RF Quiet Wall warts

On Sat, 26 Aug 2017 18:33:12 -0400, krw@notreal.com wrote:

>Over a mile away?  Did you sic the FCC on them?

Nope.  Complaining to the FCC doesn't work unless one has a VIP (very
important person) behind the complaint.  Basically, the FCC is out of
the enforcement business, which is now handled by the Dept of Justice,
which won't touch any case unless the recovered fines exceed the cost
of investigation, enforcement, and the inevitable compromises.  If I
had discovered something that might generate a few thousand dollars in
fines for the Justice Dept, I might have gotten their attention.

>Good point, though,
>they should ban all those new fangled fluorescent and LED lights and
>go back to good old tungsten.

Personally, I prefer candles, natural gas, and kerosene lamps.  The
solution to technical problems is usually not to ban the new
technology.

>Nonsense.  Receivers are sensitive to the in-band energy.  Spreading
>it out *helps* EMI.  It has other issues but it does do exactly what
>it was intended to do.  BTW, EMI tests aren't just peak tests.  The
>unit has to test quasi-peak and average masks, as well.

Look at a spread spectrum clock on a spectrum analyzer:
<https://en.wikipedia.org/wiki/Spread_spectrum>
Some PWM chips have a spread spectrum (SS) on/off pin or function. Try
it both ways.  What you get with SS turn on is something like the
photo:
<https://en.wikipedia.org/wiki/Spread_spectrum#/media/File:Aaronia_Spectrum_Analyzer_Software.jpg>
What you get with SS off is a single coherent frequency, but with an
amplitude equal to all the energy (area under the curve for each
sideband) piled on top of the fundamental.  However, you're right
about SS helping EMI.  Few of todays clocked devices would ever pass
EMI/RFI testing without it.

>Which would do absolutely nothing for harmonics (the real problem)
>unless all of the harmonics of your magic switching frequency were
>also reserved.  13.56MHz would be rather useless as a reserved
>switcher frequency, considering that this is rather high for the
>fundamental (see above for harmonics).  

Agreed.  Ideally, all the harmonics would be reserved.  However, since
the fundamental would probably need to be a fairly low frequency
(50Khz to 2MHz), the number of reserved frequencies needed to handle
the harmonics could easily be huge.  The assumption was that the
higher harmonics are easier to filter with ferrite beads and easier to
shield.  I don't recall how many harmonics we expect to ask for
allocations, but I think it was 11.

>The issue is proper design.  It can be, and is, done.  Some industries
>have stricter limits than the FCC's and they deal with switching power
>supplies without too much trouble.

They have stricter limits today because the FCC limits didn't work for
them, partly because they allowed spread spectrum clock modulation.
Please remember that this was about 1977 and the life was a bit
different.  We didn't have decent power MOSFETs, IGBT devices, SiGe,
and silicon nitride devices.  Most switchers of the day used BJT,
crude IC's, and no spread spectrum.

>>So, while you're enjoying your switcher powered gizmos and gadgets,
>>there are a few of us that are doing what we can to clean up the mess
>>that your attitude seem to have left behind.

>Not too self-important.  No, not you!

Sorry.  There are a few topics that produce an emotional response,
instead of a technical discussion.  This is one of those and was a
source of frustration for many years.  I usually try to suppress my
emotions but lost it this time.  I'll try to be more diplomatic in the
future.

-- 
Jeff Liebermann     jeffl@cruzio.com
150 Felker St #D    http://www.LearnByDestroying.com
Santa Cruz CA 95060 http://802.11junk.com
Skype: JeffLiebermann     AE6KS    831-336-2558

On Sunday, 27 August 2017 01:58:55 UTC+1, Jeff Liebermann  wrote:

> I'm fairly sure the scheme would work if the switchers were frequency
> stabilized with either a crystal oscillator or ceramic resonator. Same
> as with any uP or clocked device.  Most of the energy would be in the
> fundamental and 3rd harmonic.  As the harmonic frequency increases, it
> becomes easier to shield and filter.
> 
> Square waves have 67% of the energy in the fundamental, and the
> remaining 33% in odd order harmonics.  As the duty cycle changes the
> waveform become asymmetrical, producing more and more even order
> harmonics.  Eventually, for really low duty cycles, it becomes a pulse
> train which produces a comb line of equal power signals.  
> <http://www.analog.com/en/analog-dialogue/articles/understanding-switching-reg-output-artifacts.html>
> 
> Obviously, we're not suggesting that all switchers operate at
> 13.56MHz, which incidentally was one of the original diathermy
> frequencies.
> <https://en.wikipedia.org/wiki/Diathermy#Short_wave>
> There are other frequencies, but none low enough to be usable in a
> switching power supply:
> <https://en.wikipedia.org/wiki/ISM_band>

Even without any filtering the interference would be confined to a list of frequencies. Perhpas it could even have done away with the need to filter at all?

You could maybe cater for a range of SMPSU frequencies by permitting eg unfiltered operation at a ladder-like range of frequencies, leaving at least most of the spectrum free of junk, rather than just one frequency, which would never work long or medium term.


NT

On Sat, 26 Aug 2017 17:58:58 -0700, Jeff Liebermann <jeffl@cruzio.com>
wrote:

>On Sat, 26 Aug 2017 23:27:55 +0200, Piotr Wyderski
><peter.pan@neverland.mil> wrote:
>
>>Jeff Liebermann wrote:
>>
>>> About 30 years ago, a few people saw what was coming and tried to
>>> convince the FCC and the ITU to allocate specific frequencies for
>>> switching power supplies.  Such a frequency would be declared the
>>> dumping ground for switcher RFI, such as 13.56MHz.  We worked out a
>>> preliminary frequency plan, which was far from ideal, but at least a
>>> good start.  The idea was summarily killed when the spread spectrum
>>> trick was accepted by the FCC as the "solution" to the EMI/RFI
>>> problem.
>
>>But would it ever work? The frequency of the switchers is mostly limited
>>by the magnetics and then by the MOSFET drivers/losses. That puts them 
>>in the 50kHz...1MHz band, which is/used to be densely populated. They
>>use square waves of variable duty cycle, with the harmonic garbage way
>>above their "baseband". The resonant soft switchers are a novelty from
>>the mass market perspective. So how would a static allocation of, say,
>>100kHz help here?
>>
>>	Best regards, Piotr
>
>I'm fairly sure the scheme would work if the switchers were frequency
>stabilized with either a crystal oscillator or ceramic resonator. Same
>as with any uP or clocked device.  Most of the energy would be in the
>fundamental and 3rd harmonic.  As the harmonic frequency increases, it
>becomes easier to shield and filter.

That assumption is simply wrong.  It's never the fundamental or third
harmonic that's the problem, rather the 101st.  The problem is almost
always in the parasitics around the switch node. Snubbers help but
layout is the key. 

>Square waves have 67% of the energy in the fundamental, and the
>remaining 33% in odd order harmonics.  As the duty cycle changes the
>waveform become asymmetrical, producing more and more even order
>harmonics.  Eventually, for really low duty cycles, it becomes a pulse
>train which produces a comb line of equal power signals.  
><http://www.analog.com/en/analog-dialogue/articles/understanding-switching-reg-output-artifacts.html>

Irrelevant.

>Obviously, we're not suggesting that all switchers operate at
>13.56MHz, which incidentally was one of the original diathermy
>frequencies.
><https://en.wikipedia.org/wiki/Diathermy#Short_wave>
>There are other frequencies, but none low enough to be usable in a
>switching power supply:
><https://en.wikipedia.org/wiki/ISM_band>
>So, we had to find an empty frequency and go through the ordeal
>process of getting the FCC, ETSI, CEPT, and later the ITU to allocate
>or reassign new frequencies.  At the time (about 1977) it was apparent
>that the long wave marine navigation beacon band (160KHz to 530KHz)
>was eventually going to be replaced by NAVSTAR (now known as GPS)
>system.  Since it was expected to take at least 20 years to get
>approval for a reallocation of the LW spectrum, this was a reasonable
>strategy.  As it turned out, most of the US beacons were turned off in
>1997, almost exactly 20 years later.  There was also a problem with
>the US Navy, which wanted some of the low end of the LW beacon band
>for submarine communications.  However, none of this happened when
>spread spectrum PWM power supply modulation arrived and killed the
>plan.

