Power Electronics

eGaNTM FET-Silicon PowerShoot-Out Part 3: Power over Ethernet

The eGaN FET is a viable and efficient alternative to standard MOSFET solutions in Power over Ethernet (PoE) applications. These FETs enable higher operating frequencies that can be leveraged into reduced converter size and cost. Both 13 W and 26 W PoE eG

Find a downloadable version of this story in pdf format at the end of the story.

PoE converters are widely used in digital networks to provide operating power to peripheral equipment through existing communication wiring, hence eliminating the need for separate power lines or local power adaptors. PoE controllers are located in the receiving/powered device (PD) as well as the power sourcing equipment (PSE).

PSE converts line voltages to a DC voltage range of 36 V to 57 V with various power ratings to distribute to the peripheral equipment. On the other end, PoE-PD equipment provides the necessary isolation and converts the raw DC power for internal use mostly with fly-back or forward converter topologies. PoE-PD converters are limited by IEEE 802.3at specifications to 13 W and 26 W (25.5 W) of power drawn from the line.

There are three important factors in PoE designs:


    Cost: The number one concern for every system designer of a PoE-PD. The key is that any direct improvement in efficiency can be transformed into a cost reduction.

  2. Efficiency: There are cases where power conversion efficiency is more important than cost and demands a premium - especially where the output power is limited by either the 13 W or 26 W standards. This is more common for 13 W fly-back applications, but as more innovative products are developed for PoE this is likely to spread to 26 W fly-back / forward applications as well.

  3. Size: There is a size /cost relationship that is dominated by circuit magnetics. Any reduction in core size means a reduction in cost. In the future, as functionality increases and output power levels grow, this will become important and size reduction alone will demand a premium.

As any magnetic designer will attest, optimization of a transformer can be challenging. Multiple solutions exist in part due to discrete steps in utilizing standardized core and wire sizes that could make the realizable implementation far from optimum in size and performance. In an attempt to avoid this quagmire of magnetic design variations, our aim is to focus on efficiency improvement resulting from implementing eGaN devices over MOSFETs and keep the magnetic structures constant. The implications of this improvement on cost and size will be discussed in the magnetics section.

To evaluate the physical and electrical performance of eGaN FETs for PoE, three different designs for representative eGaN converters were evaluated and compared to MOSFET equivalent versions of the same design. The results for all three converters can be summarized as follows:


  • Small size 13 W flyback converter
  • High efficiency 13 W flyback converter
  • High efficiency 26 W forward converter

All three converter designs were based as close as possible to the controller manufacturers' suggested circuits. As the driver requirements for eGaN FETs are somewhat different from traditional MOSFETs, it was necessary to add discrete external drivers to the eGaN converter versions, as shown in Fig. 1. Further information on recommended gate drives can be found on the EPC website [1].

eGaN FETs have a better figure of merit (FoM) than equivalent state-of-the-art silicon devices [2] and this advantage increases with rated drain voltage. However, the eGaN devices also have a higher “body diode” forward drop as discussed in [3], which can be detrimental at lower voltages and requires improved controller timing to overcome this limitation. The MOSFET and eGaN FET devices used for each of the converters are listed in the table.

Continue on next page


Fig. 2 shows an implementation of a 48 V to 5 V, 13 W PoE-PD power supply utilizing the LT1725 IC from Linear Technology [4] which is a general purpose fly-back controller. The data sheet for the LT1725 specifies a maximum operating frequency of 250 kHz, however, in this implementation the frequency was adjusted to 400kHz to show the higher frequency advantages of the eGaN FET.

Both eGaN FET and MOSFET based converter efficiencies for operation at 300 kHz and 400 kHz are shown in Fig. 3. The MOSFET (FDS2582, 150 V, 66 mΩ [5]) is compared to an EPC1012 die (200 V, 100 mΩ [6]). It can be seen that the eGaN FET efficiency results are consistently about 2% higher than the MOSFET equivalent converter for all but heavy load despite a 50% increase in the RDS(ON). It is worth noting that eGaN FET efficiency at 400 kHz is still higher than the equivalent MOSFET 300 kHz version over most of the load range.

