SMPS mains circuit

[John Woodgate]

I'm afraid that this article is OT, since it isn't concerned with gun
control or US politics.

For investigations into problems that have arisen with mains harmonics
emission tests in Europe, I need to show the relevant standards
committee some results of simulations of SMPS with various supply source
impedances and supply voltages. I don't need to simulate the whole
thing, just from the mains input to the filter capacitor, plus a load
simulating the rest of the circuit. Because the actual outputs of the
SMPS are regulated, that load has effectively a negative resistance
element (it's an approximately constant power load, I think; as the d.c.
output voltage across the filter capacitor increases, the current drawn
by the switching circuit decreases).

I'd like to know typical component types and values for the part of the
circuit up to the filter capacitor, say for a 250 W supply, and how to
simulate that load with the negative resistive element. Any help will be
gratefully received.

 

[Allan Herriman]

I'm afraid that this article is OT, since it isn't concerned with gun
control or US politics.

For investigations into problems that have arisen with mains harmonics
emission tests in Europe, I need to show the relevant standards
committee some results of simulations of SMPS with various supply source
impedances and supply voltages. I don't need to simulate the whole
thing, just from the mains input to the filter capacitor, plus a load
simulating the rest of the circuit. Because the actual outputs of the
SMPS are regulated, that load has effectively a negative resistance
element (it's an approximately constant power load, I think; as the d.c.
output voltage across the filter capacitor increases, the current drawn
by the switching circuit decreases).

I'd like to know typical component types and values for the part of the
circuit up to the filter capacitor, say for a 250 W supply, and how to
simulate that load with the negative resistive element. Any help will be
gratefully received.

Wouldn't a typical 250W power supply use PFC?
If so, it would only have a negative resistance for long term changes
in the input voltage; short term (i.e. within a cycle) changes will
result in a positive resistance reading.

Regards,
Allan.

 

[legg]

I'm afraid that this article is OT, since it isn't concerned with gun
control or US politics.

For investigations into problems that have arisen with mains harmonics
emission tests in Europe, I need to show the relevant standards
committee some results of simulations of SMPS with various supply source
impedances and supply voltages. I don't need to simulate the whole
thing, just from the mains input to the filter capacitor, plus a load
simulating the rest of the circuit. Because the actual outputs of the
SMPS are regulated, that load has effectively a negative resistance
element (it's an approximately constant power load, I think; as the d.c.
output voltage across the filter capacitor increases, the current drawn
by the switching circuit decreases).

I'd like to know typical component types and values for the part of the
circuit up to the filter capacitor, say for a 250 W supply, and how to
simulate that load with the negative resistive element. Any help will be
gratefully received.

I'm not sure this is neccessary to simulate harmonic levels, unless
you are concerned with load changes that are harmonically linked - a
highly unlikely situation, and one currently avoided in typical
testing procedures.

A static power level that changes for repeated trials should be quite
adequate. All subject devices without a PFC front end are subject to
similar load variations over the capacitor voltage cyclic change - so
why would the test method itself be atsource of a problem?

Also, 'typical' values of low frequency input components would have to
cover a range and combination of values to give results that were not
unknowingly skewed, anyways.

C rectifier low frequency is sized for <20% droop.
Cac between 0.22 and 1.2uF in two positions.
Lcm between 2.0 and 10mH once or twice. Physically sized for power
loss.
Ldm between 100 and 470uH. Physically sized for power loss.

Commercial canned filter values are a good guideline to the last three
component functions.

I'm interested in the projected problem with the current test
procedures that you are attempting to illustrate. What is this
problem? Or are you just suggesting that simulations need to be of
increased accuracy?

Or are you concerned with accurate wattage cutt-off points in the
standard's application? There can't be a lot of profit in limboing
under standard limitations - why the concern?

