A method for manufacturing a planar transformer based on a multilayer printed circuit board. Payton Planar Transformers and Chokes (2005)

In the previous article, the advantages of using planar transformers in small and mobile devices were considered. The characteristics of ferrite cores used to design planar transformers were also given. This publication proposes a method for calculating planar transformers for pulse forward and reverse converters.

Introduction

Planar transformers can be built as plug-in components, as a single-layer PCB assembly or as a small multi-layer PCB, or embedded in a multi-layer power supply PCB.

Important advantages of planar magnetic components are:

• very small sizes;
• excellent temperature characteristics;
• low leakage inductance;
• excellent repeatability of properties.

Measurements of the operating parameters of planar transformers with E-shaped cores and windings made on the basis of a multilayer printed circuit board show that the thermal resistance of these devices is significantly (up to 50%) lower compared to conventional wire-wound transformers with the same effective core volume V e . This is due to the higher core surface area to core volume ratio. Thus, with increased cooling capacity, planar transformers are able to handle higher throughput power densities while keeping temperature rise within acceptable limits.

This brochure describes a quick and easy method for designing planar power transformers and provides examples of devices designed using this method.

The run test results show that the measured temperature rise is in good agreement with the calculated data.

Rice. 1. Planar transformer disassembled

Rice. 2. Design options for planar transformers

Calculation procedure

Determination of the maximum magnetic induction

Losses in the core and copper conductor during the operation of the transformer lead to an increase in temperature. The magnitude of this increase must not exceed the allowable limit in order to avoid damage to the transformer or the rest of the circuit. At thermal equilibrium, the value of the total losses in the transformer Ptrafo is related to the increase in the temperature of the transformer D T by a relation similar to Ohm's law:

where R T is the temperature resistance of the transformer. In fact, P trafo can be thought of as the cooling capacity of a transformer.

An empirical formula can be established that directly relates the value of the thermal resistance of a transformer to the effective magnetic volume V e of the ferrite core used. This empirical formula is valid for wirewound transformers with RM and ETD cores. A similar relationship has now been found for planar transformers with W-shaped cores.

Using this relation, it is possible to estimate the temperature rise of the transformer as a function of the magnetic induction in the core. Due to the limited available winding space for planar magnetic components, it is recommended to use the highest possible values ​​of magnetic induction.

Assuming that half of the total transformer losses are core losses, the maximum core loss density P core can be expressed as a function of the transformer's allowed temperature rise as follows:

The power loss in our ferrites was measured as a function of frequency (f, Hz), peak magnetic induction (B, T) and temperature (T, °C). The core loss density can be approximately calculated using the following formula:

Here C m , x, y, c t0 , ct 1 and ct 2 are the parameters found by fitting the empirical loss curve. These parameters are specific to a particular material. Their dimensions are chosen so that at a temperature of 100 °C the value of CT is equal to 1.

Table 1 shows the values ​​of the parameters listed above for several grades of Ferroxcube high power ferrites.

Table 1. Approximation Parameters for Calculating Core Loss Density

Ferrite grade	f, kHz	cm	x	y	ct2	ct 1	ct0
3C30	20–100	7.13x10 -3	1,42	3,02	3.65x10 -4	6.65x10 -2	4
	100–200	7.13x10 -3	1,42	3,02	4x10 -4	6.8x10 -2	3,8
3C90	20–200	3.2x10 -3	1,46	2,75	1.65x10 -4	3.1x10 -2	2,45
3C94	20–200	2.37x10 -3	1,46	2,75	1.65x10 -4	3.1x10 -2	2,45
	200–400	2x10 -9	2,6	2,75	1.65x10 -4	3.1x10 -2	2,45
3F3	100-300	0.25x10 -3	1,63	2,45	0.79x10 -4	1.05x10 -2	1,26
	300-500	2x10 -5	1,8	2,5	0.77x10 -4	1.05x10 -2	1,28
	500-1000	3.6x10 -9	2,4	2,25	0.67x10 -4	0.81x10 -2	1,14
3F4	500-1000	12x10 -4	1,75	2,9	0.95x10 -4	1.1x10 -2	1,15
	1000-3000	1.1x10 -11	2,8	2,4	0.34x10 -4	0.01x10 -2	0,67

The maximum allowable value of Pcore is calculated by formula (2). This value is then substituted into equation (3). Now you can calculate the maximum allowable magnetic induction Bpeak by rewriting equation (3) as follows:

Note: the maximum allowable value of B can be found in another way - by writing a computer program that calculates the power loss for an arbitrary waveform using formula (3) for given values ​​of the approximation parameters . The advantage of this approach is that it allows you to calculate the losses taking into account the actual mode of oscillations B, as well as to select the optimal grade of ferrite for a particular case.

Having determined the maximum allowable peak magnetic induction, it is possible to calculate the number of turns of the primary and secondary windings using known formulas, including the topology of the converter and the type of transformer (for example, reverse and forward).

It is necessary to make a decision on how the windings will be distributed between the existing layers. The currents flowing in the tracks will cause the PCB to rise in temperature. For reasons of heat dissipation, it is recommended to distribute the windings in the outer layers symmetrically with respect to the windings in the inner layers.

Rice. 3. B peak in the formulas is equal to half the amplitude of the induction fluctuations in the core

From the point of view of magnetism, the best option would be to intersperse the primary and secondary layers. This will reduce the so-called proximity effect (see page 4). However, the low winding height in planar design and the number of turns required for a particular application do not always allow choosing the optimal design.

From a cost point of view, it is recommended to choose printed circuit boards with a standard copper layer thickness. Common thicknesses used by PCB manufacturers are 35 and 70 microns. The temperature increase in the winding, induced by the flowing currents, essentially depends on the thickness of the copper layers.

Safety standards such as IEC 950 require a distance of 400 µm in the PCB material (FR2 or FR4) to ensure secondary decoupling from the mains. If decoupling from the mains is not required, a distance of 200 µm between winding layers is sufficient. In addition, it is also necessary to take into account the layer for the stencil - 50 microns on both sides of the board.

The width of the tracks that form the windings is determined based on the magnitude of the current and the maximum allowable current density. The distance between the turns depends on the capabilities and budget of the production. There is a rule of thumb: for tracks 35 µm thick, the width of the tracks and the distance between them should be more than 150 µm, and for tracks 70 µm thick, more than 200 µm.

Depending on the PCB manufacturer's manufacturing capability, the dimensions may be smaller, but this will most likely result in a significant increase in the cost of the PCB. The number of turns in one layer and the distance between the turns are denoted by Nl and s, respectively. Then, with the available winding width bw, the track width wt can be calculated using the following formula (see Fig. 4):

Rice. 4. Track width wt, track distance s and winding width b w

If decoupling from the mains is required, the situation changes somewhat. The core is considered as part of the primary winding circuit and must be separated by 400 µm from the secondary circuit. Therefore, the creepage distance between the secondary windings close to the left and right sides of the core and the core itself must be 400 µm. In this case, the track width should be calculated using formula (6), since 800 µm must be subtracted from the available winding width:

In formulas (5) and (6) all dimensions are given in mm.