The idea is simply nuts.  It's not surprising no one listened.

On Sat, 26 Aug 2017 18:30:08 -0700, Jeff Liebermann <jeffl@cruzio.com>
wrote:

>On Sat, 26 Aug 2017 18:33:12 -0400, krw@notreal.com wrote:
>
>>Over a mile away?  Did you sic the FCC on them?
>
>Nope.  Complaining to the FCC doesn't work unless one has a VIP (very
>important person) behind the complaint.  Basically, the FCC is out of
>the enforcement business, which is now handled by the Dept of Justice,
>which won't touch any case unless the recovered fines exceed the cost
>of investigation, enforcement, and the inevitable compromises.  If I
>had discovered something that might generate a few thousand dollars in
>fines for the Justice Dept, I might have gotten their attention.
>
>>Good point, though,
>>they should ban all those new fangled fluorescent and LED lights and
>>go back to good old tungsten.
>
>Personally, I prefer candles, natural gas, and kerosene lamps.  The
>solution to technical problems is usually not to ban the new
>technology.

Yet that's what you're proposing.
>
>>Nonsense.  Receivers are sensitive to the in-band energy.  Spreading
>>it out *helps* EMI.  It has other issues but it does do exactly what
>>it was intended to do.  BTW, EMI tests aren't just peak tests.  The
>>unit has to test quasi-peak and average masks, as well.
>
>Look at a spread spectrum clock on a spectrum analyzer:
><https://en.wikipedia.org/wiki/Spread_spectrum>

Oh, good grief!

>Some PWM chips have a spread spectrum (SS) on/off pin or function. Try
>it both ways.  What you get with SS turn on is something like the
>photo:

OF course they do.  Are you daft?

><https://en.wikipedia.org/wiki/Spread_spectrum#/media/File:Aaronia_Spectrum_Analyzer_Software.jpg>
>What you get with SS off is a single coherent frequency, but with an
>amplitude equal to all the energy (area under the curve for each
>sideband) piled on top of the fundamental.  However, you're right
>about SS helping EMI.  Few of todays clocked devices would ever pass
>EMI/RFI testing without it.

YEs, and a receiver is sensitive to the area under its sensitivity
curve, too.  The idea is to spread the interference signal further out
than the interferred.
>
>>Which would do absolutely nothing for harmonics (the real problem)
>>unless all of the harmonics of your magic switching frequency were
>>also reserved.  13.56MHz would be rather useless as a reserved
>>switcher frequency, considering that this is rather high for the
>>fundamental (see above for harmonics).  
>
>Agreed.  Ideally, all the harmonics would be reserved.  However, since
>the fundamental would probably need to be a fairly low frequency
>(50Khz to 2MHz), the number of reserved frequencies needed to handle
>the harmonics could easily be huge.  The assumption was that the
>higher harmonics are easier to filter with ferrite beads and easier to
>shield.  I don't recall how many harmonics we expect to ask for
>allocations, but I think it was 11.

It's not uncommon for the interference to be the 100th harmonic, or
even higher.  I've seen combs into the FM band and above.  
>
>>The issue is proper design.  It can be, and is, done.  Some industries
>>have stricter limits than the FCC's and they deal with switching power
>>supplies without too much trouble.
>
>They have stricter limits today because the FCC limits didn't work for
>them, partly because they allowed spread spectrum clock modulation.
>Please remember that this was about 1977 and the life was a bit
>different.  We didn't have decent power MOSFETs, IGBT devices, SiGe,
>and silicon nitride devices.  Most switchers of the day used BJT,
>crude IC's, and no spread spectrum.

Who cares about "of the day"?  Even so, you're talking about 13MHz,
which would have been silly in 1977.
>
>>>So, while you're enjoying your switcher powered gizmos and gadgets,
>>>there are a few of us that are doing what we can to clean up the mess
>>>that your attitude seem to have left behind.
>
>>Not too self-important.  No, not you!
>
>Sorry.  There are a few topics that produce an emotional response,
>instead of a technical discussion.  This is one of those and was a
>source of frustration for many years.  I usually try to suppress my
>emotions but lost it this time.  I'll try to be more diplomatic in the
>future.