RL

 

[John Woodgate]

I read in sci.electronics.design that Allan Herriman <allan.herriman.hat
[email protected]> wrote (in <dvqpe0te54t2apurk96i8c0r2o04t6qj
[email protected]>) about 'SMPS mains circuit', on Thu, 8 Jul 2004:

Wouldn't a typical 250W power supply use PFC?
If so, it would only have a negative resistance for long term changes
in the input voltage; short term (i.e. within a cycle) changes will
result in a positive resistance reading.

Judging by the intense efforts being made by the computer industry to
get the harmonic current emission limits relaxed, PFC is not their
favourite technique. And the problems, of course, don't normally arise
when PFC is present. So, I don't want PFC included in the exercise.

 

[John Woodgate]

I'm not sure this is neccessary to simulate harmonic levels, unless
you are concerned with load changes that are harmonically linked - a
highly unlikely situation, and one currently avoided in typical
testing procedures.

You are trying to second-guess the basic problem. Please believe me
instead, when I say what I need to do.

A static power level that changes for repeated trials should be quite
adequate. All subject devices without a PFC front end are subject to
similar load variations over the capacitor voltage cyclic change - so
why would the test method itself be atsource of a problem?

Also, 'typical' values of low frequency input components would have to
cover a range and combination of values to give results that were not
unknowingly skewed, anyways.

I can detect and allow for that. But the problem doesn't just occur with
a small percentage of equipment, so significant skewing doesn't seem to
occur.

C rectifier low frequency is sized for <20% droop.
Cac between 0.22 and 1.2uF in two positions.
Lcm between 2.0 and 10mH once or twice. Physically sized for power
loss.
Ldm between 100 and 470uH. Physically sized for power loss.

Commercial canned filter values are a good guideline to the last three
component functions.

I'm interested in the projected problem with the current test
procedures that you are attempting to illustrate. What is this
problem? Or are you just suggesting that simulations need to be of
increased accuracy?

The problem is that the results from tests of products offered for sale
can differ significantly, in mathematical terms, from those obtained
from the type-test sample on which the Declaration of Conformity was
based. We know why, but we need to explain in the standard when these
differences are significant in terms of compliance with the EMC
directive and when they are not. In order to develop a suitable text, we
need to present examples and explanations of the effect. The simulations
are intended to provide the examples.

Or are you concerned with accurate wattage cutt-off points in the
standard's application? There can't be a lot of profit in limboing
under standard limitations - why the concern?

Oh, there can be. Consider a laptop computer, in which space is at a
premium. If its active power consumption is below the 75 W lower bound
of the Class D limits, it doesn't have to include PFC. If it's above 75
W, it does, and PFC takes up a significant amount of board area. But
that is not what my enquiry is about.

 

[Allan Herriman]

I read in sci.electronics.design that Allan Herriman <allan.herriman.hat
[email protected]> wrote (in <dvqpe0te54t2apurk96i8c0r2o04t6qj
[email protected]>) about 'SMPS mains circuit', on Thu, 8 Jul 2004:


Judging by the intense efforts being made by the computer industry to
get the harmonic current emission limits relaxed, PFC is not their
favourite technique. And the problems, of course, don't normally arise
when PFC is present. So, I don't want PFC included in the exercise.

Oh, now I see. The manufacturers want to save $7 (or whatever a PFC
costs) so they want the harmonic current spec relaxed.

Regards,
Allan.

 

[Genome]

| I'm afraid that this article is OT, since it isn't concerned with gun
| control or US politics.
|
| For investigations into problems that have arisen with mains harmonics
| emission tests in Europe, I need to show the relevant standards
| committee some results of simulations of SMPS with various supply
source
| impedances and supply voltages. I don't need to simulate the whole
| thing, just from the mains input to the filter capacitor, plus a load
| simulating the rest of the circuit. Because the actual outputs of the
| SMPS are regulated, that load has effectively a negative resistance
| element (it's an approximately constant power load, I think; as the
d.c.
| output voltage across the filter capacitor increases, the current
drawn
| by the switching circuit decreases).
|
| I'd like to know typical component types and values for the part of
the
| circuit up to the filter capacitor, say for a 250 W supply, and how to
| simulate that load with the negative resistive element. Any help will
be
| gratefully received.
| --
| Regards, John Woodgate, OOO - Own Opinions Only.
| The good news is that nothing is compulsory.
| The bad news is that everything is prohibited.
| http://www.jmwa.demon.co.uk Also see http://www.isce.org.uk

See post with same title in ABSE

DNA

 

[John Woodgate]

See post with same title in ABSE

Thanks in advance, because it doesn't seem to have arrived there yet.