Waveforms in Fig. 4 shows eGaN FET gate and drain operation at 400 kHz, as well as the converter's output voltage.


Fig. 5 shows the schematic of the flyback converter using the LM5020 [7] from National Semiconductor. This circuit is virtually identical to the application given in the LM5020 datasheet. The converter efficiency was measured at 300 kHz and 500 kHz. The results are shown in Fig. 6. It can be seen that the 300kHz MOSFET based and eGaN based efficiency results are almost identical despite a 50% increase in RDS(ON) for the eGaN FET compared to the MOSFET. This can be accounted for by the eGaN having lower switching loss, as well as measurement accuracy and part-to-part variation in both the devices and magnetic components.

The key point to note here is, how much lower the drop in efficiency for the eGaN solution is when the switching frequency is increased to 500 kHz - about 0.5% for the eGaN solution vs. about 2% for the MOSFET solution. Waveforms in Fig. 7 shows eGaN FET gate and drain voltage at 500 kHz, as well as the converter's output voltage.


At higher power, PoE-PD converters can achieve higher efficiencies by using forward converters and synchronous rectification. A simplified forward converter schematic is shown in Fig. 8 based on Linear Technology's LT1952 [8] in conjunction with the manufacturer's LTC3900 [9] synchronous rectifier controller on the secondary side. The LTC3900 receives synchronization pulses from the primary side through an isolation /pulse transformer. Minimizing delay and on-off timings of the secondary side synchronous rectifiers is key to reducing body diode losses and improving overall efficiency. A MOSFET and an eGaN FET based converter were operated at 300 kHz and 500 kHz respectively to demonstrate the feasibility and advantages offered by the eGaN FET solution.

The turn-on and turn-off waveforms for both the eGaN FET and MOSFET are compared in Fig. 9. The significant difference between their respective gate rise times is due to a 10:1 ratio in respective input capacitance (CISS). The improved turn-on time of the eGaN is due to the five times lower Miller charge (QGD), while the reduced drain rise time is due to the eGaN FET having about half the output capacitance (COSS) as compared to the MOSFET.

Continue on next page

Fig. 10 shows the efficiency of both MOSFET and eGaN FET converters operating at 300 kHz and 500 kHz respectively. There are a few important points that can be made from evaluation of the efficiency results:

  1. At 300 kHz, eGaN FETs offer a small improvement in efficiency at all loads and have comparable efficiencies at full load despite a 50% increase in resistance for the eGaN FET in the primary socket when compare to the MOSFET RDS(ON).

  2. As in the previous cases, the eGaN FET converter efficiency performance relative to the MOSFET design significantly improves as frequency increases and is about 2% more efficient at 500kHz.

  3. The eGaN efficiency actually increases with increase in frequency. This is, however, almost entirely due to the use of two different output inductors for 300 kHz and 500 kHz versions chosen to maintain similar output current ripple between the two respective operating frequencies. This resulted in an approximately 5 mΩreduction in inductor DC resistance (DCR) and halving the inductor volume. Subtracting this inductor improvement, the efficiency between 300 kHz and 500 kHz versions remain largely unchanged.

It should be noted that due to the higher forward drop of the eGaN FET body diode, the gate driver timing was adjusted to reduce diode conduction using a single adjustment over the entire load range.

The table “Comparing MOSFETs and eGaN FETs” summarizes the performance parameters for both devices.


The above three examples of PoE converters show, to greater or lesser extent, the same result - despite a 50% increase in RDS(ON) in the primary socket, eGaN FETs will outperform their MOSFETs counterparts as the switching frequency is increased. This improvement is in most part due to the primary side switching device; as the synchronous rectifier (SR) for the forward converter is of limited benefit, due to both the absence of switching loss and increased diode conduction losses that are present without proper synchronous rectification timing. In most applications, efficiency improvement is all that is required.


The question becomes how can this efficiency improvement be converted into a cost advantage? To answer this question, consider a magnetic core with a specific cross-sectional core area and specific winding window area - this “core area product” number is commonly used to design magnetic structures [10] and can be directly related to the volume of the core. Even though the core cross-sectional area and winding windows areas may be different for different cores, a constant core area product results in similar losses and converter efficiencies for a given operating frequency.

