A practical technique to measure transformer losses in high frequency SMPS

Electronics has become an indispensable part of all facets of human life. Entertainment, communication, lighting, health care, information, transport etc. are all invariably linked to utilizing electrical energy. However, the electric supply from the utility grid is in the form of AC and the electrical gadgets need a stable DC supply for their operation. This calls for power supplies which take AC and convert to a stable DC power to the target equipment [1].

In view of the product design requirements, compactness, in terms of size and volume also has become a major specification in the product. This has necessitated the adoption of switch mode power converters which operate at high frequencies [2].

A general purpose AC to DC switch mode converter is detailed below.

AC–DC converter has in its front end a typical full bridge rectifier, which converts the incoming AC mains voltage into a unidirectional time varying voltage at twice the mains frequency. This is followed up by a large high voltage bulk reservoir capacitor which filters out the time varying voltage to an unregulated steady DC voltage. Typically, for an input AC voltage of 230 V AC, the rectified DC voltage at the reservoir capacitor would be about 325 V DC. However, this voltage has some ripple content, based on the reservoir capacitor value and the load current drawn from this source. In addition, the DC voltage would be unregulated and varies with mains voltage. For an input voltage range of 180 V AC to 264 V AC, The DC voltage can be in the range of 225 V DC to 390 V DC. The power converter is supposed to handle this large input DC voltage variation and deliver a rock steady DC output voltage in the range of say 5 V DC, or 12 V or 24 V or 48 V as needed by the end equipment and also should provide galvanic isolation from the input mains supply for safety and EMI/EMC related issues [3].

As a next step, the unregulated DC voltage is connected to a transformer with a ferrite core and a semiconductor switch, either a power transistor or a power MOSFET. The transformer would be capable of operating at high frequencies (around 100’s of kHz) and the semiconductor switch is driven ON and OFF at high frequency. Thus, the circuit acts as a chopper at high frequency and the transformer sees a high frequency AC voltage. Secondary of the transformer, which is electrically isolated from the primary, is rectified with fast recovery rectifier diodes and then connected to a low pass filter comprising of a ferrite core inductor and electrolytic capacitors. Output of the low pass filter is a DC voltage which can be regulated to a precise value, by controlling the ON or OFF time of the semiconductor power switch [3].

A stable reference, high gain error amplifier, sampled output voltage together generate an error voltage which controls the ON or OFF time of the power switch and is transferred to power switch with a galvanic isolated drive. Thus, the AC DC converter delivers a stable lower DC voltage to the end appliance with total isolation from the high voltage utility mains [3].

As discussed earlier, automation has created a sudden demand for the power converters and also for the electrical energy. However, the generation of electricity which is predominantly by fossil fuels has impact on the environment [1]. Hence, the focus is on improving the energy efficiency of electronic gadgets so that the available resources are best utilized. The importance of discussion on the power losses and especially in transformers can be well appreciated by referring to [4, 5], as it discusses the power loss in an electronic power transformer as against a conventional power transformer.

Having said about the efficiency improvement, the first element in the chain of a product is the power converter sitting right at the front end and it would be critically looked at for efficiency improvement. Front end converter handles the complete power requirement of the electrical gadget and therefore needs to be highly efficient.

Miscellaneous losses such as in the bias circuits, housekeeping circuits, filter capacitors, PCB track losses etc. can be neglected as they form a small percentage of loss and their improvement may not add much to the target. Among them the losses in the semiconductor devices are primarily computable, with the available math equations [6].They can be easily verified also, by mounting them on heat sinks which are pre calibrated for their thermal resistance R_th. Thermal resistance is defined as the temperature rise above ambient for every watt of power dissipated and has the units °C/watt [7]. By mounting the semiconductors on the heat sinks with known R_th and measuring the temperature rise of the heat sink directly measures the power loss in the device.

The power loss measurement of the ferrite transformers/inductors is not that simple [8, 9, 10]. Thermal resistance of the cores as depicted from the data sheet cannot be adopted as it is. Data sheets provide thermal resistance for only the core and not for the complete transformer [11, 12]. A wound transformer has a different property altogether.

Both losses are very complicated, non-linear in nature and too involved to theoretically compute [13]. Core loss depends on the frequency of operation, flux swing, geometry of the core and also the grade of ferrite. Likewise copper losses depend on Skin effect, proximity effect, type of topology of converter and the construction of the transformer layers [14].

Therefore, it is matters of importance to at least quantify the exact amount of core loss in ferrite transformer and also characterize the thermal resistance of the transformer. With this, half of the total transformer loss can be precisely estimated. If the accurate thermal resistance value can be obtained for the transformer, then we can also indirectly compute the copper loss.

We propose in this paper a technique to precisely measure the core loss in the designed transformer and also estimate the thermal resistance of the transformer. With these two, exact losses in the transformer can be measured.

Further, this technique can also be used to create a data base for the core losses in various cores at different frequencies and a variety of flux density swings.

The proposed methodology is very simple and easily implemented in the real life applications.

Hysteresis losses of a ferrite core are due to the work done during every switching cycle in magnetizing and demagnetizing the core. The same is well discussed widely under the heading called B-H curve [15]. Core losses depend on frequency of operation, flux density swing, core volume and core material grade. Lot of literature is available from the manufacturers’ data sheets.

