Transformer Design

[NickS]

One of the many holes in my electronics understanding is proper strategy for transformer design. It was never discussed in my schooling(I think we talked about how transformers work for 2 days). I wind my own inductors for filters but I have never attempted a transformer.

Lately I have been going back through my textbooks and it is sad how little they are discussed. I have noticed that there are a few members here that seem to be very well versed in transformer design and I would very much like to know more of what they know. So if you are willing, can we use this thread to work through the process of identifying the critical steps.

So to start a couple of basic questions
(1) for a given input frequency is there a sweet spot that you try to hit for your primary coil inductance?

(2) Obviously the gauge of wire used is important for current flow. How do you choose an appropriate gauge. Is it simply a matter of wire rating ohms/m and the length used? Somehow it seems there is more to it.

(3) Saturation how do you plan for a certain primary coil saturation headroom. I think this is determined mostly by the core. But what is the process for choosing a proper core.

Or perhaps I am approaching this all wrong. I would love to see the steps you take to specify and design a new transformer. I am not really looking for the LMGTFY kind of answers here. I would rather converse with someone who does this sort of thing(if they would be so kind).

Thanks in advance
Nick

 

[shrtrnd]

I wouldn't even attempt it these days. More power to you if you're going to do it.
With the advent of different wire composition and construction, iron core composition, and a dozen other variables, I leave it to the transformer companies to figure out.
I've got five books on transformer construction (Pre-1980). You're going to have to talk
to somebody who actually works in transformer design, to get a responsible answer.
(And we didn't even discuss your input to the transformer). I know it's no help, but I wanted to give you my two-cents worth, ...before you invest the time.

 

[NickS]

That has been my attitude about it too. But there are special situations that warrant a custom design because nothing off the shelf suits your need. So I am not saying that I necessarily wind the copper myself. But I need to be able to specify how it is to be made to a shop that does.

Here is an example. I worked on an electro-surgical knife that had a 2 custom transformers(one for small signal isolation and another for the power amp). The isolation one was simple no need to waste the time on a custom part. But the one for the PA was a legitimate piece of engineering with its very specific turns ratio, its use of litz wire to handle the high power being pushed and its catch winding. We did not revisit the making of the PA transformer during the re-design because it was a known working part and we lacked the expertise to know if it needed to be changed. This is the sort of thing that I want to have a better handle on.

I am not suggesting that this thread be an all inclusive resource, but perhaps it can layout the appropriate steps.

 

[shrtrnd]

I don't know who makes up-to-date reference material on transformers. You might actually email a transformer company for suggestions on where to look for reference materials. Especially for critical design. I've used transformers that vendors claimed were equivalent to what I needed, only to have eddy current hum screw me from a
poor lamination job. Sales people at companies are useless, but sometime you find a telephone answerer, who can connect you with somebody in their company who can be really helpful. Considering your interest, I'd find me a transformer fanatic in a company that makes them. Some guys LOVE to impress you with their expertise.
I salute you for attempting this quest, ...most people like me, just give up.

 

[Militoy]

(1) for a given input frequency is there a sweet spot that you try to hit for your primary coil inductance?

For most power transformers, a specific inductance isn’t the actual goal – but operating within an optimum range of core flux density. It seems like it would work out to the same thing – but there is actually a very important difference. A very small difference in core gap can show up as a significant difference in open-circuit inductance – but will typically be an insignificant factor in full-load operation.

(2) Obviously the gauge of wire used is important for current flow. How do you choose an appropriate gauge. Is it simply a matter of wire rating ohms/m and the length used? Somehow it seems there is more to it.

Total copper loss and resulting load regulation and temperature rise are the main factors that determine the optimum wire gauge for a specific design. “Rules of thumb” are handy starting points for a less experienced designer – but can limit flexibility in specific design cases. A common guideline quoted for line-frequency (50-400Hz) transformers from 5-1000 VA is to start with 500 circular mils/amp. So a 2A secondary would be wound with 20 AWG magnet wire. It’s important to know though, when to throw out the rules. At higher frequencies, AC copper losses increase due to skin effect and proximity effect. Then, conductor shape and structure, and winding geometry become significant factors.

(3) Saturation how do you plan for a certain primary coil saturation headroom. I think this is determined mostly by the core. But what is the process for choosing a proper core.

Each application will have a hierarchy of factors that dictate all the elements of a design. In military work for instance, size and weight, temperature range and other environmental stress factors will far outweigh cost considerations. In automotive and consumer applications, where millions of units are common production numbers, competitors will cut each other’s throats to knock 5 cents off the cost of the core. Within the scope of those initial considerations, once a core type is selected, flux density is typically set up to avoid saturation during any anticipated condition of normal or abnormal operation – and to meet the other goals of the design within the design goal hierarchy. Sometimes all-day efficiency is the top goal; sometimes ability to withstand occasional or regular overload; sometimes MTBF (mean time between failure) is paramount; sometimes limiting inrush current. Every design has to be considered separately.

