Transformers (rant)

[Paul Andrews]

Hopefully I'm not totally off base in what follows. That would be embarrassing!

I've been looking at building my own SMPS to drive my nixies from 5V or thereabouts. I think it would be hard to beat John Taylor's teeny power supplies on this front, but sometimes I might want to squeeze it into a different form factor, plus the completist in me just plain wants to do it.

It seems that using a flyback transformer would be a good way to go, but this is itself an example of exactly the sort of thing that has me all riled up! Now it is more than possible that I might be under some gross misunderstanding in what follows, but a transformer is a transformer. There is no such thing as a flyback transformer. You can use a transformer in a flyback design, but that doesn't make the transformer special. It is still just a transformer, and the same transformer could be used in a variety of other designs and applications.

However, the world seems to have decide that there are a wide variety of different types of transformer. This is a small sample: PoE, SMPS, Pulse, CFL, Power, Energy harvesting, Coupled inductors (this last one deserves its own place in hell, isn't that just another term for transformer?) etc.

Distributors and manufacturers split their offerings into these categories, making it especially difficult to actually find a transformer when you are trying to search on windings ratio!

Oh. And it turns out that a lot of these actually have a 1:1 windings ratio. Sigh.

But wait. it gets worse! Even assuming that you can actually find the windings ratio, the world has then decided that it is a good idea to talk about primary and secondary coils. Now it could be that there is some useful distinction between these two coils in a real world design, but as far as I know transformers are symmetrical devices. I can feed either coil, and use the other as the output, as long as I don't exceed the specs for each.

It gets even worse! It seems like the world of transformers is especially fluid. Parts come and go like leaves in a storm (o.k., there might be a better metaphor, but I'm on a rant here!).

At this point, I'm figuring that I'll just have to wind my own, which is possibly a step further than I had wanted to go! Anyway I've bought an LCR meter to help me out, and I'm poking around in the old parts from the boiler in my basement. 20kV anyone? I feel a digression coming on.

BTW I'm trying to keep some very impatient parrots off my iPad while I write this, so apologies for formatting errors, spelling errors and general incomprehensibilties.

[John Rehwinkel]

> It seems that using a flyback transformer would be a good way to go, but this is itself an example of exactly the sort of thing that has me all riled up! Now it is more than possible that I might be under some gross misunderstanding in what follows, but a transformer is a transformer. There is no such thing as a flyback transformer.

That's not entirely true. There are many parameters that affect how useful a transformer is for a specific application, and in particular, transformers designed for flyback duty include a gap in the magnetic path, to store energy to be released in the "flyback" phase when the magnetic field collapses. Transformers for some other uses are not gapped. Note that the gap itself is not always a physical gap in the core, and some powdered cores (there are several designs) have a discontinuous magnetic path on a small scale, which can act as a "distributed gap".

> You can use a transformer in a flyback design, but that doesn't make the transformer special. It is still just a transformer, and the same transformer could be used in a variety of other designs and applications.

Yes and no. Gapped transformers have different behaviour when it comes to magnetic saturation, and some designs depend on those differences to work well.

> However, the world seems to have decide that there are a wide variety of different types of transformer. This is a small sample: PoE, SMPS, Pulse, CFL, Power, Energy harvesting, Coupled inductors (this last one deserves its own place in hell, isn't that just another term for transformer?) etc.

Many of these are optimized for various parameters. All transformers are a compromise between a bewildering number of parameters (several loss mechanisms, leaking inductance, capacitance, high voltage standoff capability, etc.), and which choices are appropriate depend on how the transformer is going to be used.

> Distributors and manufacturers split their offerings into these categories, making it especially difficult to actually find a transformer when you are trying to search on windings ratio!

Quite. The design notes for the LT8304 point to some useful places to obtain transformers with a variety of ratios. For high-ratio transformers, I find that CCFL drivers are a reasonably common and affordable category. I sent a request to a manufacturer of mains transformers a while back, curious about their CRT power supply units, which are still listed in their catalogue, but their distributors didn't carry them. I was quoted nearly US$500 apiece for them.