 

[Harry Dellamano]

See post with same title in ABSE

DNA

Maybe you should add a LISN and some input wires parasitics to get a more
realistic EMI spectrum.
Regards
Harry

 

[John Woodgate]

I read in sci.electronics.design that Harry Dellamano

Maybe you should add a LISN and some input wires parasitics to get a more
realistic EMI spectrum.

This is about mains harmonics emissions. No LISN allowed or required.
See IEC/EN 61000-3-2. Supply source impedance is a variable in the
investigation.

 

[Genome]

| I'm afraid that this article is OT, since it isn't concerned with gun
| control or US politics.
|
| For investigations into problems that have arisen with mains harmonics
| emission tests in Europe, I need to show the relevant standards
| committee some results of simulations of SMPS with various supply
source
| impedances and supply voltages. I don't need to simulate the whole
| thing, just from the mains input to the filter capacitor, plus a load
| simulating the rest of the circuit. Because the actual outputs of the
| SMPS are regulated, that load has effectively a negative resistance
| element (it's an approximately constant power load, I think; as the
d.c.
| output voltage across the filter capacitor increases, the current
drawn
| by the switching circuit decreases).
|
| I'd like to know typical component types and values for the part of
the
| circuit up to the filter capacitor, say for a 250 W supply, and how to
| simulate that load with the negative resistive element. Any help will
be
| gratefully received.
| --
| Regards, John Woodgate, OOO - Own Opinions Only.
| The good news is that nothing is compulsory.
| The bad news is that everything is prohibited.
| http://www.jmwa.demon.co.uk Also see http://www.isce.org.uk

As an aside a Google search on Middlebrook Criterion gives

http://www.smpstech.com/filtertl.htm

http://www.smpstech.com/filter00.htm

DNA

 

[Ken Smith]

I'd like to know typical component types and values for the part of the
circuit up to the filter capacitor, say for a 250 W supply, and how to
simulate that load with the negative resistive element. Any help will be
gratefully received.

Based on my repair of exactly one IBM PC style power supply:

The input had an off the shelf EMI filter/power plug

After it was a pair of black objects with radial leads that seemed to be
thermistors or simply power resistors. Their resistance was at most about
the same range as the shorted lead resistance of the Ohmmeter. I'm
guessing they were in fact resistors of about 1/2 Ohm.

Next came the bridge and 110/220V switch. In the 220 mode the input is a
bridge in the 110 mode it is a doubler. It used a series pair of 220uF
capacitors as the filter cap.

Next came a common mode choke and another series pair of 220uF capacitors.

The 5V load side had a 1000uF capacitor, a small value inductor, and
another capacitor with a value I don't remember. Perhaps the second
cap. was also 10000uF. I recall it as the same mechanical size.

The frequency of switching seemed to vary but was about 200KHz.

I doubt the gain cross over frequency of the servo loop would be much
above 20KHz. Perhaps it is lower. If the harmonics you are concerned
with ar all below 1KHz, we can assume that the DC-DC converter section had
no delay in its servoing.


I pulled the lid off of a spare PC supply in the shop:

input cap = 4 each 200v 160u rubycon 230404

 

[John Woodgate]

I read in sci.electronics.design that Ken Smith <[email protected]>

If the harmonics you are concerned
with ar all below 1KHz, we can assume that the DC-DC converter section had
no delay in its servoing.

Thanks for your information. IEC 61000-3-2 requires control of harmonics
up to the 40th, which means 2.4 kHz in sixty-cycle land. The problem I
am investigating is not due to restricted servo bandwidth; any problem
due to that is already covered by the test duration requirement.