Determination of PCB Temperature Rise Caused by Flowing Currents

The last step to be taken is to determine the temperature rise in the copper traces caused by the currents flowing. To do this, it is necessary to calculate the effective (rms) values ​​of the currents, based on the input data and the desired output parameters. The calculation method depends on the topology used.

The example section shows calculations for standard forward and reverse converter technology. An example of the relationship between temperature rise and effective current values ​​for different cross-sectional areas of printed circuit board conductors is shown in fig. 5. In cases where there is a single conductor, or where the inductances are not too closely spaced, the width, thickness, and cross-sectional area of ​​the conductor can be directly determined from this diagram, as well as the maximum allowable currents for various given temperature rises.

Rice. 5. Relationship between current, PCB trace sizes and temperature rise

The disadvantage of this design method is the assumption that the heat generated in the winding is caused by DC current flowing, when in reality there is AC current causing skin and proximity effects.

The skin effect is due to the presence in the conductor of a magnetic field created by the current that flows in this conductor itself. A rapid change in current (at high frequency) induces a variable induction which induces eddy currents. These eddy currents, which contribute to the main current, are in the opposite direction to it. The current vanishes at the center of the conductor and moves towards the surface. The current density decreases exponentially from the surface to the center.

The depth of the surface layer d is the distance from the surface of the conductor in the direction of its center, at which the current density decreases by a factor of e. The depth of the skin layer depends on material properties such as electrical conductivity and magnetic permeability, and is inversely proportional to the square root of the frequency. For copper at 60 °C, the skin depth can be approximated by the following formula:

If a conductor with a thickness w t less than 2d is taken, the contribution of this effect will be limited. This gives a track width of less than 200 µm for a frequency of 500 kHz. If a large winding width is available for the required number of turns, the best magnetic solution is to divide them into parallel tracks.

In real situations, eddy currents will be present in the conductors, caused not only by the changing magnetic field of their own current (skin effect), but also by the fields of other conductors located nearby. This effect is called the proximity effect. If the primary and secondary layers alternate, the influence of this effect is much less. The fact is that the currents in the primary and secondary windings flow in opposite directions, so that their magnetic fields cancel each other out. However, adjacent conductors of the same layer will still contribute some amount to the proximity effect.

Empirical results

Temperature measurements in several types of printed circuit board designs with alternating currents flowing in the windings show with acceptable accuracy that, at frequencies up to 1 MHz, each increase in frequency by 100 kHz gives an increase in the temperature of the printed circuit board by 2 °C more than the values ​​determined for the case direct currents.

The goal is to design a horizontal transformer with the parameters given in the table.

As a first step, an assumption is made that at a given frequency, a large value of the peak magnetic induction can be taken - 160 mT. Later we will check if this is possible for given values ​​of core loss and temperature rise.

Example 1 Flyback transformer

Table 2 shows the calculated number of turns for the six smallest standard combinations of planar E-cores and Ferroxcube inserts. In addition, the values ​​of the self-inductance of the primary winding, the width of the air gap and the currents calculated according to the formulas from inset 1 are given.

Table 2. Calculation of design parameters of several line transformers

Core	Ae, mm 2	Ve, mm 3	N1	N2	NIC	G, µm	Other calculated parameters
E-PLT14	14,5	240	63	7,4	7,2	113	L prim = 638 uH
E-E14	14,3	300	63	7,4	7,2	113	I p (rms) = 186 mA
E-PLT18	39,5	800	23	2,7	2,6	41	I o (rms) = 1593 mA
E-E18	39,5	960	23	2,7	2,6	41
E-PLT22	78,5	2040	12	1,4	1,4	22
E-E22	78,5	2550	12	1,4	1,4	22

From Table 2, it can be seen that the required number of primary windings for the E-E14 and E-PLT14 core sets is too high for the winding to be made on the basis of a multilayer printed circuit board. Therefore, combinations of E-E18 and E-PLT18 cores look like the best option. Rounding the calculation results N1, N2 and NIC gives the numbers 24, 3 and 3 respectively.

To determine the losses in the case of a unipolar triangular induction wave with a frequency of 120 kHz, a peak induction of 160 mT and an operating temperature of 95 °C, a program based on expression (3) was used. For high-power ferrites 3C30 and 3C90, the expected core losses are 385 mW/cm3 and 430 mW/cm3, respectively.

The allowable loss density at D T=35°C is 470 mW/cm3 for E-PLT18 and 429 mW/cm3 for E-E18 (from expression (1)).

The conclusion is that 3C30 and 3C30 ferrites can be used in both core combinations. Lower quality ferrites with high power losses will lead to too much temperature rise.

The 24 turns of the primary can be distributed symmetrically over 2 or 4 layers. The available winding width for E-18 cores is 4.6 mm. This shows that the option with two layers of 12 turns each will be difficult to implement, and therefore expensive. This will require the use of very narrow tracks with a very small step. Therefore, a variant with four layers, 6 turns in each, is selected. Fewer layers in a multilayer PCB will result in lower cost. Therefore, we will provide 3 more turns of the primary winding (for IC voltage) and 3 turns of the secondary winding, and one layer for each of them. Thus, it is possible to construct a structure with six layers, as shown in Table 3.

Table 3. An example of a transformer design with six layers

Layer	Number of turns	35 µm	70 µm
stencil		50 µm	50 µm
primary	6	35 µm	70 µm
insulation		200 µm	200 µm
primary	6	35 µm	70 µm
insulation		200 µm	200 µm
primary IC	3	35 µm	70 µm
insulation		400 µm	400 µm
secondary	3	35 µm	70 µm
insulation		400 µm	400 µm
primary	6	35 µm	70 µm
insulation		200 µm	200 µm
primary	6	35 µm	70 µm
stencil		50 µm	50 µm
TOTAL		1710 µm	1920 µm

Depending on the amount of heat generated by the flowing currents, you can choose between 35 µm or 70 µm thick copper tracks. A distance of 400 µm is required between the layers of the primary and secondary windings to ensure decoupling from the mains. The E-PLT18 combination has a minimum winding window of 1.8mm. This is sufficient for a track thickness of 35 µm, which gives a total PCB thickness of about 1710 µm.

To reduce the cost of construction, we chose the distance between tracks equal to 300 μm. Calculating the track width of the secondary winding using formula (5) gives a result of 1.06 mm, including decoupling from the mains.

Using the diagram in Fig. 5 and the calculated (see Table 2) effective value of the current in the secondary winding, equal to 1.6 A, we get a temperature rise of 25 °C for tracks with a thickness of 35 microns and about 7 °C for tracks with a thickness of 70 microns.

We have assumed that the temperature rise due to winding losses is about half of the total temperature rise, in this case 17.5 °C. Obviously, with a track thickness of 35 µm, the temperature rise caused by an effective current of 1.6 A will be too great, so 70 µm thick tracks will have to be used.

The width of the tracks of turns of the primary winding can be calculated by formula (5). It will be equal to approximately 416 microns. With this track width, an effective primary current of 0.24 A is unlikely to cause any temperature increase.

Since the frequency is 120 kHz, an additional increase in PCB temperature of about 2 °C is expected compared to a situation where only direct currents flow. The total PCB temperature rise caused by current flow alone will remain below 10°C.