On Sat, 26 Aug 2017 10:18:02 -0700, Jeff Liebermann <jeffl@cruzio.com>
wrote:

>On Thu, 24 Aug 2017 21:35:20 -0400, Martin Riddle
><martin_ridd@verizon.net> wrote:
>
>>It's been a year or two since I listend to AM Radio, and in that time
>>theres a new Roku box and linksys router.  Well, I couldn't hear
>>anything  on the AM band when I checked a week ago. So the quest was
>>to find the culprits. Its these SMPS wall warts, they literly wipe out
>>the entire AM band, and also some chinese LED lights( 15w equiv
>>Grimaldi's ).
>>Tossing the led lights are easy, the wall warts not so.
>>I found the old DSL modem transformer wall wart and that is quiet.
>
>Welcome to my nightmare.  Ham radio operators have been dealing with
>the wall wart and other devices EMI/RFI for many years.  It only gets
>worse with new and more insidious sources of electrical noise arriving
>every day.  For example, I recently found a garden path full of these
>LED lights.  Each one has a built in switcher with little filtering
>and no shielding.
><http://www.ebay.com/itm/222204398564>
>Ferrite beads around the leads were a big help.
>
>"A Ham's Guide to RFI, Ferrites, Baluns, and Audio Interfacing"
><http://audiosystemsgroup.com/RFI-Ham.pdf>
>I do some computah work for the author.  I recently unloaded about 50
>lbs of linear wall warts on his doorstep.  Over the years, he has
>systematically replaced as many switchers with linear power supplies
>as possible.  The few switchers that can't be replaced, have large
>common mode toroids on both input and output leads.  Something like
>these:
><http://palomar-engineers.com/rfi-kits/computer-rfi-kits>
>
>See list of his other publications for other articles:
><http://audiosystemsgroup.com/publish.htm>
>such as:
>"Power Line Filters"
><http://audiosystemsgroup.com/PowerFilters.pdf>
>
>If you're going to be chasing down sources of EMI and RFI, the best
>technique is to kill the power to sections of the house at the breaker
>box.  This doesn't do anything for battery backed up devices, such as
>digital clocks, or deal with neighborhood sources, but otherwise works
>well.  For fine tuning, I use an AM portable receiver tuned to empty
>HF frequencies.  I use an Alinco DJ-X2 but any portable AM/FM/SW/AIR
>receiver will work.  When I'm dealing with more exotic noise sources,
>I use a spectrum analyzers.
>
>>Perusing the Innernet, I see some mention Apple supplies and the
>>Powerstream SMPS supplies are quiet too, any one here with experiance
>>with them? 
>
>I haven't done any bench tests with an LISN (line impedance
>stabilization network) comparing various switchers.  Powerstream does
>not actually manufacture the products they sell but buys them from
>various vendors:
><https://www.powerstream.com/power2.html>
>My guess(tm) is that I would expect wide variations in EMI radiation
>characteristics.  The various genuine Apple power supplies that I own
>seem fairly quiet for both conducted and radiated noise, but not
>totally quiet.  An external filter would still be useful.  I don't
>know how they compare with counterfeit and cloned Apple power supplies
>or similar switchers.  

Thanks for the info.

The RIF-Ham.pdf is very helpful.

Palomar-Engineers that's a name I haven't heard in years, I have a box
of cores from them that I got in the mid 90's.
I see they have a wall wart kit for EMI, it's under this link:
<http://palomar-engineers.com/rfi-kits/acdc-power-line-chokes>

The Clip on bead they use (square one) looks like the Lab favorite
from Fairite, the one I have does not have the PN on it anymore.
I've done the multi-turn thru the cores before ( like George
suggested) with success, but these SMPS's are just screamers and hard
to easily suppress.