There really isn’t a simple pat answer to any of your very thoughtful questions. Magnetics design is viewed as a kind of “black art”. That’s not because there’s any magic to it – it’s just pure and simple physics. But there are a whole lot of variables to consider, and a whole lot of choices to make in any given design.

 

[NickS]

Thanks guys I think we are making some progress here. Militoy please don't take offense to the fact that I would like to crawl inside your head with a gray matter xerox machine. It sounds like you are the transformer guru I was looking for and I hope I can continue to draw on your knowledge.

Perhaps we could go through an arbitrary design to sort of quantify some of the aspects we have already discussed loosely. I will do some homework tonight and come up with a set of parameters(out comes the physics book).

 

[Mitchekj]

Good stuff Militoy. I haven't done any in depth transformer design, but I've cobbled a couple together when I needed a flyback SMPS transformer in a hurry for some testing.

Correct me if I'm way off here, I don't really have any magnetics gurus around work to ask, but I had always chosen the primary inductance to get me a constant of about 0.4 for the ripple factor, based on frequency, DC bus voltage, max input power, and duty cycle max. That's for continuous conduction mode, anyhow. There's a formula somewhere, (I use an excel calculator,) is that sounding right?

For wire gauge, there are yet more forumlas, but try to stay under 1mm diameter (eddy currents, if I recall) and just go with a gauge to keep a good headroom on current capacity, based on peak primary current. Number of turns and turns ratio all depend on voltages required, duty cycle max, current limit level, primary inductance, and some others which I can't rightly recall at the moment. An aux coil (for Vcc generation) is also calculated here.

Saturation is all about the core, as I understand it. I try to get a core rated for double the peak current I expect, to avoid saturation. Saturation will destroy a flyback without prejudice, so I make sure to go overboard when feasible. There are many types of cores, I don't have enough experience to say why you would use an EE22 PC40 vice any of the others, other than saturation rating / gap / material etc. Bobbin type and size affect things too.

May be way off base, but I've wound workable transformers for prototype work. It's a pain in the arse, and I try to avoid doing it at all costs. For production stuff, I'd wager it's a whole new ballgame... I know there are all kinds of UL/CSA/etc. requirements on insulation, wires, bobbin, yada yada. And for EMI supression I've seen all manner of copper layers wound in and/or around the transformer.

That's just my amatuer take on transformers, and I'm not even sure if it would apply to an actual transformer (as a flyback xfmr is more of a coupled inductor.) Actually, I'm pretty sure most of the steps involved in making your inductors apply equally as much to transformers.

Please give me any feedback if you wouldn't mind, magnetics gurus.

 

[Militoy]

Good stuff Militoy. I haven't done any in depth transformer design, but I've cobbled a couple together when I needed a flyback SMPS transformer in a hurry for some testing.

Correct me if I'm way off here, I don't really have any magnetics gurus around work to ask, but I had always chosen the primary inductance to get me a constant of about 0.4 for the ripple factor, based on frequency, DC bus voltage, max input power, and duty cycle max. That's for continuous conduction mode, anyhow. There's a formula somewhere, (I use an excel calculator,) is that sounding right?

For wire gauge, there are yet more forumlas, but try to stay under 1mm diameter (eddy currents, if I recall) and just go with a gauge to keep a good headroom on current capacity, based on peak primary current. Number of turns and turns ratio all depend on voltages required, duty cycle max, current limit level, primary inductance, and some others which I can't rightly recall at the moment. An aux coil (for Vcc generation) is also calculated here.

Saturation is all about the core, as I understand it. I try to get a core rated for double the peak current I expect, to avoid saturation. Saturation will destroy a flyback without prejudice, so I make sure to go overboard when feasible. There are many types of cores, I don't have enough experience to say why you would use an EE22 PC40 vice any of the others, other than saturation rating / gap / material etc. Bobbin type and size affect things too.

May be way off base, but I've wound workable transformers for prototype work. It's a pain in the arse, and I try to avoid doing it at all costs. For production stuff, I'd wager it's a whole new ballgame... I know there are all kinds of UL/CSA/etc. requirements on insulation, wires, bobbin, yada yada. And for EMI supression I've seen all manner of copper layers wound in and/or around the transformer.

That's just my amatuer take on transformers, and I'm not even sure if it would apply to an actual transformer (as a flyback xfmr is more of a coupled inductor.) Actually, I'm pretty sure most of the steps involved in making your inductors apply equally as much to transformers.

