Simple Circuit Design Tutorial For PoE Applications
Simple Circuit Design Tutorial For PoE Applications
applications
Robert Mayell, Staff Applications Engineer, Power Integrations - April 12, 2006
As the number of Power Over Ethernet (PoE) powered device (PD) applications grows, market
pressures are driving designers to lower the cost and complexity of the DC-DC converters that
power them, while improving their performance. This paper provides simple circuit designs that
meet those goals. It presents a discrete PD interface circuit that has been shown to work in
University of New Hampshire Interoperability Consortium (UNH-IOC) tests. That circuit can be
implemented with only a few low-cost components. It gives practical design hints and tips on
designing a PD DC-DC converter stage. It also presents a galvanically isolated, Flyback DC-DC
converter example.
Lastly, it addresses designing PDs to operate from alternate power sources, such as AC power
adapters. Since PoE enabled PDs will become consumer commodity items, the simplicity, cost-
effectiveness and performance of the circuits presented in this paper make them attractive for most
PD applications.
The system on the right hand side uses an older switch that does not provide power, so the power
must be injected somewhere between the switch and the patch panel. Since the power is added at
the starting “end” of the enabled switch, it is referred to as an “endpoint” system, versus a midspan,
which injects the power somewhere in between the switch and the patch panel.
Figure 1. PSE power injection schemes: Endpoint (left) vs. Midspan (right)
Essentially, midspan PSEs are merely a way to allow PoE to be added to older systems without
replacing the switch hubs (with newer, PoE enabled units). The current specification only allows
power on two of the four wire pairs in the CAT-5 cable. As a general practice, Endpoint PSEs place
their power onto the data pair of wires in the CAT-5 cable (see Figure 2), while midspan PSEs are
restricted to using the spare pair of CAT-5 wires (see Figure 3 and 802.3af, pages 29 and 30). The
specification also allows three different power connection options (802.3af, page 31), which
determine the number and orientation of diodes found on a PD front-end.
Figure 2. Endpoint PoE configuration with PSE injecting power onto data pair
Figure 3. Midspan PoE configuration with PSE injecting power onto spare pair
The front-end
The front-end interface of powered devices
The input diode bridge (see Figure 9) protects the circuitry within the PD from being connected to a
reverse polarity input voltage. The first phase (detection) of the powering sequence occurs when the
PSE polls the newly connected device to see if it provides the correct impedance signature. The PSE
accomplishes this by ramping up a current limited (5 mA) detection voltage (from 2.5 V to 10 V)
across the designated pairs of CAT-5 wires (at about a 2 ms repetition rate) and measuring the
voltage and the current at the end of the ramp time. If the PSE detects the proper signature
impedance (802.3af, pages 38 and 39, Table 33-2), it determines that there is a valid PD at the end
of the link. The PSE then proceeds to the next step in the process -- Classification.
In the Classification step, the PSE applies more voltage (a 0 to 20.5 V ramp) to the link, and
measures the current drawn, in order to determine the classification (its power requirement;
802.3af, page 53, Table 33-10) of the PD. Once the PSE has classified the PD, it is supplied with full
operating voltage so that it can connect its pass switch and start up its DC-DC converter.
In-rush protection was required in some older, pre-PoE systems. However, in the newer systems, in-
rush current is so low that there is no need to have any current limiting components in the PD.
Under-voltage lockout (UVLO) is required due to the length of CAT-5 cable (100 M, or about 330 ft)
over which power must be supplied. The voltage drop across a 100 M cable can be up to 8 V when
the maximum available power (12.95 W) is drawn by a PD. The DC-DC converter’s UVLO circuit
must ensure that the voltage at the PD is high enough (≥42 V) that it will not drop below the
minimum working voltage (34 V) once load current is drawn from the cable. This prevents the DC-
DC converter within the PD from starting up and then shutting down repeatedly as power is initially
applied to the link.
Figure 5. The detection impedance resistor (R51) of a simple, Class 0, PD interface circuit
Classification: How much power will you be needing? During the classification phase, the PSE uses
the current drawn by the PD to determine how much power it will require. For Class 0 (the general,
default Class), the detection impedance draws the correct amount of current to classify the PD as a
Class 0. This makes the front end of the Class 0 PD the simplest and least expensive solution.
The circuit diagram of the Class 0 circuit (Figure 5) consists of six components that cost about 12
cents (at high volume manufacturing pricing) in materials. A less expensive (but lower efficiency)
bipolar transistor based pass switch circuit requires 3 additional parts (to boost the drive into the
base of the transistor) but costs only about nine cents in materials.
Figure 6. Detection, Classification and Pass-Switch interface circuit for Classes 1–3
Identifying a PD as a Class 1, 2 or 3 requires a few additional parts, as can be seen in Figure 6. The
10 components (VR1, Q1, Q2, Q3, Q4, R2, R3, R4, R5 and U1) that are located between the signature
impedance resistor and the MOSFET pass-switch connection circuit only conduct current once the
PSE’s probing voltage exceeds the Zener voltage (11 V) of VR1. The PSE ramps the voltage across
the link from 0 to 20.5 V. It measures the current drawn by the PD while the voltage is ramping from
14.5 to 20.5 V. During this phase of the ramp, the PSE expects to measure a current within the
range of 9 and 30 mA to detect a valid PD of Class 1, 2 or 3 (see the table in Figure 6).