The last RFI problem I had was with one of those 12DC auto cell phone
chargers for a Nokia phone.  Wound up putting a multi turn bead on and
loading the output with 0.1uf 0.01uf and a 100uf capacitor.  What was
interesting was that the little switcher they used was not wired per
the data sheet for the Nokia phone, once I wired it for my Motorola
phone (5V) it created nice birdies in the AM band.

Cheers

On Sat, 26 Aug 2017 22:12:37 -0400, krw@notreal.com wrote:

>On Sat, 26 Aug 2017 18:30:08 -0700, Jeff Liebermann <jeffl@cruzio.com>
>wrote:

>>Look at a spread spectrum clock on a spectrum analyzer:
>><https://en.wikipedia.org/wiki/Spread_spectrum>
>
>Oh, good grief!

That's all I could find while eating dinner.  This article explains it
in more detail and with nicer looking spectrum analyzer screen shots.
<https://interferencetechnology.com/spread-spectrum-clock-generation-theory-and-debate/>

>YEs, and a receiver is sensitive to the area under its sensitivity
>curve, too.  The idea is to spread the interference signal further out
>than the interferred.

Right and spreading the signal over a wider bandwidth allows it to
interfere with more narrowband receivers operating on adjacent
channels.  Instead of trashing one receiver, the SS noise causes
problems with all of them.  Of course, the noise power bandwidth will
be less, so the interference would be less, which is why the FCC
accepted the idea.  Whether this is a good tradeoff is subject to
considerable debate.

>It's not uncommon for the interference to be the 100th harmonic, or
>even higher.  I've seen combs into the FM band and above.  

Sure, but like I previously muttered, a common ferrite bead on the
in/out wires can stop 100 MHz trash quite easily.  Trying to do the
same at 100 KHz would require a much larger ferrite bead.  If you're
seeing a comb line at 100 MHz (assuming the harmonics weren't created
in the spectrum analyzer front end from overload), then someone didn't
do a very good job of designing the switcher.

>Who cares about "of the day"?

I was commenting about the lack of modern solutions, techniques, and
devices that were not available in about 1977 when we were working on
a political solution.

>Even so, you're talking about 13MHz,
>which would have been silly in 1977.

Yep.  See my comments elsewhere about the selection of frequencies and
the lack of harmonically related possibilities among the ISM
frequencies.  Basically, we were trying to have allocated marine LW
beacon band frequencies and their harmonics after the LW beacon band
transmitters were shut off.  The allocation of RF spectrum is normally
not done for the convenience of power supply manufacturers and we
didn't have even a preliminary list of usable frequencies.  13.56 MHz
would probably have been a harmonic of some lower frequency.


-- 
Jeff Liebermann     jeffl@cruzio.com
150 Felker St #D    http://www.LearnByDestroying.com
Santa Cruz CA 95060 http://802.11junk.com
Skype: JeffLiebermann     AE6KS    831-336-2558

On Sat, 26 Aug 2017 22:45:11 -0400, Martin Riddle
<martin_ridd@verizon.net> wrote:

>Thanks for the info.

Y'er welcome.

>The RIF-Ham.pdf is very helpful.

It's RFI-ham.pdf.  It is much better than most articles on RFI/EMI
suppression because it has a good mix of theory, experience, tests,
and sources.

>Palomar-Engineers that's a name I haven't heard in years, I have a box
>of cores from them that I got in the mid 90's.
>I see they have a wall wart kit for EMI, it's under this link:
><http://palomar-engineers.com/rfi-kits/acdc-power-line-chokes>

I'm not sure of the current status of the company.  The owner died in
2013 and things seem to be rather confused since then:
<http://www.eham.net/ehamforum/smf/index.php?topic=109940.0>
You can probably build all of the ferrite filters and chokes yourself.
>The Clip on bead they use (square one) looks like the Lab favorite
>from Fairite, the one I have does not have the PN on it anymore.
>I've done the multi-turn thru the cores before ( like George
>suggested) with success, but these SMPS's are just screamers and hard
>to easily suppress.

The smash-on split bead variety are quite common and cheap:
<https://www.ebay.com/sch/i.html?_nkw=split+ferrite+bead>
However, you can get more turns and heavier gauge wire through a large
toroid.