Please give me any feedback if you wouldn't mind, magnetics gurus.

It sounds as if you have done quite a bit of homework - nothing I read above strikes me as off base. Mose of the ferrite core companies have a lot of good ap notes for different topologies up on their websites - and that's a good place to start. Mag Inc used to have a core size/shape vs frequency/topology/power in their catalog. Don't know if they still have that up, but it was a nice little cheat sheet to get into the ballpark with. I have some familiarity with UL/CSA/VDE specs - but since I've been buried in military work for the past 30 years, I haven't had to comply directly with consumer electronics standards very often - and I won't attempt to comment intelligently on those.

It's refreshing to read of a company designing an SMPS power supply into their own product, as opposed to just dropping in an off-the-shelf module. That seems to be more unusual these days.

 

[JuliaKhanam]

It is intuitive that if 1 cc of core material produces a certain power loss, then 2 cc of that material will produce twice as much.

So we are dealing with the Volume of the core.

The core volume is proportional to the cube of any of its linear dimensions.







Last edited by JuliaKhanam; 08-29-2010 at 04:45 AM..

 

[Militoy]

It is intuitive that if 1 cc of core material produces a certain power loss, then 2 cc of that material will produce twice as much.

So we are dealing with the Volume of the core.

The core volume is proportional to the cube of any of its linear dimensions.

While the above might seem intuitive - it's not quite true. A core with twice the cross sectional area in the magnetic path will support twice the core flux - but as the core volume increases as the cubed product of the dimensions, the radiating surface area of the core only increases as the squared product of those dimensions. This means that as transformer size increases in volume (and power rating), it becomes increasingly more difficult for it to shed heat. That's one reason why pole-mounted distribution transformers are filled with oil - and often fitted with radiator fins and oil circulation tubes. Engineers take every possible measure to help them get rid of heat, without making them unnecessarily large.

 

[Mitchekj]

It's refreshing to read of a company designing an SMPS power supply into their own product, as opposed to just dropping in an off-the-shelf module. That seems to be more unusual these days.

Thanks for the comments. In the LED lighting field, the drivers which are suitable are too pricey to spec into our products and be competitive. Not to mention the total lack of design control, and unknown (sometimes downright crap) quality. It's just better for us to produce our own drivers. The only major headaches are all the regulatory requirements, either law (FCC) or market driven (UL, EnergyStar.) Said requirements tend to shoot the price nice and high when you ask a vendor for a supply that has to do "everything!"

 

[NickS]

Militoy,
Thanks for the Mag-inc plug. I have 15pdf app notes open on my desktop and I am trying to digest it all.

Mitch,
So for your SMPS's do you typically design your own magnetics or do you try to find a COTS solution and design around it. I would think you would keep the cost down by using COTS coils but then again if you quantities are high enough then custom coils could edge out.

 

[Mitchekj]

For most of the SMPS designs we do, 9/10 times there are no off the shelf transformers that would meet everything we need. Generally we just give the specs to one of our magnetics vendors and they give us a quote. They do all the in-depth engineering work, or they may have a "custom" transformer already in production that would work for our application as well.

Wurth Electronics is great for this kind of custom thing.

 

[Militoy]

I haven't heard of Wurth before - though we've used several other offshore companies to build lower $ magnetics that it didn't make sense for us to wind ourselves.

Roger on the "major headaches" dealing with regulatory agencies. I built a batch of parts for a medical product around 10 years ago, and had to get the design listed to UL544. The process was even more painful than putting together a full technical provisioning package for the US Navy (usually around as thick as a phone book!). After delivery, we were visited and billed quarterly for the following 2 years by UL - just to verify that we hadn't had any additional production activity on that product. No, gracias.

 

[NickS]

Interesting. Can you elaborate on what specs you furnish. I would assume you are giving them Input Voltage and frequency and the Output voltage and current. Do you also tell them what core type and wire you want used or do you leave that up to their discretion? I would assume to meet commercial standards you would have to call out your maximum core loss which will inevitably impact your core size and cost. Finally what is a typical NRE you expect to pay for such a transformer?

 

[Mitchekj]

If I recall, they require Vin range, Vout, Iout (Pout), switching frequency, number of secondary/aux windings (and associated Vout/Iout), size requirements, topology, conduction mode, and I think a couple of other specs. Any specific UL requirements, max temp rise, etc. The guys we've worked with have been real good at asking questions to make sure we get what works best for us.

They will generally do the bobbin/core/wire, turns, etc. And yeah, efficiency vs. cost vs. size is always a fun teeter-totter. They will usually give a list of options and you can pick what works best. I've not dealt with them first hand, but have observed the process a bit. Seems pretty painless.