When any transformer primary is excited with alternating current (AC) voltage and the secondary is left unconnected, then the current drawn from the primary source voltage indicates the core loss. This is very simple in case of low frequency (50 Hz) transformer with sinusoidal voltage input. Whereas, in high frequency applications with non sinusoidal input voltage and varying ON and OFF times, the measurement of core loss is not that simple. Therefore a simple technique to measure the core loss in a high frequency transformer with non sinusoidal excitation and varying ON or OFF times is essential in practical applications. The following methodology suggests a solution for the above problem. Most important of this technique is that the actual core loss is obtained in the in sit u conditions for the exact field conditions.

Let us consider a power converter set up of Full bridge bridge configuration as shown in Fig. 1 wherein the transformer is fed with bidirectional voltage in every cycle and does not contain any DC component due to presence of a coupling capacitor in series with the transformer primary [13].

Open image in new window

Calculation of the power loss in a high frequency transformer is very involved and a lot of study has been done and reported [8, 16]. For a normal practicing power supply manufacturers and engineers, the theoretical part appears to be too complicated to follow. However, it is very important to know the actual losses in all the power components of a converter, in order to improve the product performance, reliability and product design. Therefore a practical technique to assess the losses in magnetic components is well appreciated and need of the hour.

Since the core losses are non-linearly related to flux swing and frequency of operation, it is in the best interest that the power converter chosen is of variable frequency type. This greatly simplifies in evaluating the core loss at different frequencies and different Flux density levels.

In order to demonstrate the efficacy and usefulness of the scheme proposed, a practical fly back converter of 150 W was designed, fabricated and evaluated to accurately assess the power loss in the power transformer.

The chosen topology for the practical converter was a soft switching fly back configuration due to the fact that it has only a single transformer and fewer components. Input voltage chosen was 190 V DC. Output voltage was selected as 150 V DC so that the output current is only 1.0 A and hence does not add up to unnecessary PCB track losses and diode losses. The configuration of the fly back was a soft switching Boundary Control Mode BCM control topology as cited in [18]. BCM allows the diode losses to be at minimal. BCM control IC FAN7527 was used as the controller. This IC senses the complete energy transfer point by measuring the zero current in the secondary and again turning ON the primary Mosfet switch. BCM control automatically has the switching frequency as a variable for various Line and load conditions. The Block diagram of the flyback converter and the photograph of the fabricated unit are shown in Figs. 4 and 5.

Open image in new window

Open image in new window

Specifically, the soft switching converter with BCM controller is chosen, so as to minimize all the losses which makes it easy to concentrate on the transformer loss alone.

Bias and drive losses are catered to by a separate bias supply. Such a unit was designed and fabricated and tested. The test data such as input power, output power is recorded for a stable DC input voltage. With this data, total losses were arrived at.

In order to evaluate the exact core losses in the above transformer, a 250 W bridge converter is built to operate at 175 kHz. Input voltage and output load were adjusted to obtain 150 V DC bipolar and for 2.5 ms to be impinged across the transformer. Next, an Identical transformer as the one used in the flyback converter is fabricated and the secondary consisting of 15 turns, is connected in parallel to the main transformer of the half bridge converter. The primary of the flyback transformer with 20 turns was left open. With this condition the half bridge converter was run for about 4 h to attain thermal equilibrium.

Under this condition, the observed increment in the input power of the half bridge converter is noted to be 1.75 W. This is the core loss in the fly back transformer at 175 kHz and 250 mT flux swing.

Simultaneously, the rise in the flyback transformer temperature was recorded as 32 °C above ambient.

The Flyback converter with designed transformer is then run at 150 W load continuously till thermal equilibrium and temperature rise of transformer is recorded. The obtained switching frequency was 175 kHz.

Temperature rise of the flyback transformer after thermal stabilization = 68 °C above ambient.

The input bridge rectifier loss is measured to be 1.3 W.

The current sense resistor and the output sense network together consumed about 1.3 W.

With this the total losses measured in the converter of 150 W are 6.1 W, not accounting for transformer. The total converter losses are noted to be 9.54 W. Hence, from this angle; the estimated power loss in transformer is 9.54 − 6.1 = 3.44 W. However, the power loss computed from the proposed scheme by measuring the temperature rise and R_th is 3.72 W which is in close acceptance range. Thus, the proposed method of core loss measurement is validated in a practical set up

Measuring the power loss in a high frequency ferrite transformer is very essential for all power supply engineers. Improving the efficiency of the converter can be done, only if the losses occurring at various levels are known. Therefore, any effort in contributing to the loss estimation is a welcome phenomenon and this paper proposes a simple, implementable technique to practically estimate the losses in high frequency transformers.

Core loss data for two types of cores are measured and presented. Thermal resistance derived practically is also depicted. A practical 150 W flyback converter is built and evaluated for its losses and verified with the proposed technique. Quantum of increased copper loss due to skin and proximity effects is clearly arrived at.

The proposed scheme accurately arrives at the losses in a high frequency power transformer. This enables designers to try out various design options and arrive at an optimum configuration for maximizing the efficiency. The designed 150 W fly back converter achieved an efficiency of 94.06%.