30W offline isolated flyback and use of shield windings.

[ocset]

We all know that with an offline 30-60W, isolated Flyback SMPS project, for the first prototype, the conducted EMC scan often shows a fail at the 28-30MHz point. As we know, this is down to common mode emissions.
(this fail is often there, even with meticulous PCB layout, and correct use of a Y capacitor across the isolation barrier.)

Often we initially try and get away with just one common mode choke, and no shield windings in the transformer. (also, we don’t like to snub our secondary diode too heavily, as this needs bigger snubber resistors. Also, we don’t like to increase the fet gate resistor too much, because although this conveniently reduces drain voltage dv/dt, it also increases switching losses)

So, in many cases, the transformer shield windings, or the second common mode choke with a higher SRF, is needed to clinch the conducted EMC pass and solve the 28-30MHz conducted problem.
You almost always find that people elect to solve it with shield windings in the transformer rather than with a second, smaller common mode choke. ..
Why is this?
..especially since offline flybacks of 30-60w are almost inevitably sandwich wound, so two shield windings are unfortunately needed in order to fully shield the primary from the secondary.

So why do people always elect for shield windings rather than a second common mode choke with a higher SRF?

(After all, shield windings can often mean having to use a bigger, more expensive transformer, in order to fit the shield windings on to the former.)

[T3sl4co1l]

The use of alleged certainty is amusing.

Fortunately, we all know why this is; how has it come to be that you need to ask?  Who neglected to tell out about shield windings and common mode chokes?

Tim

[MagicSmoker]

Don't assume everyone reading this is an expert on switchers, and the use of first person plural ("as we know," etc) is also annoying.

That said, switchers fail EMC testing at a variety of frequencies and for a variety of reasons and shield windings really shouldn't increase the size of the transformer so much that the next size up is needed. After all, good practice is to oversize the core a bit, anyway, to minimize flux swing and winding layer count, both of which serve to minimize losses (and temperature rise).

In contrast, employing a second common mode choke - which, btw, I have never seen in a commercial switcher - takes up more board space and usually ends up costing more than the shield winding, too.

Controlling the ringing across the secondary diode and switch during turn-off with properly sized RC dampers (rather than just winging it with a generic 10R + 100pF) helps more than anything else, though.

[Yansi]

That said, switchers fail EMC testing at a variety of frequencies and for a variety of reasons and shield windings really shouldn't increase the size of the transformer so much that the next size up is needed. After all, good practice is to oversize the core a bit, anyway, to minimize flux swing and winding layer count, both of which serve to minimize losses (and temperature rise).

In contrast, employing a second common mode choke - which, btw, I have never seen in a commercial switcher - takes up more board space and usually ends up costing more than the shield winding, too.

Controlling the ringing across the secondary diode and switch during turn-off with properly sized RC dampers (rather than just winging it with a generic 10R + 100pF) helps more than anything else, though.

But the oversized transformer, second common-mode choke and properly engineered big RC dampers cost money. Yo boss wont like that as tell you to cut corners, as is the typical mode of operation of most corpos.

They will probably tell you t otry downsizing the transformer a bit, so it barely works.

[MagicSmoker]

But the oversized transformer, second common-mode choke and properly engineered big RC dampers cost money. Yo boss wont like that as tell you to cut corners, as is the typical mode of operation of most corpos.

They will probably tell you t otry downsizing the transformer a bit, so it barely works.

I've been self-employed for most of the last 20+ years so if the pointy-haired boss cuts corners I have only myself to blame... 

[Yansi]

You are a happy person then. 

[ocset]

The attached is 70W offline Flyback SMPS. It has 2 common mode chokes in the AC section, as can be seen. The upstream one is 4mH  common mode  inductance, and of size 0.8mm diameter torroid, and the downstream one is 48mH  of common  mode inductance, and of diameter ~1.5mm torroid.
The main 70W flyback transformer does have a  copper tape shield, which is connected to DC Positive.
Primary side ground  is actually connected to the ferrite core (PQ3220) by way of a 3cm bare copper wire which is taped to the underside of the ferrite core.
The supply also features a 1W pilot  Flyback which has two  shield windings.
There are four small 1n Y caps bridging the isolation barrier….2 lots of two in series…I suspect the use of two small ones in series means less overall  stray inductance.
The primary FET is ceramic TO220, which I suppose means the extensive heatsink doesn’t inject so much noise everywhere since its not electrically connected to the drain of the fet (switching node)
Do you know why primary ground is connected to the  ferrite core? (of both transformers)

[chris_leyson]

I suspect the use of two small ones in series means less overall  stray inductance.

Nope, it's done for safety in case one cap fails short circuit.
EDIT: Good luck with your shield windings.

To be honest I haven't got aound to opening that can of worms yet, mainly for two reasons. Firstly, I haven't found many good papers on the subject, and secondly, there seem to be a lot of patents covering shield winding and construction techniques. And finally, we haven't really needed to use shields to get common mode down, well not yet anyway but it wouldn't hurt to try shield windings.
If anyone knows of any early literature on the subject I would love to hear. The thing that really scares me is having to go through potentially a lot of patents to make sure that you're not infringing any.   