A six-layer printed circuit board with 70 µm tracks should function in accordance with the calculated parameters. The nominal thickness of the PCB will be about 1920 µm, which means that the standard E-PLT18 E-core and plate combination will not work in this case. A standard E-E18 combination of two E-shaped cores with a winding window of 3.6 mm can be used. However, such a large winding window seems redundant here, so a non-standard core with a window of about 2 mm would be a more elegant solution.

Measurements on a comparable design with two 3C90 E-shaped ferrite halves showed a total temperature rise of 28°C. This is consistent with our calculations, which gave a temperature rise of 17.5°C due to core losses and 10°C due to winding losses.

The coupling between the primary and secondary windings is good as the leakage inductance is only 0.6% of the primary winding inductance.

Example 2: Forward running transformer

Here the goal is to design a direct transformer with a choice of one of the four transformation ratios that are often used in low power DC/DC converters. The desired characteristics are shown in the table above.

First you need to check if the combinations of the smallest cores from the standard nomenclature, E-PLT14 and E-E14, are suitable for this case. Calculating the maximum allowable core loss density at 50 °C temperature rise, we get 1095 mW/cm3 for the E-E14 combination of two E-cores and 1225 mW/cm3 for the E-PLT14 combination of E-core and plate. Next, we calculate the loss density in the core using formula (3) in the case of a unipolar triangular induction wave with a frequency of 500 kHz for several values ​​of peak induction.

The results obtained show that at a peak magnetic induction of about 100 mT, the losses are less than the maximum allowable losses calculated by formula (2). The number of turns and effective currents are calculated using the formulas given in Box 1. With a peak magnetic induction of 100 mT and the parameters given above, it turns out that at 530 kHz, combinations of E-E14 and E-PLT14 are suitable for use, and the number of turns is acceptable. The calculation results are shown in Table 4.

Table 4. Calculation of design parameters of several direct transformers

Core	V in , V	V out , V	N1	N2	L prim , μH	I o (eff.) , mA	I mag , mA	I p(eff.) , mA
E-PLT14	48	5	14	3,2	690	2441	60	543
	48	3,3	14	2,1	690	3699	60	548
	24	5	7	3,2	172	2441	121	1087
	24	3,3	7	2,1	172	3669	212	1097
E-E14	48	5	14	3,2	855	2441	48	539
	48	3,3	14	2,1	855	3669	48	544
	24	5	7	3,2	172	2441	97	1079
	24	3,3	7	2,1	172	3669	97	1080

The final determination of the core loss density at an operating temperature of 100 °C for the specified induction waveform at 530 kHz yields results of 1030 mW/cm3 for 3F3 ferrite and 1580 mW/cm3 for 3F4 ferrite. Obviously the best option is 3F3. The temperature rise in the E-PLT14 core is:

(calculated loss density in 3F3/admissible loss density) X 1/2DT = (1030/1225) X 25°C = 21°C.

For the E-E14 combination, the temperature rise is 23.5 °C. For the primary winding, depending on the input voltage, 7 or 14 turns are required. In the case of a conventional direct transformer, the same number of turns is required for the demagnetizing (restoring) winding. In order to be able to use 7 or 14 turns and the same number of turns for the degaussing winding, a design with 4 layers of 7 turns each has been chosen. When 7 turns of the primary and demagnetizing windings are needed, the turns of the two layers are connected in parallel. This will give an additional effect - a halving of the current density in the winding tracks.

When 14 turns of the primary and demagnetizing windings are needed, the turns of the two layers are connected in series, so that the effective number of turns becomes 14.

The available winding width for the E-14 core is 3.65 mm. For an economical design with a track spacing of 300 µm, the track width at 7 turns per layer is 178 µm.

The thickness of the tracks should be 70 µm, since at an input voltage of 24 V, the effective current in the primary winding will be about 1.09 A. This gives (see table. 2) with an effective track width of 356 µm (the width is doubled as a result of the parallel connection of the parts of the winding when using 7 turns) temperature rise 15 °C. An input voltage of 48 V will produce an effective current of approximately 0.54 A.

In this case, the contribution of losses in the winding to the total temperature increase will be about 14 °C for a track width of 178 µm (14 turns connected in series).

A track width of 178 µm with a track spacing of 300 µm and a track thickness of 70 µm deviates somewhat from our rule of thumb (track spacing and track width > 200 µm). This can lead to somewhat higher production costs for multilayer printed circuit boards. The secondary winding requires 3 or 2 turns. When one layer is allocated to each of the turns, the track width is 810 and 1370 µm, respectively. Effective currents in the secondary windings of 2.44 and 3.70 A cause a temperature rise in the windings of approximately 25 °C, which, taking into account the temperature increase in the primary windings, turns out to be too much. In this case, the best solution is to use 2 layers for both windings. When these layers, each with 3 turns, are connected in parallel, the current density is halved. From fig. 5, it can be determined that the contribution of winding losses to the total temperature increase in this situation will be about 6 °C. The total temperature rise in the PCB will be approximately 21°C plus the additional rise due to AC losses. Since the frequency is 500 kHz, about 10 °C more must be added, which means that the temperature of the PCB will increase by 31 °C.

The number of turns and width for each layer of this design are shown in Table 5. At least one layer, indicated in the table as optional, is required to make connections. However, this will give us a total of 9 layers, which in terms of production is equivalent to 10 layers (the next even number). For this reason, the top and bottom layers of the PCB are used as additional layers - also because it has the added benefit of halving the current densities in the tracks. The traces on these layers are connected to the traces in the inner layer through copper-plated holes and "bring" the inputs and outputs of the primary and secondary windings to two sides of the printed circuit board. Depending on how the inputs and outputs are connected on the primary and secondary sides, 4 different transformation ratios can be obtained.

Table 5. 10-layer design example

Layer	Number of turns	70 µm
stencil		50 µm
extra layer		70 µm
insulation		200 µm
primary degaussing	7	70 µm
insulation		200 µm
primary	7	70 µm
insulation		200 µm
secondary	3	70 µm
insulation		200 µm
secondary	2	70 µm
insulation		200 µm
secondary	2	70 µm
insulation		200 µm
secondary	3	70 µm
insulation		200 µm
primary	7	70 µm
insulation		200 µm
primary degaussing	7	70 µm
insulation		200 µm
extra layer		70 µm
stencil		50 µm
TOTAL:		2600 µm

The total nominal thickness of the PCB will be about 2.6 mm, which exceeds the available winding window of the E-PLT14 core combination of 1.8 mm. An E-E14 combination can be used, however it has a minimum winding window of 3.6mm - much larger than is actually required. A better solution would be a non-standard core with a reduced window size.

Temperature measurements of this printed circuit board were made using thermocouples under various conditions. For testing, the 24/5 V conversion option was used, which gives the highest current densities. First, direct currents equal to the calculated ones were separately applied to the primary and secondary windings. A direct primary current of 1079 mA resulted in a temperature rise of 12.5°C, and a secondary current of 2441 mA gave a temperature rise of 7.5°C. As you might expect, when both currents were applied to the PCB at the same time, the temperature rise was 20°C.

The procedure described above was repeated for alternating currents of several frequencies with effective values ​​equal to those calculated. At a frequency of 500 kHz, the total temperature rise in the printed circuit board was 32 °C. The largest additional temperature rise (7 °C) caused by AC losses was observed in the secondary windings. This is logical, since the effect of the skin effect is more pronounced in the wide tracks of the secondary windings than in the narrow tracks of the primary windings.