What happens when the switching frequency increases? Without changing the magnetics, and assuming the higher switching frequency is still within the viable range of the core material used, the core losses will decrease at a higher rate than the frequency increases due to the non-linearity of these losses vs. flux density [11]. This effect can be seen in the eGaN FET forward converter results where the overall efficiency remains relatively unchanged with an increase in switching frequency even though the eGaN switching losses are increasing.

Continue on next page


Can the core size be reduced in exchange for most of this core loss improvement? In other words, can the initial MOSFET efficiency be maintained by using eGaN FET at a higher frequency and using a smaller / less expensive transformer?* Consider an example where the switching frequency is increased from 300 kHz to 500 kHz. The core cross-sectional area can be reduced to increase flux density back to its original value at 300 kHz resulting in 60% of the original cross-sectional area (~77% per side for a square core area). Cross-sectional views of two such cores are shown in Fig. 11 and yield the following results:

  1. Core volume has decreased to 60% of the original size.

  2. Core losses per unit volume may have increased (depends on material).

  3. Winding volume and mean length of turn has also reduced to about 85% - 90% (depends on the length to width ratio). This translates into a lower DC winding resistance and copper/conduction loss.

  4. AC winding resistance may have increased due to reduced skin depth (depends on wire thickness)

Overall, this will result in a more efficient transformer as typically (a) > (b) and (c) > (d). The “new” core area product as percentage would be roughly equal to the square root of the ratio of the lower frequency divided by the higher frequency, (about 77% in this example), but would be dependent on core aspect ratio. The new core volume would also be reduced to a similar percentage.

In high enough production volumes where material cost dominates the transformer cost, it would be fair to assume the cost of a transformer would decrease accordingly. Since the magnetic component tends to be the largest and most expensive single item, this savings can be substantial.

The eGaN FET's increased switching frequency further benefits designers. These include reduced output capacitance, the ability to use a smaller inductor in the case of the forward converter, as well as improved control bandwidth and dynamic and transient response.


  1. http://epc-co.com/epc/ToolsandDesignSupport/ProductTraining.aspx

  2. http://powerelectronics.com/power_semiconductors/power_mosfets/fom-useful-method-compare-201009/

  3. http://epc-co.com/epc/documents/product-training/Appnote_GaNfundamentals.pdf

  4. http://cds.linear.com/docs/Datasheet/1725fa.pdf

  5. http://www.fairchildsemi.com/ds/FD%2FFDS2582.pdf

  6. http://epc-co.com/epc/documents/datasheets/EPC1012_datasheet_final.pdf

  7. http://www.national.com/ds/LM/LM5020.pdf

  8. http://cds.linear.com/docs/Datasheet/19521fd.pdf

  9. http://cds.linear.com/docs/Datasheet/3900fa.pdf

  10. Colonel William T. McLyman, “Transformer and Inductor Design Handbook” CRC Press, 2004

  11. C. P. Steinmetz, “On the law of hysteresis,” AIEE Transactions, vol. 9, pp. 3ñ64, 1892, Reprinted under the title “A Steinmetz contribution to the ac power revolution,” introduction by J. E. Brittain, in Proceedings of the IEEE 72(2) 1984, pp. 196-221

*Transformer is used loosely here and refers to both transformers and coupled inductor magnetic structures.


Primary MOSFET FD52582 150 4.1 66 19 4.4 1254 290
eGaN FET EPC1012 200 3 100 1.9 0.9 190 90
S/R MOSFET SIR464 40 50 4.2 28.2 9 118 38
S/R eGaN FET EPC1015 40 33 4 11.5 2.2 46 9

Table: Comparing MOSFETs and eGaN FETs

Download the story in pdf format here.

Hide comments


  • Allowed HTML tags: <em> <strong> <blockquote> <br> <p>

Plain text

  • No HTML tags allowed.
  • Web page addresses and e-mail addresses turn into links automatically.
  • Lines and paragraphs break automatically.