The problem I am working on is that products taken from the field and
measured give significantly (in the mathematical sense) different
results from the original type-test product on which the Declaration of
Conformity was based. Many of these variations are of no significance in
terms of passing or failing the test, but this has to be explained in
the standard, and we need to find the best way of doing that, **and***
justifying what we write.

There are three causes of the problem, that we know about;

- component tolerances in the equipment; this is the responsibility of
the manufacturer but there is widespread lack of knowledge about it,
according to the information I have;

- variations in the test supply impedance from one test-house to
another;

- incorrect test-set ups.

But we need more analysis and perhaps more field data. I'm trying to
contribute to the analysis.

 

[legg]

The problem I am working on is that products taken from the field and
measured give significantly (in the mathematical sense) different
results from the original type-test product on which the Declaration of
Conformity was based. Many of these variations are of no significance in
terms of passing or failing the test, but this has to be explained in
the standard, and we need to find the best way of doing that, **and***
justifying what we write.

If the variations have no signifigance, then perhaps it's because
there is already sufficient margin in the standard to cover variations
of this sort.

As the standards are written in absolute current value limits, the
assumption that worst case occurs at full load might be incorrectly
adopted. Loading tolerances might also be signifigant as product
inefficiencies are introduced, for economy's sake, after innitial
certification.

Has the testing body or representative ever been called to task for a
report that is unrepeatable? I doubt it.

Has retesting ever produced unexplainable non-compliant results in a
fully-functional product?

RL

 

[John Woodgate]

If the variations have no signifigance, then perhaps it's because
there is already sufficient margin in the standard to cover variations
of this sort.

That's not the point. The people involved are concerned about the
variations they observe, irrespective of whether they actually result in
non-compliance.

As the standards are written in absolute current value limits, the
assumption that worst case occurs at full load might be incorrectly
adopted.

The problem is more significant for Class D products, where the limits
are proportional to active input power, because of the spectral
characteristics of the emissions.

We don't assume that worst-case is at full load; in fact for Class D
products, it usually isn't.

Loading tolerances might also be signifigant as product
inefficiencies are introduced, for economy's sake, after innitial
certification.

That has been taken into account as well.

Has the testing body or representative ever been called to task for a
report that is unrepeatable? I doubt it.

More than one testing body is involved. The associations representing
testing bodies have been informed about the problem, as have the
assessors in at least two countries. I have no doubt that when the
analysis is complete, some 're-training' will occur where necessary.

Has retesting ever produced unexplainable non-compliant results in a
fully-functional product?

Well, not non-compliant, as far as I know, if the standard is understood
correctly (which in some cases it wasn't), but variations in measured
value of 4:1 have been observed, which is not exactly reassuring unless
the reason is fully understood.

 

[Ken Smith]

I read in sci.electronics.design that Ken Smith <[email protected]>


Thanks for your information. IEC 61000-3-2 requires control of harmonics
up to the 40th, which means 2.4 kHz in sixty-cycle land. The problem I
am investigating is not due to restricted servo bandwidth; any problem
due to that is already covered by the test duration requirement.

I think you misunderstood my reasoning for putting that comment in. At
frequencies well below the servo's gain cross over, the input impedance of
the switcher section will look like a negitive resistance. This is
because the servo can turn up and down the pulse width fast enough. As you
get near the servo cut off, the variation in current becomes delayed and
above it, the variation decreases with increasing frequency.

Since the upper bound of frequency you are dealing with is well below that
point, the negitive resistance will look almost purely real. This should
simplify your calculations.

[..]

There are three causes of the problem, that we know about;

- component tolerances in the equipment; this is the responsibility of
the manufacturer but there is widespread lack of knowledge about it,
according to the information I have;

If we assume everyone is being smart, the makers will have allowed for the
tolerances and all the field units should be better than the makers
claim. (Or at least no worse)

Is this what you've seen so far?

Is there a spec for the temperature at which the test is to be
made? Inductors decrease at higher temperatures.

