added
capacitors.  The orange unit on the left is the plastic capacitor that suppresses the voltage differential between
the case and power supply output that contained the switching energy seen in Figure 4, above.
Click on the image for a larger version.

Putting the cover back on and testing it - even using it a few times during Field Day over the years - I have not observed that this power supply has caused any detectable interference, even when being placed next to a balanced-wire antenna tuner.

A Radio Shack model 22-510 power supply:

A couple years after getting the Samlex 1223 I noticed that the Radio Shack 22-510 power supply was on sale and grabbed one.  Rated for 25 amps, it is almost identical in size and shape to the Samlex 1223 and it had a permanently-attached power cord rather than a detachable computer-type cord.  Popping the cover I could tell that it was better filtered than the old Samlex in that it already had a 2-stage AC input line filter that strongly resembled that depicted in figure 1.  As with the Samlex, I noted that across its DC terminals the output was fairly clean, but like the Samlex, I observed a waveform that was nearly identical to that in Figure 4 between the chassis and ground.

Figure 7:
The modification to the output of the Radio Shack 22-510 power supply.  The orange capacitor is a 3.3uF unit connected between V- and the case while you can see a 1000 uF, 25 volt capacitor connected directly across the DC output terminals.
The added, series inductor - from the high-current output of a junked PC power supply - is contained within the
insulating piece of yellow heat-shrinkable tubing seen in the upper-right corner of this picture, above the head sink.
Click on the image for a larger version.
Figure 7 shows the modification to clean this up, but in this case I used a screw with a ring lug to make the connection to ground and in the picture you can see the 3.3uF capacitor connected between it and the V- output terminal.

Since I had noticed a small amount of switching noise (not bad, but not as clean as that in Figure 5) I rummaged around and found a small choke on the 5 volt, high current output of a junked PC power supply that had been wound with #12 AWG wire and would thus be capable of handling 25 amps without much voltage drop.  Clipping the red (V+) lead, I soldered this inline and it is shown in Figure 7, insulated with yellow heat-shrink tubing.  Across the V- and V+ outputs I attached a 1000 uF, 25 volt capacitor and the combination of these two components made its output at least as clean as that shown in Figure 7.  I probably would have been fine not doing this, but since I was already working on the power supply, anyway...

After I did all of the above I noticed that there was still some low-level noise on the power supply that was not at the switching frequency, but rather in the range of a few hundred Hz:  It wasn't strong enough to be a problem, but it annoyed me that it was there at all and I was curious as to its source.  What I soon realized was that this extra noise was coming from the small cooling fan on the chassis:  Unlike the fan on the Samlex power supplies which are thermostatically-controlled with an electronic heat sensing circuit, the fan on this power supply always runs and is connected across the DC output.

One insidious problem with these brushless DC fans is that they seem to have the uncanny ability to put some of their electronic commutating noise onto their power supply leads despite the fact that they draw only a hundred milliamps or so and are, in this case, connected to a high-current power supply with lots of filtering!  The fix for this is quite simple, however:  A series 10 ohm resistor and a 220uF, 16 volt capacitor.

Figure 8:
The added filtering for the fan supply to keep its "whine" out of the DC output, consisting of a 10 ohm resistor in series with the positive lead and a 220uF capacitor on the "fan" side between the fan's V+ and its ground.
Click on the image for a larger verion.

Figure 8, above, shows this modification with the capacitor connected across the "fan" side of the resistor on the fan's power supply leads.  Because it was convenient to do so I used RTV ("Silicone") adhesive to attach these added components to the back wall of the power supply, but I could have also enclosed them in heat-shrink tubing.  With this modification the fan electrical noise was completely removed from the power supply's output and the fan ran slightly slower due to the voltage drop across the 10 ohm resistor and would likely last a bit longer - but it still moved more than enough to keep it cool under full load.

When I was done this power supply, too, was now very "clean".

A newer Samlex 1223 power supply:

Earlier this year I spotted a brand new, in-the-box Samlex 1223 for a really good price at a swap meet and couldn't resist getting it.  When I opened it up I could see that since my older '1223 had been built, they'd made some improvements:
  • Like the Radio Shack unit, it now had a 2-stage input line filter.
  • They'd changed the output connector from binding posts to screw-type compression terminals.
  • Both the V- and V+ output leads were routed through one large ferrite bead.
At this point I'll mention, again, the ferrite bead:  While it is a common technique to run power supply leads through such a device, simply passing wires through one of these will not likely add enough reactance to provide a significant degree of RFI suppression at lower HF frequencies!  What's more, the effect of this added reactance is not well-utilized unless you add some capacitors to the output as well as shunt the residual RFI to the chassis.

To be sure, I had not tested the unmodified power supply against the others that I'd modified to see how "clean" they were in terms of causing interference to HF operations but reports indicate that these newer Samlex 1223's are better than the older version in that regard but were still known to cause objectionable interference in some cases.