> Oh. And it turns out that a lot of these actually have a 1:1 windings ratio. Sigh.

With a gapped transformer, and sometimes even without, you can get a large step-up with a 1:1 unit. There are plenty of geiger counter supplies out there built around 1:1 audio coupling transformers, deriving better than 400V from a 9V supply. I eyeballed those, but decided that since I'm aiming for much higher voltages (I'm looking to power things like lasers and CRTs that operate on many kilovolts and consume several watts), I'd explore other concepts.

> But wait. it gets worse! Even assuming that you can actually find the windings ratio, the world has then decided that it is a good idea to talk about primary and secondary coils. Now it could be that there is some useful distinction between these two coils in a real world design, but as far as I know transformers are symmetrical devices. I can feed either coil, and use the other as the output, as long as I don't exceed the specs for each.

Mostly true, but you can get some unexpected results, especially when operating at high frequencies (the regime where flyback circuits usually operate). I've seen some doozies in my time, one due to the fact that the outermost winding happened to be an efficient RF radiator.

> It gets even worse! It seems like the world of transformers is especially fluid. Parts come and go like leaves in a storm (o.k., there might be a better metaphor, but I'm on a rant here!).

Oh, absolutely. And most transformers are produced for a given buyer, and the specifications aren't available (although there are ways to derive them). A few builders here design (and even wind) their own transformers. They can probably offer much more concrete explanations than I can.

I do understand your frustration: most writing on the subject treats transformers in a fairly simple idealized manner as just two coupled inductances with a turns ratio, leads with pages filled with imposing equations, or just considers magnetics as "black magic". It took me a while to find an accessible primer on the subject, but I can recommend a reasonably compact and fairly readable reference that covers a lot of the basic details without getting bogged down in the physics and mathematics involved: The "Trilogy of Magnetics", by Würth Elektronik, a very nice book (in English) by a German transformer company (an earlier version was entitled something like "The ABCs of Transformers and Coils", and retailed for $14). Unfortunately, the current version is a bit dear at €49, and used copies are tricky to find.

> At this point, I'm figuring that I'll just have to wind my own, which is possibly a step further than I had wanted to go! Anyway I've bought an LCR meter to help me out, and I'm poking around in the old parts from the boiler in my basement. 20kV anyone? I feel a digression coming on.

Ah, you share my interest in high voltage! Digress away! At those voltages, I start eyeing the "LOPT" or "flyback" transformers used in CRT circuits and things like "neon sign" and "ignition" (both oil burner and automotive) transformers. That stuff is fun, the failures are spectacular, and safety is absolutely critical (and difficult: things that seem like insulators at lower voltages aren't really, most test equipment can't protect you, you can fill an entire room with a powerful RF field by mistake, and arcs can jump surprisingly far to get you). I've been working with high voltage for decades, and I've made every one of those mistakes.

- John

[David Forbes]

Paul,

I understand your frustration. Every device designed for mass production
has a custom transformer in it, as does every thing with a motor have a
custom motor.

When I decided to build a CRT clock in 2000, I was faced with the same
problem. So I sat down with a Morris coil winder and a bunch of cores
and bobbins that I scored in an estate sale, and a switching power
supply design book, and learned how they work.

I ended up designing my own power supply topology to run a CRT from a
single transformer. There are forward converters and flyback converters
and voltage multipliers. A voltage multiplier is both a forward
converter and a flyback converter, and if one wants to regulate several
supplies running from the same core, one needs to take both halves of
the cycle into account. So I used doublers on every secondary winding,
and achieved excellent regulation of both the LV and HV outputs.

Oops, was I supposed to patent that?

On 6/26/2017 4:40 AM, Paul Andrews wrote:
> Hopefully I'm not totally off base in what follows. That would be embarrassing!

--
David Forbes, Tucson AZ

[gregebert]

One thing that is confusing between linear and flyback supplies is that the transformer ratio does not really determine the output voltage of flyback converter. It's the feedback/control mechanism that ultimately sets the output voltage. Theoretically, a 1:1 transformer ratio could produce kilovolts from a 5V source, but you would never be able to build that with real components.