Finally, temperature measurements were made with standard cores (combination E-E14) installed on the printed circuit board, under conditions corresponding to the working conditions of a direct transformer. PCB temperature rise was 49°C; the point of maximum heating of the core was on its upper side and the temperature in it was 53 °C. In the central part of the core and its outer part, an increase in temperature of 49 °C and 51 °C, respectively, was observed.

As predicted by calculations, this design is somewhat critical for a set of two E-shaped cores, since a temperature of 53 °C was recorded at the point of maximum heating, which is above 50 °C. However, when using flatter (non-standard) E-shaped cores, the temperature is within acceptable limits.

In the next article, we will consider an example of calculating a 25-watt DC / DC converter based on a planar transformer.

Literature

1. Mulder S. A. Application note on the design of low profile high frequency transformers. Ferroxcube Components. 1990.
2. Mulder S. A. Loss formulas for power ferrites and their use in transformer design. Philips Components. 1994.
3. Durbaum Th., Albach M. Core losses in transformers with an arbitrary shape of the magnetizing current. EPE Sevilla. 1995.
4. Brockmeyer A. Experimental evaluation of the influence of DC premagnetization on the properties of power electronic ferrites. Aachen University of Technology. 1995.
5. Ferroxcube Components technical note. 25 Watt DC/DC converter using integrated planar magnetics. 9398 236 26011. 1996.

Planar transformers are an excellent alternative to standard transformers and wire wound chokes. Planar transformers are based on multilayer printed circuit boards.

Today, the development of planar transformers requires the use of components with minimal dimensions, because the dimensions of electronics are constantly decreasing.

Planar power transformers

The design of planar power transformers can be carried out either with shed components, for example in a single layer or small multilayer board, or as a multilayer printed circuit board.

Advantages of planar transformers:

• are small in size;
• have excellent temperature characteristics;
• have low leakage inductance;
• have excellent repeatability properties.

Due to the higher ratio of the surface area of ​​the core to its volume, the thermal resistance of such devices can be 2 times lower than in conventional wire-wound transformers.

Fig 1. Design of planar transformers

Therefore, due to their increased cooling capacity, planar transformers can handle higher power throughput densities while keeping temperature rise within acceptable limits.

Planar transformers based on multilayer printed circuit boards

When it comes to semiconductor components, including passive ones such as capacitors and resistors, there is quite a lot to choose from.

However, today we will talk about planar transformers.

As a rule, in many cases, developers use standard transformers and chokes that are wire wound. But we will describe planar transformers (PT) based on multilayer boards.

Since the cost of multilayer boards tends to decrease, planar transformers are gradually replacing conventional ones. Especially in cases where a small-sized magnetic component is required.

In planar transformer production technology, the windings are tracks on a printed circuit board or copper sections, which are printed and separated by various layers of insulating material.

Also, the windings can be made from multilayer boards. They are placed between small-sized ferrite cores.

Regarding the design of planar transformers, they can be divided into several types.

• Mounted planar components – they stand closest to conventional inductive components. They can replace conventional parts on single or multilayer printed circuit boards. The height of the hinged planar component can be reduced by plunging the core into the PCB cutout. In this case, the winding should lie on the surface of the board.
• Hybrid type of planar transformers. This type involves embedding part of the windings in the motherboard. At the same time, the other part of the windings is on a multilayer printed circuit board, which is connected to the motherboard. But in this case, the motherboard must have holes for the ferrite core.
• The winding is fully integrated into the multilayer printed circuit board. The halves of the cores are connected by gluing or clamping. It all depends on the preferences of the customer and the manufacturer.

Benefits of planar technology

Compared to conventional wire winding, the planar technology for manufacturing magnetic components has a number of advantages.

Planar transformers found their very first application in power conversion. For this purpose, medium and high-frequency ferrites were used in planar transformers. It was possible to buy a planar transformer from the manufacturer.

If you are interested in the development of planar transformers to order, then you can increase the inductance of the mains filter choke if you replace the powerful ferrite materials with high magnetic permeability.

In pulsed signaling, a broadband transformer between the pulsed generator IC and the cable provides decoupling and impedance matching. In the case of an S- or T-interface, this must also be high permeability ferrite.

Not so long ago, I was approached by a company that needed to develop a line of LED drivers. I will not name the name of the company and the performance characteristics of the drivers, I did not sign the NDA, but ethics are ethics. It seems to be an ordinary order for a driver, which dozens are recruited per year, but there were two mutually exclusive requirements: price And dimensions.

The task from the point of view of circuitry is simple, but from the point of view of production and design it turned out to be very interesting. And so - it was necessary to make a network driver for LED with a power factor corrector (power about 100 W), which cost was around $3 on the series and had dimensions in height no more than 11 mm! Many will say: "What's the problem with making a Deshman driver?" one more requirement - it is possible to give without fear 5 years warranty. And here the most interesting begins.

The choice of topology, circuitry was made, everything fit into the dimensions and cost, but such a wonderful picture was spoiled by the "classic" transformer. It is huge, it is expensive, it is technologically difficult to manufacture. It remained to solve the last problem, and after two days of thought and calculation, it was found - planar transformer.

If you are wondering between what and what the choice was made, on what arguments it was based and how it was possible to get the cost of a transformer less than $ 0.5, then I invite you to a tackle. Well, to improve the "appetite" I am attaching you a photo of the finished transformer:

The main disadvantages of "classic" transformers

I think it's no secret to anyone what an ordinary transformer looks like, but suddenly someone missed the last 150 years of the industrial revolution, so let me remind you:

It looks like a conventional transformer wound on a frame from an RM12 core. Why is he so bad? There are several reasons for this, of course, some of them lose their relevance in certain tasks, but the story will be conducted in the context of the task that confronted me. And here are the main ones:

• Height. Even a person with a poor eye can roughly estimate the size of a transformer from a photograph and say with confidence: "It is definitely more than 11 mm." Indeed, the height of the transformer on the RM12 is about 24 mm, which is more than 2 times the required value.
• Manufacturability. When you need to wind 1-2 transformers, then you take the frame, wire and wind it. When you need to wind 100-200 pieces, you can order winding in your country, the price does not bite yet. When you need to wind 10,000 pieces, and then another 50,000, then there are a lot of nuances: price, quality, choosing another contractor in Asia. All this increases the final cost of the product, when I just need super cheap and very high quality.
• Repeatability. It is very difficult to wind and assemble two identical transformers, it is impossible to make 10,000 identical transformers. I have experienced this in my own skin more than once, especially when it comes to production in SA. Now imagine
  that you will have to "finish with a file" these 10,000 transformers at the final assembly. Represented? Did you feel sad about the amount of labor, and therefore the cost? I think it has.
• Cost price. This is generally a very difficult point, but let's look at the photo above and see that to assemble a classic transformer, we need a frame, core, staples, copper wire, insulation, and all this by hand or on a semi-automatic machine. Let's say it all costs X dollars. For the manufacture of a planar transformer, only the core is needed. I think it’s obvious here that 1 part is clearly cheaper than 1 of the same part + 4 more components?