[dmills]

Have a look at the literature around audio frequency transformers from the vacuum fet era, they often used an 'interwinding screen' and I **think** you will find some discussion in the period literature (Maybe start with one of the later editions of the "Radio designers Handbook" by Langford-Smith).

Regards, Dan.

[ocset]

Nope, it's done for safety in case one cap fails short circuit.

Thanks, thats what i always thought, (and still do to an extent) but i had some doubts after reading the final post of this...
https://www.edaboard.com/showthread.php?381543-Radial-Y-capacitors-made-with-too-small-pitch&p=1636346

Though in truth , it would surprise me, since the tolerance on Y caps wouldn’t be that bad as to make the voltage across each very unequal. But then again, even though a big and a small y cap might be rated to “250VAC”, I still suspect the  physically bigger one would have a greater voltage withstand than either of the two little ones.....not toghether i would admit.

I’m sure we all agree that shield windings do work to an extent . Though sometimes I wonder if they are put in to transformers  from Chinese manufacturers  in order to make the transformer more physically hard to wind….ie, increasing the manufacture cost…….i mean, it probably wouldn’t increase the manufacture cost in China, but in UK it would significantly do so…..so , in other words, shield windings pre-dispose a transformer to have to be made in China, rather than in UK….thus pre-disposing a UK customer to pay  the Chinese to do it rather than get it done domestically in UK. (for "UK" read NZ, Aus, USA, Europe, etc etc)

…ditto the “pigtail” ground_ferrite  connection I spoke of above?

[dmills]

Pretty sure Pace would wind me something with a screen if I asked them for not a lot more then winding one without a screen, same goes for most of the other custom magnetics vendors. One turn of copper tape, terminated at one end is not a big deal in the scheme of things.

Besides to use a screen or not is a DESIGN decision, you make it based on what the best way to get the performance you require is, I don't see a big conspiracy here, interwinding screens are just another tool in the designers toolkit.   

Custom magnetics are one of those things that get surprisingly cheap even at only a few hundred pieces, they are well within reach for small volume manufacturing.

73 Dan.

[ocset]

Thanks, presumably if you wind an offline, isolated  Flyback SMPS with a screen in it, and then you do two conducted EMC scans, one with the secondary fully isolated from the primary, and the other scan with a  small shorting wire shorting the primary and secondary together so its non-isolated…….then if the scans are the same, then there was no need to actually use the screen. Would you agree?
(since  in a non isolated Flyback SMPS, the capacitive coupling to the secondary doesnt matter too much because  the primary and secondary grounds are shorted together,  and  any noise coupled to the secondary will just get shorted back to the primary anyway)

[chris_leyson]

@dmills Thanks Dan, I was considering looking through the Radio Designers Handbook, it's within easy reach but it was getting late. Been going through some of the patents and tracing references back to "prior art" but to be honest I'm putting "shield windings" on the back burner for now as there are a lot of other subtle things that creep in like winding geometry when it comes to common mode currents.

There is an interesting tear down of an Apple iPhone charger on Ken Shirriff's Blog, http://www.righto.com/2012/05/apple-iphone-charger-teardown-quality.html. If you look at the transformer tear down the auxiliary winding is HUGE given that it only needs to supply a few tens of mW. One side of the auxiliary winding is "grounded" to the primary smoothing cap so it's a good example of a primary side shield winding.

[dmills]

Yep, transformers a subtle. Lots of fiddly details and they all matter at least some of the time.

Treez,
I would not try to extrapolate from a single design, the parasitics are just too much of a factor, and seriously, the experiments are not that hard to do. A few days with some bobbins, a turns counter, and a collection of suitable wire, tapes and core halves is an education well worth taking the time to acquire if you want to play in this space.

[T3sl4co1l]

Thanks, presumably if you wind an offline, isolated  Flyback SMPS with a screen in it, and then you do two conducted EMC scans, one with the secondary fully isolated from the primary, and the other scan with a  small shorting wire shorting the primary and secondary together so its non-isolated…….then if the scans are the same, then there was no need to actually use the screen. Would you agree?

I wouldn't.

The screen shunts the capacitance to the gradient of voltage along the primary winding.  Shorting the common grounds with a wire doesn't do quite the same thing.  The gradient gets induced onto the secondary, perhaps increasing the RF ripple (differential) aspect.  The ground wire size and length may not be the same, and the capacitances loading them down are not, either.

A non-isolated converter is a much easier thing in general, though.  I would not generally expect to need a shield in such a beast.  But a LinkSwitch (or equivalent), for example, is an even more compact solution than a flyback.

Tim

[ocset]

Thanks, but pertaining to Flybacks. we agree that a non isolated flyback  smps suffers less common mode noise (on its conducted EMC scan), then an isolated flyback?

(Albeit that yes, on occasions,  non isolated flyback  sometimes doesnt make sense)

[T3sl4co1l]

If it's wired correctly, then by definition there is no common mode!