 

[Ken Smith]

value of 4:1 have been observed, which is not exactly reassuring unless

Holy cow. I think that kills the theory that the temperature of the unit
under test is to blame. I find it hard to believe that temperature
variations will be that big.

Are these products with removable power cords?

Where the power cords included in the testing?

I know that unit to unit and maker to maker variations in power cords can
be quite significant. Some have a fair twist in the wires and others
almost none.

 

[John Woodgate]

I read in sci.electronics.design that Ken Smith <[email protected]>

Holy cow. I think that kills the theory that the temperature of the unit
under test is to blame. I find it hard to believe that temperature
variations will be that big.

Are these products with removable power cords?

Mostly not, but it's not significant.

Where the power cords included in the testing?

Yes, you have to do that.

I know that unit to unit and maker to maker variations in power cords can
be quite significant. Some have a fair twist in the wires and others
almost none.

But that isn't going to make a difference at mains harmonic frequencies,
surely? The length matters in principle, because it affects the source
impedance seen by the rectifier, but not much.

 

[John Woodgate]

I read in sci.electronics.design that Ken Smith <[email protected]>

I think you misunderstood my reasoning for putting that comment in.

It seems so. Thanks for the clarification.

At
frequencies well below the servo's gain cross over, the input impedance of
the switcher section will look like a negitive resistance. This is
because the servo can turn up and down the pulse width fast enough. As you
get near the servo cut off, the variation in current becomes delayed and
above it, the variation decreases with increasing frequency.

Since the upper bound of frequency you are dealing with is well below that
point, the negitive resistance will look almost purely real. This should
simplify your calculations.

Indeed. In fact, I had assumed that, but your confirmation is helpful.

[..]

There are three causes of the problem, that we know about;

- component tolerances in the equipment; this is the responsibility of
the manufacturer but there is widespread lack of knowledge about it,
according to the information I have;

If we assume everyone is being smart, the makers will have allowed for the
tolerances and all the field units should be better than the makers
claim. (Or at least no worse)

Is this what you've seen so far?

No. But we haven't measured all the components in the tested units to
investigate the effects of tolerances 'in vivo'. We can do this by
simulation (electronic 'in vitro').

Is there a spec for the temperature at which the test is to be
made? Inductors decrease at higher temperatures.

Well, there is the usual permitted range of ambient temperatures, but
there are microclimates in an SMPS, some of which are pretty tropical.
Common-mode choke inductance may not be big enough to be significant,
but if there is a series inductor to widen the conduction angle its
inductance variation would be significant.

 

[R.Legg]

Has retesting ever produced unexplainable non-compliant results in a

Well, not non-compliant, as far as I know, if the standard is understood
correctly (which in some cases it wasn't), but variations in measured
value of 4:1 have been observed, which is not exactly reassuring unless
the reason is fully understood.

If this 4:1 variation occurs at harmonics higher than n=11, then I
don't find it very surprising. The occasional higher harmonic can
unexpectedly null in a unit, for purely serendipidous combinations of
components and load.

There isn't a corresponding unexpected peaking, as the stray is
undocumented series leakage inductance elements in CMFilter. This
depends on winding method, orientation, turns accuracy and just
sometimes just plain sloppiness. There is an upper limit achievable
without turning the choke into a recieving antenna at other
frequencies.

This stray leakage begins to look like 100mOhms or larger at the
+decade harmonic, and appears twice in every common-mode position (not
neccessarily with identical measurable values). Even low impedances
strongly influence the sharpness of the impulse input current
waveform, and the higher harmonics that this represents.

Your comments that an LISN is not required in the test procedure is a
little ingenuous. The networks were developed to allow predictable and
reproducible high frequency EMI performance. The fact that a similar
line source impedance is not specified in the testing of low frequency
harmonics, when 100mOhm impedances are capable of producing errors
(that may be of concern), is laughable. At the very least, the test
method should indicate that any LISNs present for any other test
purposes should be removed, if their LF impedance is likely to affect
producable harmonic levels.

RL