At this point, you must be tormented: "If everything is so bad, then why are conventional transformers so common?" A little earlier, I said that some of these disadvantages in certain tasks are not a disadvantage. For example, if you open UPS on-line, you will see that the transformer is not the largest element there. And if you collect small batches of up to 100–200 devices per month, then the cost will probably even out, because. 100-200 pieces can already be made in Russia or hire a winder, buy a Chinese machine or make it yourself for 100-200 thousand rubles. and enjoy life.
And perhaps the main place where planar transformers will not replace conventional ones is converters with a rated power more than 2000 W.

Planar transformer device

In the very first picture you see this type of transformer already assembled, the view is very unusual, isn't it? Although people who have opened modern TVs, laptop chargers (not cheap ones) have probably already seen such transformers or similar.

Planar transformers can be made in different designs, there is no clear classification as far as I know, but I divide them into 2 types:

Whatever type of planar transformer is considered, they have one thing in common - all windings are made in the form of copper tracks on a printed circuit board.

If you decide to get acquainted with this technology in more detail and go to Google, then you will probably come across the phrase in many articles: “... and finally, in recent years, planar transformers have become affordable. This is due to the fact that multilayer boards have fallen in price. When I was designing my first planar transformer, in 2010-11, this phrase baffled me. I naively thought that planar planes are made exclusively on multilayer printed circuit boards. At that time, I was still studying at a university, and although I worked and received a good scholarship, this type of payment was not very affordable for me financially. I thought about it and decided to make my own Facebook! to reduce the cost of this technology, as it turned out later - he invented a bicycle.

The essence of the reduction in price was the use of a "pie" of several two-layer printed circuit boards of small thickness (0.8 or 1 mm). For me, it seemed ingenious and simple solutions. The only problem was that, as always, I looked at the solutions of top power electronics companies, such as Texas Instruments, Linear, Infineon, Murata, and they used printed circuit boards in 6–8 layers, and in 2010 they even used standard 4 class (0.15 / 0.15 mm) were very expensive. Then it turned out that I was invited to a good company for a summer practice, and there they told me and showed me that they have been doing such “pies” for planar transformers for 10 years already. So did other companies lower in rank than TI and Infineon. The main one is the idea was right and such a decision is not only correct, but also time-tested.

All elements of the "pie" are ordinary two-layer boards of a standard accuracy class, which means they are sooooo cheap and any printed circuit board manufacturer can make them. The elements of the “pie” of a planar transformer look like this:

As you can see, there are only 3 elements in my transformer, although there could be more. Why 3? According to my calculations, in order to gain the desired inductance in the primary winding, I need 6 layers. 2 layers gives me the main board + 2 layers of "piece of pie" + 2 layers of "piece of pie". The secondary winding fit in only 2 layers, from here another "piece of the pie". As a result, it has a stack of 4 two-layer printed circuit boards. Further arithmetic is even simpler: I use the ELP18 / 4/10 core, which means that the distance under the “windings” is 4 mm. We divide this distance by the number of boards: 4 mm / 4 boards = 1 mm - the thickness of each printed circuit board. Everything is simple!

If you suddenly don’t understand where the 4 mm gap came from, you can see the datasheet for the core here. And for those who are not comfortable following links or don’t want to spend traffic on a large pdf-ku, a small clipping:

As you can see, the size of the core window on one half is 2 mm, on the second half it is also 2 mm. We get the total window size in height - 4 mm.

Now you can make out what the cost of a planar transformer consists of. In fact, there are only 2 components: the core and 3 printed circuit boards. The core costs $ 0.14 in bulk, printed circuit boards are 3 pieces at $ 0.11 each, also on a series. We get $0.47 for the transformer itself. I did not include the core bonding compound here, because if you scatter its cost over the entire batch, then even 1 cent does not work there and did not count the assembly work. Work is not considered for one simple reason - the transformer is assembled at the stage of manual installation, and it costs a penny in Asia. For comparison, soldering 2 transistors in the TO-220 package costs the same as installing a planar transformer, that is, again, a minuscule comes out. This is how we get the number 0.5$ for 1 transformer up to 100W.

A little about my results ... I managed to fit in the height dimension and even do it better - instead of the limiting 11 mm, I got 9.6 mm. On the one hand, it is hardly noticeable, but in practice this is a decrease in size by about 13%. Moreover, the main height dimension was no longer set by the transformer, but by electrolytic SMD capacitors at the input and output.
At cost - I can't tell you the exact figure, but it turned out to meet the requirement. Here it is worth noting the efforts of the customer himself, he managed to find suppliers who, for a large series, were able to give prices at the level, and sometimes a little lower than on digikey. Personally, my merit is that I solved the technical problem and did it cheaply, and the customer himself has already done it super-cheaply without losing quality.

Technical possibilities offered by a planar transformer

Further, my article becomes more technical than narrative, and if you are not interested in power electronics, dry calculations and other nasty things, then you can stop reading and move on to discussions in the comments. There will be no more beautiful pictures. If you are planning to adopt this technology for yourself, then everything is just beginning for you.

So that you can more clearly assess the full potential of this type of transformer, I can say that in this project, on one pair of ELP18/4/10 cores, I managed to build a 65 W resonant converter. And now look at its overall dimensions, is it not bad for such a trifle?

Planar Transformer Calculation Method

There are a lot of methods that allow you to calculate this type of transformers. True, the main literature, including scientific, is mainly in English, German and Chinese. I tried several in practice, they were all taken from English-language sources and all showed an acceptable result. In the process of working for several years, I made small changes that allowed me to slightly increase the accuracy of the calculations, and I will demonstrate this technique to you.

I do not have any ambitions for its uniqueness, nor do I guarantee that its results are sufficiently accurate in all frequency and power ranges. Therefore, if you plan to use it in your work, then be careful and always follow the adequacy of the results.

Calculation of a planar transformer

When calculating any transformer, the first step is to find the maximum value of magnetic induction. Losses in the core and in copper conductors lead to heating of the transformer, therefore, calculations must be carried out relative to the maximum allowable overheating of the transformer. The latter is selected based on the operating conditions and requirements for the device.

We make an empirical assumption in which we assume that half of the total losses on the transformer are losses in the core. Based on this assumption, we calculate the maximum loss density in the core using the empirical formula:

Where is the value of the effective magnetic volume VE is taken from the documentation for the core in [cm 3 ], the value of the maximum superheat ∆T is chosen based on calculations (for example, I usually take into account 50–60 degrees). The dimension of the resulting value - [mW/cm3].

Please note that many of the formulas that I describe are empirically obtained. Others are written in their final form without scheduling their mathematical derivation. For those who are interested in the origin of the latter, I advise you to simply familiarize yourself with foreign literature on magnetic materials, for example, there are also books by Epcos and Ferroxcube.

Now, knowing the maximum loss density in the core, we can calculate the maximum inductance value at which the overheating temperature above the calculated one will not be exceeded.


Where CM, ST, x, y- parameters obtained empirically by the method of approximating the loss curve, and f- conversion frequency. You can get them in two ways: by processing the data (graphs) from the documentation for your core, or by building these graphs yourself. The latter method will allow you to get more accurate data, but you will need a full-fledged thermal imager.