Isat (saturation current) is probably the most important transformer parameter; stay well below it otherwise you will have poor efficiency and excessive heating.

Make sure your driver transistor can handle the kickback that happens when it shuts off; higher turns-ratios will reduce this and higher leakage inductance will increase it.

Be sure to run a lot of spice simulations to get a feel for the design. And dont be too disappointed when your bench-test results, especially near full-load,  are not as good as your simulations.

Coilcraft has a variety of transformers with turns-ratios up to 1:100, but you wont need anything that high. 1:10 is a good starting point.

My nixie watch boosts 3.7V to about 150V for the display, using a coilcraft LPR6235. I exceeded Isat and it will overheat if left running for more than 20 seconds.

I painted myself into a corner due to size constraints, but it basically works OK for a wristwatch. Still running after 2 years on it's original battery charge and keeping accurate time.

[Paul Andrews]

Thanks guys. As usual, lots of useful feedback. I would like to ask more about a couple of the comments though:

As regards air gaps: My understanding is that they reduce the magnetic permeability and thus the inductance (while increasing Isat?). As such, I figure that for off-the-shelf parts that that should already be factored in to the specs, air gap or no? If I am winding my own, and I have separable cores, then I should definitely include that though?

When used in flyback power supplies, I assume that the output voltage still follows the winding ratio but that the input voltage is actually that of the back-emf induced by the field collapsing in the input coil? Is there something else happening?

What happens if the two coils wrap around cores of different permeability? Is the voltage ratio actually the square-root of the inductance ratio? i.e. if inductance ratio is 1:100, then voltage ratio is 1:10 regardless of the number of windings? A bit of a leap there: I assume the inductance varies with the square of the number of turns?

Also need to keep an eye on the isolation voltage - a few I looked at had very low isolation voltages, I assume that would not end well for my purposes?

That LPR6235 is a perfect example of a transformer I couldn't find! After you pointed it outAnd yes I tried to order samples, but they rumbled that I wasn't going to go into production and said it wasn't their policy to send samples out to people like me (I paraphrase, I am not trying to imply that they were nasty about it or anything). Understandable, but a little disappointing. I guess I'll just have to try and buy one from somewhere!

And yes, High voltages are fascinating :-)

[gregebert]

I just buy them from Coilcraft; their website nags me to order free samples but they gladly sell in single quantities. I usually buy a few other coils at the same time, for future projects.

As far as what voltage a flyback converter produces, I follow these lines of reasoning:

• When the primary is energized, it stores energy. Basic 1/2LI^2 . You can use volt-seconds if you wish.
• When the primary is de-energized, the inductor will spew all of that stored energy wherever it can. Think of it as a mechanical spring that pops loose.
• Your job as the designer is to harvest as much of that energy as possible
• The transformer, coupled-inductor, etc will produce any voltage necessary to get rid of it's energy, which includes exceeding the breakdown voltage of the driver transistor.
• Make sure all of the energy is removed from the inductor before you energize the primary again. Volts*seconds at the input must equal volts*seconds at the output.
• A flyback converter requires a feedback/control mechanism
• To reduce the output voltage, you need to reduce the energy-transfer. This can be done by reducing the duty-cycle at the input.
• Opposite for increasing the output voltage, but only up to a point. Dont go up to 90% duty-cycle, for example. Instead, increase the input voltage or lower the primary inductance. Beware that current increases.

With that in mind, start trolling for research papers on flyback designs; lots of good stuff out there to read and learn from.

[Dekatron42]

You can find a lot of good information with maths and circuits on Ronald Dekker's website http://www.dos4ever.com/flyback/flyback.html, also check out his 90V Anode battery.

/Martin

[Paul Andrews]

Of course I forgot one more question: I've used LTSpice, but I recently installed OrCAD PSpice because I need to input my own transistor models and it looks like it has good support for inputting models. That's as far as I have got because I really need to focus on some pieces of hardware other than power supplies (though I am itching to try and model some of the designs I have seen out there, so I can hopefully figure out some of the design choices).