>The last RFI problem I had was with one of those 12DC auto cell phone
>chargers for a Nokia phone.  Wound up putting a multi turn bead on and
>loading the output with 0.1uf 0.01uf and a 100uf capacitor.  What was
>interesting was that the little switcher they used was not wired per
>the data sheet for the Nokia phone, once I wired it for my Motorola
>phone (5V) it created nice birdies in the AM band.

I built constant current dummy load for testing power supplies and
chargers.  These should work better than my mess:
<http://www.ebay.com/itm/152159721225>
<http://www.ebay.com/itm/282437868849> (I just ordered one of these)
<https://www.ebay.com/sch/i.html?_nkw=USB+Adjustable+Constant+Current+electronic+load+Tester>
At some point along the load curve, I can often get a cheap Chinese
bottom of the line charger to go unstable and oscillate.  


-- 
Jeff Liebermann     jeffl@cruzio.com
150 Felker St #D    http://www.LearnByDestroying.com
Santa Cruz CA 95060 http://802.11junk.com
Skype: JeffLiebermann     AE6KS    831-336-2558

Phil Hobbs <pcdhSpamMeSenseless@electrooptical.net> wrote:
> On 08/26/2017 02:36 PM, Rob wrote:
>> Jeff Liebermann <jeffl@cruzio.com> wrote:
>>> I do some computah work for the author.  I recently unloaded about 50
>>> lbs of linear wall warts on his doorstep.  Over the years, he has
>>> systematically replaced as many switchers with linear power supplies
>>> as possible.
>> 
>> Swiching power supplies are much more efficient, both in operation
>> and in manufacturing.  They use a lot less resources to manufacture.
>> 
>> That is what the world needs more than clean HF bands.  Live with it.
>> 
>
> Gee, it's great that there are people like you, People Who Know Best. 
> Otherwise we the great unwashed would have no one to tell us how to 
> think, and then where would we be?
>
> There's nothing that prevents one from making quiet switching supplies. 
> They just cost more.

I'm not "the one who knows best", I am the one that realistically sees
that the use of frequency bands in the world has irreversibly changed.

In the past the HF bands were used to do radio transmissions, now they
are emitted as unwanted byproduct by modern equipment, and they are
no longer used for radio.  At least not like before.

For sure it would be good to design switching supplies with low emission,
but don't think that would solve the problem of "HF bands not being
as quiet as they used to be".  There are too many other sources of
emissions that are not so easy to handle.  E.g. ethernet over powerline.

One can argue that many of them are just "bad ideas" and "irresponsible
engineering" but it is just irrelevant in a world where nobody cares
about HF radio reception except some old-hat radio amateurs.
In the modern society they are irrelevant and only hampering innovation.

Piotr Wyderski <peter.pan@neverland.mil> wrote:
> Rob wrote:
>
>> This is also the reason why nobody cares about the interference from
>> wall warts.  The AM (HF) bands are no longer used by the public.
>
> Germany and a fair fraction of Europe work thanks to the AM
> transmissions (DCF77@77.5kHz for time broadcast, DCF49@129kHz
> for remote control of power meters etc.). Unfortunately, the
> lower LF band is where the EMI is highest. There are PM signals
> too, but no consumer-grade devices use that due to the involved 
> complexity. There is also LORAN, but it is not very useful on land.

I know, I have clocks like that and I am seeing all the time how they
have more and more difficulty receiving the signal.

Both at home and at work, the reception of DCF77 is erratic at best:

     remote           refid      st t when poll reach   delay   offset  jitter
==============================================================================
 GENERIC(0)      .DCFa.           0 l 289m   64    0    0.000   -0.717   0.000

So, I now use GPS as a reference.

Rob wrote:

> I know, I have clocks like that and I am seeing all the time how they
> have more and more difficulty receiving the signal.

But they are just consumer products, they can fail without many
serious implications. Remote meter switching for the purpose of
proper bill calculation is another story, can end with a lawsuit.

> So, I now use GPS as a reference.

GPS requires direct sky visibility, which sometimes is a serious
problem. The VLF signals can be received easily deep in the concrete
buildings, it's just the noise that makes it hard to decode.

	Best regards, Piotr