Interestingly, when I placed the oscilloscope between the V- and the chassis of the power supply I got almost exactly the same waveform as I'd gotten with the older Samlex 1223 and the Radio Shack 22-510 power supply depicted in Figure 4 so I knew what I had to do.

Figure 9:
Modifications to the new version of the Samlex 1223 power supply.  On the left can be seen the orange 3.3uF capacitor along with a heavy (#12 AWG) wire running from the V- lead to the added 1000uF, 25 volt capacitor.
Click on the image for a larger version.
As can be seen in Figure 9 I did exactly the same modification as was done on the Radio Shack power supply in Figure 7 - the only difference was that didn't need to add the extra choke in the V+ lead and I also had to route a piece of insulated heavy copper wire from the V- terminal across the top of the terminals to the added 1000uF, 25 volt capacitor since there wasn't enough room to locate it elsewhere.  In this case it was important to use a heavy-gauge (#12 AWG or heavier) wire for this connection since not only was the capacitor's lead not long enough to reach in the first place, but its small gauge lead (perhaps #22 or #24) offered enough resistance/reactance that the capacitor's suppression of some of the residual switching energy was degraded:  This just goes to show how, when dealing with high frequency switching supplies and high currents, how even a little bit of extra wire can cause a difference in performance!

Conclusions:

I've made these modifications to these power supplies over the years as I've acquired them and was somewhat surprised to see that they all have the same issue in common:  Significant switching energy between their cases and their DC output lines.  Fortunately, the "fixes" outlined above seem to be very effective and add minimal safety risk to their use and these three switching supplies that no longer cause any noticeable RFI, even when placed very close to the feedpoint of an HF antenna.

In addition to keeping these power supplies clean at HF, I also wanted to make sure that they caused minimal disruption at MF, LF and VLF frequencies (e.g. those below the AM broadcast band - where there are amateur allocations at 600 and 2200 meters) where I occasionally listen.  Because of these lower frequencies it is much more difficult to keep them from causing interference for several reasons:
  • Rather than being several 10's of times higher than the switching supply frequency, I might actually be listening on the switcher frequency - or on one of its first few harmonics.
  • At these lower frequencies the amount of inductance and capacitance in the filters may not be adequately high to effectively remove enough of the switching energy.
It was for this reason that I used the fairly large (3.3uF) case-to-V- coupling capacitor as well as adding the 2nd bifilar choke to the older Samlex power supply - not to mention the extra 1000 uF capacitors across the output leads of the supplies themselves.
Most of the time I don't even notice any of these power supplies causing interference, but on those occasions when I do (e.g. if I'm listening around 30-300 kHz, I may hear it) I can just shut them off for the duration.

Unfortunately, I also have other power supplies around the ham shack and the house (for the computer/monitor, the DSL modem, in the compact fluorescent and LED lighting, etc.) that are far "dirtier" and, at some point, these will require some action to clean them up - but that's another article!

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


[End]

This page stolen from ka7oei.blogspot.com

4 comments:

  1. Thanks this was very helpful. What are some typical values of inductors you have seen?

    ReplyDelete
    Replies
    1. On the AC input inductors - which are typically dual winding to suppress common-mode currents - the inductance values are typically in the range from 100uH to 10's of milliHenries: The ones depicted in the pictures are in the milliHenry range, potentially providing many ohms of reactance at frequencies of hundreds of kHz and up.

      Since it is practical to use only a fraction of a microfarad of capacitance across the AC mains and to "ground" to minimize "leakage" currents due to efficiency and safety issues - which can have ohms of reactance at the low frequency end - this high inductance is appropriate and necessary to suppress these signals.


      On the high-current output the inductance is typically in the 10's of uH range, partially as a matter of practicality as it would be difficult to get a lot of inductance with a small amount of heavy-gauge wire with a reasonable-sized inductor that could also handle 10's of amps.

      At DC and low voltage it is practical to use large, high-quality capacitors that have fractions of ohms of reactance at the low end of the frequencies to be blocked (e.g. a few tens/hundreds kHz) so by the time one gets to HF, these relatively small amounts of inductance will have 10's or hundreds of ohms reactance and the (good quality!) bypass capacitors will do their job in shunting residual RF to the case ground, containing it within.

      73!

      Delete
  2. I didn't make extensive measurements of the output ripple of the supply under the rated load, but I've never found it to be an issue.

    I must admit that I don't even remember the AC frequency ripple across the power supply's output terminals at load, but I do know that once I had done the modifications described above to reduce the "chassis to output terminal" noise depicted in Figures 4 and 5 the supplies became "invisible" from an RFI standpoint.

    ReplyDelete
  3. This is very helpful, as I study for the Amateur Extra exam. Thanks.

    ReplyDelete





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!