Anyway, any particular spice recommendations?

@Martin, yes I had found that link. A very useful article. I am slowly poring over his web site :-).

[John Rehwinkel]

> As regards air gaps: My understanding is that they reduce the magnetic permeability and thus the inductance (while increasing Isat?). As such, I figure that for off-the-shelf parts that that should already be factored in to the specs, air gap or no?

Yes, but hopefully the specs will also show if the transformer core is gapped. Transformers for SET amplifiers and flyback mode power supplies, generally are, mains transformers and Royer converter transformers generally aren't.

> If I am winding my own, and I have separable cores, then I should definitely include that though?

If you're planning on using them in flyback mode, yes. The gap shouldn't be big, a fraction of a millimeter. E core transformers often have the outer arms full length and the center one slightly short to create the gap.

>
> When used in flyback power supplies, I assume that the output voltage still follows the winding ratio but that the input voltage is actually that of the back-emf induced by the field collapsing in the input coil? Is there something else happening?

Something else is indeed happening (as Greg mentions). The output voltage actually depends on the duty cycle as well as the winding ratio. This is why some of our nixie supplies are twitchy: in order to get a large voltage ratio, the duty cycle becomes somewhat extreme. As Greg also points out, you need to arrange the timing, voltage, inductance, and saturation so that you've put as much energy as you can into the magnetic field. Even more importantly, you need to remove all that energy too.

Remember, you can get a couple hundred volts from an ordinary buzzer running on a 1.5V cell. That's just one coil.

> What happens if the two coils wrap around cores of different permeability?

If they're different cores, I don't see how they can be magnetically coupled properly.

> Also need to keep an eye on the isolation voltage - a few I looked at had very low isolation voltages, I assume that would not end well for my purposes?

Yeah, that can be important, one of those many transformer specifications you have to check when working with high voltages.

> That LPR6235 is a perfect example of a transformer I couldn't find! After you pointed it outAnd yes I tried to order samples, but they rumbled that I wasn't going to go into production and said it wasn't their policy to send samples out to people like me (I paraphrase, I am not trying to imply that they were nasty about it or anything). Understandable, but a little disappointing. I guess I'll just have to try and buy one from somewhere!

They're not expensive. I didn't look up that one, but the FA2786-BL is four bucks apiece in quantity one (cheaper if you buy 400 at a time).

Then again, I don't buy a lot of CCFL transformers, as I keep finding nice CCFL inverters from surplus vendors for a buck apiece. I can't buy the transformers alone that cheaply, and these include the driving circuitry, output coupling capacitors, and printed circuit board. One common model even has a handy notch in the trace that joins the LV and HV sides, so I can easily cut it to separate them (I really would rather not have the ground return for my HV through the switching transistor in my regulator).

- John

[gregebert]

Regarding SPICE simulators, I use 2.

For initial design work, I use LTspice. Not just because it has an integrated GUI environment for schematic-capture & waveforms, but also because it has good models for the Linear Tech devices. My only gripe is that it's not available on Linux.

For PCB simulation, I use ngspice. I extract netlists from my PCB schematics in spice format so I can run analog simulations on everything non-digital, including my cabling. I have found and fixed several design errors on 6 projects over the years. None of those PCBs has an error or blue-wire; PCB is right-first-time. The one PCB I did not fully simulate has a stupid wiring error to the DCDC converters. Go figure.

BTW, digital simulations are done on everything, even transformers and nixie tubes. I use Modelsim. Yes, you can crudely model a transformer in a digital simulator.

[gregebert]

Here is a helpful link about air-gaps: http://info.ee.surrey.ac.uk/Workshop/advice/coils/gap/index.html

What isn't obvious is that an air-gap results in higher stored energy for a given core size (you can do the math from energy=0.5LI^2, and I=Vt/L) because current-squared is the dominant term, and lowering the inductance results in higher current (assuming voltage & time are fixed).

For a flyback converter, stored-energy is the name of the game.

Another, less-obvious, advantage to an air gap is that it reduces the variation of inductance due to variations in permeability of the core material.

This is very important for manufacturing/design quality control.