As an example, I will share with you these values ​​for cores made of material Epcos N49, its analogue from Ferrocube is also a popular and affordable material 3F3. Both materials make it possible to easily build converters with a resonant frequency up to 1 MHz inclusive. It is also worth noting that these parameters depend on the frequency, these figures for frequencies 400-600 kHz. This is the most popular frequency range and material I use.

• CM = 4.1×10–5
• ST = 1.08×10–2
• x = 1.96
• y=2.27

Next, it is worth remembering the second component of losses in the transformer - losses in the copper winding. They are considered easily, according to our favorite Ohm's law, in which quite logical points were additionally taken into account: our current is pulsed and it does not flow 100% of the time, that is, the fill factor. I won’t tell you how to calculate the resistance of a copper winding according to its geometry, it’s too banal, but I’ll probably remind you of the general formula:

Losses in copper are calculated for each winding separately, and then added up. Now we know the losses in each layer of the "pie" and in the core. Those who wish can simulate transformer overheating, for example, in Comsol or Solidworks Flow Simulation.

Continuing the topic of copper conductors, let's recall such a phenomenon as skin effect. If you explain “on the fingers”, then this is the effect when, with an increase in the frequency of the current flowing in the conductor, the current is “squeezed out” from the conductor (from the center to the surface) by another current - eddy.
Speaking more scientifically, as a result of the flow of an alternating current in the conductor, a variable induction is induced, which in turn causes eddy currents. These eddy currents have a direction opposite to our main current and it turns out that they are mutually subtracted and in the center of the conductor the total current is zero.
The logic is simple - the higher the frequency of the flowing current, the more the skin effect affects and the lower the effective cross section of the conductor. Its influence can be reduced by optimizing the geometry of the windings, their parallelization and other methods that probably deserve, if not a whole book, then a large separate article.
For our calculations, it is enough to roughly estimate the influence of the skin effect using another empirical formula:

Where ∆δ - thickness of the zone with zero current, f- converter frequency. As you can see, this effect is entirely tied to the switching frequency.

And now let's calculate how many turns and other things we need to make a forward-running transformer. First of all, we consider how many turns we need in the primary winding:

Where Umin is the minimum input voltage, D is the duty cycle, f is the operating frequency, Ae is the effective core section. Now we count the number of turns for the secondary winding:

Where N1 is the number of turns in the primary winding, D is the duty cycle, Uout is the rated output voltage, Umin is the minimum input voltage.

The next step is to calculate the inductance of the primary winding. Since the current in the winding has an impulse response, it will also depend on the inductance. We calculate it using the following formula:

Where μ0 is the effective magnetic permeability, μa is the amplitude magnetic permeability, Ae is the effective cross section of the core, N1 is the number of turns in the primary winding, Ie is the effective path length. You can take the missing parameters, such as permeability and magnetic line length, in the documentation for a specific core.

Now the final step that we need to take is to calculate the current in the primary winding. This will allow in the future to calculate the cross section for the primary winding and, accordingly, the width of the conductor. The current value is the sum of two components and looks like this:


Here it seems that all the components of the formula are already familiar and calculated, the only thing I will note is the Pmax parameter. This is not just the value of the rated output power, it is the total power of the converter, taking into account the efficiency at least approximately (I usually lay 95–97% for resonant converters) and the margin that you put into the device. In my devices, there is usually a 10% margin in power, in especially critical devices and nodes, sometimes you have to lay a 20–25% margin, but this causes a rise in price.

So we got all the parameters that are necessary for the calculation and design of a planar transformer. Of course, you will have to calculate the cross section for the windings yourself, but this is elementary arithmetic, which I do not want to clutter up the article with. Everything else has already been calculated and it remains only to design the boards in some kind of CAD.

Outcome

I hope my article will help you start using planar transformers both in your home projects and in commercial ones. This technology must be used carefully, because depending on the task, it can be more expensive than "classic" transformers.

Also, undoubtedly, the use of planar transformers opens up new technical possibilities, and modern Mosfets and new GaN transistors only contribute to this, allowing you to create converters with frequencies from 400 kHz and higher. However, the cost of these "opportunities" is not always low enough, and the design of resonant converters at such frequencies requires a large set of knowledge and experience.

But do not be upset! Any of you, even a beginner electronics engineer, can assemble topologies in a simpler way, for example, a ZVS bridge (Full bridge). This topology allows you to get a very high efficiency and does not require any super-secret knowledge. You just need to make a prototype or layout and experiment well. Good luck in exploring new horizons!

read 14146 times

The constant reduction in the size of electronic products, especially mobile devices, leads to the fact that developers have to use components with minimal dimensions. For semiconductor components, as well as passive components such as resistors and capacitors, the choice is quite large and varied. We will consider a small-sized replacement for another passive element - transformers and chokes. In most cases, designers use standard transformers and wirewound chokes. We will consider the advantages of planar transformers (PT) based on multilayer printed circuit boards. The cost of multilayer printed circuit boards is constantly decreasing, so planar transformers will be a good replacement for conventional ones.

Planar transformers provide an attractive alternative to conventional transformers where small magnetic components are required. With planar technology for the manufacture of inductive components, the role of windings can be played by tracks on a printed circuit board or copper areas applied by printing and separated by layers of insulating material, and in addition, windings can be constructed from multilayer printed circuit boards. These windings are placed between small ferrite cores. According to their design, planar components are divided into several types. Closest to conventional inductive components are planar plug-in components, which can be used instead of conventional parts on single and multilayer printed circuit boards. The height of the hinged component can be reduced by plunging the core into the cutout of the printed circuit board so that the winding rests on the surface of the board. The step forward is a hybrid type, where part of the windings is built into the motherboard, and part is on a separate multi-layer printed circuit board that is connected to the motherboard. The motherboard must have holes for the ferrite core. Finally, in the latter type of planar components, the winding is fully integrated into the multilayer printed circuit board.

As with conventional wirewound components, the halves of the cores can be joined together by gluing or by clamping, depending on the manufacturer's capabilities and preference. FERROXCUBE offers a wide range of planar E-cores for various applications.

Benefits of planar technology

The planar technology for manufacturing magnetic components has a number of advantages over conventional wire winding. The first obvious advantage is the very low height, which makes planar components promising for high-density rackmount and portable applications.

Planar magnetic components are well suited for the development of high efficiency switching power converters. Low ac copper loss and high coupling coefficient provide more efficient conversion. Due to the low leakage inductance, voltage spikes and fluctuations, which are the cause of failure of MOS components and an additional source of noise, are reduced.

Planar technology is simple and reliable in production. Tables 1-3 describe the advantages and limitations of this technology.

Table 1. Development Benefits

Table 2. Manufacturing Benefits

Table 3. Restrictions

(1) The cost of multilayer PCBs is reduced. Overall cost: no frame required, smaller core size.

Integrated vs. Attached Components

Integrated planar components are used when the complexity of the surrounding circuitry forces the use of a multilayer printed circuit board. Typical applications are low power converters and signal processing devices. They mainly use a combination of a W-shaped core and a small plate. The main design requirements here are low height and good high-frequency performance.

• Mounted components are used differently. Typical applications are high power converters; they mainly use a combination of two large size E-shaped cores. Thermal characteristics are the main design requirements here. The design of the winding depends, in particular, on the magnitude of the current.