In practice, the three-terminal device (in, gnd, out) is often expanded to a full two-port, with CM and DM, where CM is due to ground loop, or because the input is AC it's the DM plus or minus.

Tim

[ocset]

Thanks, as a finality on this theme,  i postulate that if we "de-isolate" a 60W offline flyback SMPS LED driver, by connecting its primary and secondary grounds together with a short wire piece...then that 60W flyback SMPS led driver gives essentially no more common mode noise problem than an offline  60W buckboost led driver, which uses no transformer and just uses a single winding inductor. Do you agree?

[T3sl4co1l]

No. The wire has impedance.  The buck-boost is hopefully over continuous ground plane.  Very different.

Tim

[ocset]

Thanks, OK,  but just for the sake of argument, lets now  say the flyback is also over continuous ground plane...then do you now say that their conducted EMC signatures are the same?

[T3sl4co1l]

The flyback probably isn't meeting its isolation or spacing requirement then.

You can't have both!

Tim

[ocset]

Thanks, yes i tend to agree...but i think i subtend from your reply, that in fact, the non isolated flyabck with contiguos ground plane, has much the same emc signature as the buckboost.....(ok a little worse as the flyback has two switching nodes...maybe i should have compared non isolated flyback with uncoupled inductor sepic.)

..Lets suppose I don’t have isolation and spaceing requirement with the  flyback, and just want it because I get LED load referenced to ground, and low side fet drive...isolation not needed.

[T3sl4co1l]

Thanks, yes i tend to agree...but i think i subtend from your reply, that in fact, the non isolated flyabck with contiguos ground plane, has much the same emc signature as the buckboost.....(ok a little worse as the flyback has two switching nodes...maybe i should have compared non isolated flyback with uncoupled inductor sepic.)

Actually, SEPIC with coupled inductors and no capacitor would be almost spot on.  The remaining difference is what ratio if any the flyback has.  Since, if you need a 1:1 ratio, you probably wouldn't choose flyback.

Otherwise, that's not a bad analogy.

..Lets suppose I don’t have isolation and spaceing requirement with the  flyback, and just want it because I get LED load referenced to ground, and low side fet drive...isolation not needed.

(Which, speaking of, you also get with SEPIC.  Again, assuming 1:1 ratio in both cases.)

So what?  Now that you've crafted a situation that is different from your premises, should it be surprising that its response also differs?

And this is still ignoring layout, and mechanical constraints, and so on.  The more you transform a circuit, the more those constraints will also change.  What are the chances that any particular embodiment of the isolated flyback topology, will have equivalent layout, mechanics and emissions to a SEPIC or buck-boost of some other particular embodiment?

You could create a series of converters, of each of the major topologies, crafted in such a way that their layouts and emissions are also equivalent.  But this would be a rather verbose way to create supplies for practical use.  Real practical circuits will differ on component choice and mechanical constraints, leading to very different layouts.  (In effect: the ideal layout, for a given set of components, connectivity and electrical and mechanical constraints, is highly nonlinear with respect to the topology, or the input and output ratings.)

Tim

[chris_leyson]

I have two flyback transformers, transformer A and transformer B. Transformer B is a clone of transformer A and has the same winding stackup and number of turns but the core and bobbin are slightly larger. Both are wound interleaved with a split primary. Transformer A passes conducted emissions but transformer B fails. Long story short, I needed a sample core and bobbin, transformer B style, so I resorted to taking one of the samples apart. Iransformer A, has single wire primaries wound on a single layer whereas transformer B uses two wires and is wound on two layers. Primary to secondary HF coupling is going to change because of the different winding geometry.

[Zero999]

I take it this is for street lighting? Are the EMC requirements more strict than for consumer equipment?

Does the secondary side really have to be SELV? If this is going to be used in a fully insulated luminaire, then I would've thought isolation is unnecessary?

I admit, I've never designed an offline SMPS before, since I always opted for off the shelf, because it normally works out more economical. Nowadays it's generally pointless for one to design their own offline SMPS. Any odd voltages can be generated using DC:DC converters on the secondary side.

I'm curious. I've never heard of a transformer screen, being referred to as a shield winding before. I'm guessing it's simply a piece of copper tape, with insulation at one end to avoid creating a shorted turn?

I've only seen screens on mains transformers before, but it doesn't surprise me they're used on some more high end switched mode supplies.

In an SMPS, is the shield connected to mains earth, or the rectified DC on the primary side?

I imagine, in order to satisfy the safety requirements and pass the hi-pot test, the secondary will need a lot of insulation between it and the secondary, if it's connected to the primary side. Can it be coupled via a suitable class Y capacitor so it doesn't have a direct connection to the mains and therefore less stringent insulation requirements?

Anyway, sorry for derailing the thread with noobie questions. To answer the question. I've had issues with an SMPS failing EMC before. It was a small Traco  DC:DC converter module, used in something to be installed in a military vehicle. From memory (this was about 15 years ago) it emitted in the high HF/low VHF range, similar to the OP's experience and it was resolved by adding a common mode choke, ferrite beads and some capacitors to the input side.