The NRE is in the thousands as I recall, but the exact number I've not been in contact with. I believe there's only an NRE if you want design rights (with drawings, etc.) so you can get the thing manufactured wherever you wish. I'm not sure that there is an NRE if you buy directly from them, but that would make sense to me if there weren't. Sorry I can't be more specific, as I'm not the guy dealing with those aspects. I'm just the tech, creeping my way into design.

Edit: And yeah, UL just loves to charge you a couple thousand bucks to change a letter on your file, too. Weren't they supposed to be non-profit?

 

[(*steve*)]

Roger on the "major headaches" dealing with regulatory agencies. I built a batch of parts for a medical product around 10 years ago, and had to get the design listed to UL544.

I feel your pain. I work for a company producing medical software.

We need to be able to document everything to a level where the regulator can determine who farted in which meeting, and what action was pre-planned, and subsequently taken on each occurrence.

 

[NickS]

I am still reading datasheets on core properties. In the mean time lets discuss minimum size transformers. For the sake of example lets say I have a core sitting around with the following properties
OD = 7mm
ID = 3 mm
H = 5 mm
perm = 125
100 nH/T^2
Our basic electronics training says as long as the turns ration is good we have the desired transformer. But we know that there is so much more to it than just that. So lets say I need 120/12V conversion at mains 60Hz(at whatever current we can get the more the better) and I quickly decide to wrap 20 turns of 26awg mag wire for the primary and I have just enough room left for 2 turns on the secondary. My turns ratio is 10:1 so the correct voltage should be there but the design is admittedly bad.

Can you help me quantify why it will fail to preform.

Is the primary/secondary inductance a problem?
Do I have too much leakage with only two windings on the secondary?
Where will it saturate the core?
Will it even operate at 60Hz?
And whatever else are glaring problems.

What are the limiting factors

 

[Militoy]

I am still reading datasheets on core properties. In the mean time lets discuss minimum size transformers. For the sake of example lets say I have a core sitting around with the following properties
OD = 7mm
ID = 3 mm
H = 5 mm
perm = 125
100 nH/T^2
Our basic electronics training says as long as the turns ration is good we have the desired transformer. But we know that there is so much more to it than just that. So lets say I need 120/12V conversion at mains 60Hz(at whatever current we can get the more the better) and I quickly decide to wrap 20 turns of 26awg mag wire for the primary and I have just enough room left for 2 turns on the secondary. My turns ratio is 10:1 so the correct voltage should be there but the design is admittedly bad.

Can you help me quantify why it will fail to preform.

Is the primary/secondary inductance a problem?
Do I have too much leakage with only two windings on the secondary?
Where will it saturate the core?
Will it even operate at 60Hz?
And whatever else are glaring problems.

What are the limiting factors

Interesting question - but the theoretical design you are suggesting bumps hard into the laws of physics on multiple sides. First of all - the toroidal core you describe has a mu of 125, and an Al of 100 - identifying it as an MPP, HiFlux or iron powder core - in other words, a core with a distributed gap, intended to support DC flux. This kind of core is designed to be used as a DC reactor - or other inductor or transformer with significant DC current in the windings. Typical operating flux density is on the order of 2-3 Kg (0.2 -0.3T). As a side note - the mu (permeability) of a 50/60 Hz core is typically between 30,000 and 50,000, and the saturation flux density on the order of 17.5 Kg (1.75 T). The cross section of the core described converts (in inches) to about 0.015 IN^2. This core would need about 250,000 turns on the primary, to keep the flux down below saturation. You can imagine, with a 0.118 core I.D., the wire would need to be unusually small - unbelievably small - to fit through the aperature of the core.

If you would like to discuss real-world minimum-size 60 Hz transformers - there are several legitimate tradeoffs that can be made. But I'm afraid the above concept would have to be classified as "vaporware".

 

[NickS]

Fantastic that is just the sort of reply I was hoping for(I admittedly layed out what I thought to be a bad design to spur some technical discussion). So going back to what you said about the core, I used the mag-inc 0077270A7 as a rough model for my previous post.

(1)You called it a distributed gap I am having some difficulty finding a good definition for that. I know its not a physical gap but what exactly is the gap in reference too.

(2) You say the typical perm of a 50/60Hz core is around 30k-50k with sat flux density around 17.5k gauss. Are these parameters driven by peak voltage, frequency or power? And how much flexibility exists in this(with trade offs).

(3) Cross sectional area of the core. You must be working off some sort of calculator or formulas to know how many windings(250000) it would take to keep the flux below saturation for the given core. I think that tool would be very useful here. Would you mind sharing how you came to that number?

Thanks again Militoy

-Nick-