Immersing add-on components into the board allows you to reduce the height of the assembly without changing the location of the components.

Hybrid components reduce the number of sheath windings at the expense of traces on the printed circuit board, and in the integrated version there are no sheath windings at all. Combinations of the two types are also possible. For example, a power converter may have the primary winding of the transformer and the mains filter choke built into the motherboard, while the secondary winding and the output choke are on separate printed circuit boards (Fig. 3).

Bonding versus clamping

The choice between bonding and clamping depends primarily on the capabilities and preferences of the manufacturer, but there are also application-specific requirements that may determine one or the other as more desirable.

The first field of application for planar transformers was power conversion. Accordingly, medium- and high-frequency powerful ferrites were used. The inductance of the mains filter choke can be increased by replacing the powerful ferrite with a material with high magnetic permeability. In pulsed signaling, a broadband transformer between the pulsed generator IC and the cable provides decoupling and impedance matching. In the case of an S- or T-interface, this must also be high permeability ferrite. 3E6 high permeability ferrite cores have been added to the FERROXCUBE product range. Below is a list of applications where planar technology can be beneficial.

Power conversion

• Components
  • Power transformers, output or resonant chokes, mains filter chokes.
• Rectifiers (mains power supplies)
  • Switching power supplies.
  • Chargers (mobile phones, laptops).
  • Control and measuring equipment.
• DC converters
  • Power conversion modules.
  • network switches.
  • Mobile phones (main power source).
  • Portable computers (primary power source).
  • Electric vehicles (traction voltage converter to 12 V voltage).
• AC converters (line power supplies)
  • Compact converters for fluorescent lamps.
  • Induction heating, welding.
• Inverters (battery power supplies)
  • Mobile phones (LCD backlight).
  • Notebook computers (LCD backlight).
  • Gas-discharge automobile headlights (ballast).
  • Heated car rear window (boost converter).

Pulse transmission

• Components
  • Broadband transformers.
  • S 0 -interfaces (subscriber telephone line).
  • U-interfaces (ISDN subscriber line).
  • T1/T2 interfaces (trunk line between network switches).
  • ADSL interfaces.
  • HDSL interfaces.

Table 4. Characteristics of materials

Table 5. Cores for bonding (without notches)

Table 6. Core materials for bonding

(*) - halves of cores for use in combination with an E-shaped core without a gap or a plate.

(**) - halves of cores with high magnetic permeability.

E160 - E - half core with symmetrical gap. A L = 160 nH (measured in combination with half a core with a symmetrical gap).

A25–E - half core with asymmetrical gap. A L = 25 nH (measured in combination with half core without gap).

A25 - P - half core with asymmetrical gap. A L = 25 nH (measured in combination with plate).

1100/1300 - half core without gap. AL = 1100/1300 nH (measured in combination with half core without gap/wafer).

The AL value (nH) was measured at B≤0.1mT, f≤10kHz, T = 25°C.

Tolerance A L:

Table 7. Performance versus power (gluing cores)

Table 8. Clamp-On Cores

Product range

FERROXCUBE offers a wide range of planar E-cores in the size range 14-64mm. In the basic version for bonding, the cross-section is always uniform, which allows optimum use of the ferrite volume. Each size has an E-core (denoted by the letter E) and a corresponding insert (denoted by the letters PLT). The set may consist of an E-core and a plate or two E-cores. In the latter case, the height of the winding window is doubled. For the smallest sizes, there is also a set of E-core and plate in a clamp-on version. It uses a notched E-core (denoted E/R) and a grooved insert (denoted PLT/S). The clamp (designated CLM) snaps into the recesses of the core and provides a strong connection by pressing the plate at two points. The groove prevents the insert from moving, even under strong shock or vibration, and also ensures alignment. There is no clamp connection for the combination of two E-cores.

Table 9. Clamp connection core materials

(1) - core halves for use in combination with a plate.

A63 - P - half core with asymmetrical gap. A L = 63 nH (measured in combination with plate).

1280 - half core without gap.

A L = 1280 nH (measured in combination with plate).

The A L value (nH) was measured at B≤0.1 mT, f≤10 kHz, T = 25 °C.

Tolerance A L:

Table 10 - Power characteristics (clamp connection cores)

Powerful ferrite cores 3F3 (operating frequency up to 500 kHz) and 3F4 (500 kHz - 3 MHz) are available in all sizes. The largest cores are also made from 3C85 ferrite (operating frequency up to 200 kHz), since large cores are often used in high-power low-frequency devices. Smaller cores are also available, made from 3E6 high permeability ferrite (μ i = 12000), for use in mains filter chokes and broadband transformers.

Package

Plastic film is used as standard packaging for planar E-shaped cores and plates.

Table 11. Packaging

Table 12. Core box

Table 13. Terminal box

Table 14. Tape packaging

For E14/3.5/5 and E18/4/10 cores, a tape packaging prototype was developed for use with SMD automatic assembly equipment. The packaging method is in accordance with IEC-286 Part 3. The plates are packaged in the same way as the corresponding E-cores.

Development

To take full advantage of planar technology, a different design concept must be followed than with wire wound. The following are a number of considerations to be followed in this regard.

Core selection

• Magnetic induction

The improved thermal performance allows for twice the power loss of a conventional design with the same amount of magnetic field, so the optimum flux density will be higher than usual.

• Air gap

Large gaps are undesirable in planar designs as they create stray flux. The edge flux depends on the ratio of the winding window height to the air gap width, which is smaller for flat cores. If the height of the window is only a few times the width of the gap, and the width is several times the width of the central part of the core, then a significant amount of flow will occur between the top and bottom of the core. Large values ​​of edge and intersecting flows lead to large eddy current losses in the winding.

Winding design

• DC resistance

The most commonly used copper tracks are 35, 70, 100 and 200 microns thick. If the trace cross-sectional area is insufficient to obtain acceptable DC resistance, it is possible to connect the traces in parallel for all or part of the turns.

• AC resistance

AC copper losses due to skin effect and proximity effect are less for flat copper traces than for round wire with the same cross-sectional area. Eddy currents induced in the vicinity of the air gap can be reduced by removing a few turns at the point where the induction is maximum and directed perpendicular to the winding plane. The combination of an E-core and a plate has slightly less leakage flux than a combination of two E-cores due to the location of the air gap.

• Leakage inductance

When the windings are located one above the other, the magnetic coupling is very strong, and values ​​of the coupling coefficient close to 100% are achievable (Fig. 13, a).

The previous design leads to a higher interwinding capacitance. This capacitance can be reduced by placing the tracks of adjacent windings in between each other (Fig. 13, b).

Moreover, the repeatability of the capacitance value allows it to be compensated in the rest of the circuit, as well as used in resonant designs. In the latter case, it is possible to purposefully create a large capacitance by placing the tracks of adjacent windings opposite each other (Fig. 13, c).

Production

Assembly

When using clamps, you must first snap the clamp into the recesses of the core, and then align the plate laterally.

For integrated components, assembly is combined with mounting.

Mounting

When using add-on components, through-hole boards or SMD mounting can be used. There are no significant differences from the usual process

The flat surface of the core is well suited for automatic mounting.

In the case of integrated components, installation is best done in two steps:

1. Glue one half of the core to the PCB. The same adhesive can be used for this as for mounting SMD components, and this step is logically combined with mounting SMD components on this side of the PCB.
2. Glue the second half of the core to the first. This includes the same comments that were made about the assembly of attachments.

Soldering

Applies to plug-in transformers only.

In the case of reflow soldering, the preferred heating method is hot convection rather than infrared radiation, since the first method ensures that the temperatures of the surfaces to be soldered are equalized. When heated with infrared radiation using standard materials, the good thermal conductivity of the planar component can lead to too low a temperature of the solder paste, and when the radiation power is increased, to a too high temperature of the printed circuit board. If infrared heating is used, a different solder paste and/or PCB material is recommended.

Size designation

All numbers given refer to halves of the cores. The two core halves must be ordered in the correct combination. There are four types of core halves, of which sets of three types are made:

• two W-shaped cores (E+E);
• W-shaped core and plate (E+PLT);
• Notched E-core and slotted plate (E/R + PLT/S).

The last set also includes a clamp (CLM).

The next article will provide a method for calculating planar power transformers for switching power supplies.

Payton Planar Transformers and Chokes (2005)

One of the main tasks in the development of a transformer is to reduce its overall dimensions while increasing the effective power. Today, the transformer is experiencing a second birth - the traditional technology of building a transformer is being replaced by a new planar technology. The principle of building electromagnetic devices using the new technology is to use printed circuit boards instead of a frame assembly and wire winding. The role of the winding in planar technology is performed by tracks printed on the board. The boards are stacked in several layers, separated by an insulating material, and enclosed in a ferrite core.

planar technology
Until the mid-1980s, planar transformer manufacturing technologies were limited mainly to developments in the military, aviation and space industries. At the origins of the active commercial application of planar technologies was Alex Estrov, who published in 1986 some data on his developments in the field of planar transformers operating at a resonant frequency of 1 MHz. The idea was a success. Some time later, A. Estrov organized a company (today it is called Payton Power Magnetics Ltd.), which launched mass production of planar power transformers and chokes.
What is planar technology and why is it remarkable? Consider an example that explains the principle of building planar transformers (Fig. 1). The figure shows a disassembled transformer. It consists of several plates with winding turns applied to them and insulating plates separating the winding plates from each other. The winding of the transformer is made in the form of tracks on printed circuit boards or sections of copper printed on the board. All layers are placed on top of each other and held by two pieces of a ferrite core.
The desire to reduce overall dimensions while increasing power is the main goal of the development of modern power devices. At the same time, planar transformers, unlike traditional ones, have a relatively large effective cooling area and are easier to cool - you can use various options: natural, forced, one-sided and two-sided radiator, liquid cooling.
Another positive feature of planar devices is a small spread of electrical parameters from device to device. A wire-wound transformer has a large spread of parameters, since the wire lies unevenly on the frame during the winding process, which cannot but affect the device parameters (for example, inductance, quality factor). Planar transformers are assembled on the basis of multilayer printed circuit boards. Each board is made in the same way. The tracks on the boards are also printed. Board etching is always the same process. The parameter errors of a planar transformer are hundreds of times smaller than those of a traditional wire-wound transformer.
Planar transformers are ideal for telecommunication systems, computers, aircraft onboard systems, power supplies, welding machines, induction heating systems - i.e. wherever power transformers with high efficiency and small dimensions are needed.
The main advantages of planar transformers:
high power with small overall dimensions (10 W - 20 kW);
high efficiency of devices (97–99%);
wide operating temperature range: from -40 to +130°С;
dielectric strength of devices 4-5 kV;
low leakage inductance;
the operating frequency range of planar devices lies in the range from 20 kHz to 2.5 MHz;
high power with small dimensions: planar transformers usually include from one to seven windings;
small spread of parameters in serial production of devices;
very low level of electromagnetic interference;
small dimensions and weight.

Payton Planar Transformers
Payton manufactures a wide range of planar transformers ranging from 5W to 20kW. Payton transformers are small in size (Fig. 2), capable of operating at high powers and providing good thermal performance. Table 1 provides data on power size, weight and core size.

The Payton product line includes devices rated at various power levels for use in telecommunications equipment, power supplies, AC/DC and DC/DC voltage converters, and the like. Table 2 shows the main characteristics of some types of Payton planar transformers.
Initially, Payton developers focused on the production of transformers only for switching power supplies (SMPS), for use in welding machines and induction heating systems. However, now they are used almost everywhere.
Modern Payton transformers are ideal for SMPS applications for welding machines. Transformers fit perfectly into the structure of the source, guaranteeing a long duration of its operation. It is known that SMPS of welding machines generate critically high values ​​of output currents. Therefore, in most cases there are only a few secondary turns. Planar transformers are thus suitable for handling high currents and can be used in welding equipment. The use of planar transformers can significantly reduce the size and weight of the final device.

The planar transformer also fits well into the structure of power supplies for induction heating systems. For these purposes, for example, a 20 kW transformer was produced (Fig. 3) with dimensions of 180x104x20mm.
Payton Power Magnetics offers transformers with leads for a variety of mounting options, both surface mount and through PCB mount options. The flat surfaces of the cores are suitable for automatic mounting. In addition, there are devices with terminals for surface mounting.

Payton planar chokes
Payton manufactures a wide range of planar chokes. Payton chokes, like transformers, provide significant power in small sizes. The chokes are produced using the technology of pre-magnetization of the core. Although this technology has been known for a long time, it has not been widely used due to the high cost of special magnetic materials traditionally used to make cores, the impossibility of operating devices at high frequencies, and the deterioration in performance due to demagnetization of the core. Payton's engineers have overcome these shortcomings by using ferromagnetic cores, an inexpensive and effective replacement for special magnet cores.
The pre-magnetization technology of the cores allows you to double the value of the inductor without changing the current, or double the value of the current with the same inductance. The new technology for the production of chokes makes it possible to reduce power losses by 4 times and reduce the contact area by 30–40% (Fig. 4).
Testing of chokes for the deterioration of magnetic properties showed that at operating frequencies up to 1 MHz, the deterioration of the magnetic properties of the cores does not occur even at a 10-fold excess of the field strength compared to the usual operational value.

Payton Hybrid Chokes
In addition, Payton is actively developing technologies for building hybrid planar chokes that are capable of operating at high resonant frequencies. These devices are based on a "6-knee" planar ferromagnetic core, combined with a stranded winding. This combination allows you to achieve a high quality factor at high frequencies. For example, the value of the quality factor of a choke with an inductance of 40 μH at a current of 3A and an operating frequency of 1 MHz is 500!

Choke Filters Payton
Payton also manufactures planar chokes specifically designed for common mode mitigation. The ratio between leakage inductance and device self-inductance is reduced to 0.005%. Due to their high self-capacitance, planar common mode chokes can include input and output capacitors. Therefore, this type of chokes can be used as a common mode filter. Planar choke filters are already being developed today, which will operate at currents up to 200A.

Conclusion
Due to the stable performance, high efficiency and efficient cooling method of Payton's planar electromagnetic components, their use is an attractive solution for power supply manufacturers. The trend towards cheaper production of multilayer printed circuit boards makes planar transformers more and more affordable for a wide variety of applications. It can be assumed that in the near future planar devices will completely replace traditional wire-wound transformers.