The classification circuit works as follows: Transistors Q1, Q2, R2 and R3 form a 350 μA current
source (set by the value of R3) that biases U1. Transistor Q3 and U1 form the current source that
sets the classification current, based on the value of R4. By changing the value of R4, the circuit will
conduct the right amount of classification current for a Class 1, 2 or 3 device. Once classification has
occurred and the pass-switch is connected, R5 biases Q4 on, which shunts the 350 μA current source
to ground. This disables U1 and Q3 to minimize the power consumption of the front-end once the
DC-DC converter begins operating.
Description Comments
Flyback Lowest cost for output currents < 6="" a="">
Typically used for output voltages (> 2.5 V) Recommended for output voltages (>
When to Use
12 V) For applications requiring lowest cost
Advantages No output inductor required. Output isolated from input.
Higher output ripple current (higher output capacitor cost) Medium efficiency
Disadvantages
(due to higher peak and rms currents)
Forward Lowest cost for output currents > 6 A
Typically used for output voltages (< 12="" v)="" recommended="" for=""
When to Use output="" voltages="">< 2.5="" v)="" for="" applications="" requiring=""
highest="" efficiency="">
Low output ripple current (lower output capacitor cost) Higher efficiency (due to
Advantages lower peak and rms currents) Can utilize synchronous rectification. Isolated
output.
Disadvantages Requires output energy storage inductor (higher cost)
Buck Usable if isolation not required and output current < 3="" a="">
When to Use Low power, one output, and low efficiency is acceptable
Advantages Simple, low parts count. Uses off-the-shelf inductors
Variable switching frequency required for low output voltages (< 12="" v).=""
Disadvantages one="" output="" only="" (multiple="" outputs="" not="" recommended).=""
switch="" carries="" full="" output="">
Since most PD applications require input voltages and currents that can efficiently and cost
effectively be delivered by Flyback converters, this section will focus on the operation and design of
Flyback converters.
For a single switching cycle of a Flyback converter’s normal operation: the controller turns the
MOSFET switch on, creating a path for current (ID) to flow through the primary winding of the
transformer and the MOSFET. Due to the phasing of the transformer primary and secondary
windings and the output diode orientation, no current flows in the secondary while the switch is on,
and energy is stored in the core of the transformer. The current increases linearly until the
controller turns the switch off. As the switch turns off, the voltage across the transformer windings
reverse, which forward biases the output diode. The energy that was stored in the transformer core
is then transferred to the output capacitor (which is connected across the secondary winding) as
current (IS) flows through the output diode.
The voltage rating of the Zener diode is chosen so that it does not conduct during normal operation,
but only when a line surge or load transient induces a larger-than-normal spike. With a Zener clamp,
the converter usually operates more efficiently. An RCD network is used when EMI noise generation
is a concern, since it clamps the Flyback voltage spike each time the switch turns off. The diode in
the RCD clamp should have about a 500 ns recovery time, so that some of the energy stored in the
leakage inductance can be recycled as it recovers. A series resistor limits the reverse current
through the diode. The values of the clamp resistor and capacitor are typically obtained through an
iterative process of taking measurements and adjusting values until the spike is sufficiently clamped
while the temperature rise of both components is kept to a minimum. This is required because the
values of resistance and capacitance will primarily be determined by the parasitic capacitances and
leakage inductance of the transformer, which cannot be easily predicted before a number of
transformer samples have been built, tested and characterized.
Flyback transformer core selection guidelines The following table lists the transformer cores most
commonly used in low-power, DC-DC Flyback converters.
Description Comments
EFD EFD10, EFD12 Low profile. Rectangular center leg, hence higher mean length per turn
EFD15, EFD20 (MLT). Terminating multi-filar windings to EFD bobbin pins is difficult.
Low profile. Round center leg, hence lower MLT. Large center leg gives
ER ER9.5, ER11,
gives high inductance with few turns, hence lower leakage inductance.
ER14.5
High pin count bobbins make winding termination easy.
Large center leg. Terminating multi-filar windings to RM bobbin pins is
RM RM6, RM10 difficult. RM bobbins have little room for meeting safety spacing
requirements. Cores are expensive.
Large center leg. Terminating multi-filar windings to PR bobbin pins is
PR PR14x8 difficult. PR bobbins have little room for meeting safety spacing
requirements. Cores are expensive.
Planar ELP18/10/4, E series has a square center leg, hence high MLT. ER series has a round
E18, E22, ER18, ER23 center leg, hence lower MLT. Not cost effective at PoE power levels.
Table 2: Flyback Transformer Core Selection Guidelines
Flyback transformer construction guidelines The following table lists the most important aspects of
low-power, DC-DC, Flyback transformer design.
Description Comments
Core Flux Maximum Flux Density (BM), measured in Gauss or Tesla
BM (higher efficiency) 1500 < BM < 2500 (smaller core size)
Core Gap Gap length (LG), typically measured in millimeters
LG LG > 0.1 mm (the minimum, reliable gap size)
Windings How many turns and layers (of turns) per winding
Number of turns Few turns limits leakage inductance and conduction losses
Split windings Reduces leakage inductance
Multi-filar ≤4 Limited by the number of termination pins. Reduces Skin Effect
Copper Foil Use only if > 4 strands of multi-filar wire are required
Flyback output diode selection guidelines The following table lists the most commonly used output
rectifier diodes for low-power, DC-DC, Flyback converter.
Description Comments
Low-Drop Schottky For low output voltage and high efficiency applications
Low forward voltage drop: 0.44 VFORWARD, rated for 1 A, 2 A and 4 A, and 20,
SL1x, SL2x, SL4x
30 and 40 V of reverse voltage
Schottky Reliable Schottky diodes with VREVERSE up to 100 V
SS1x, SS2x Forward voltage drop range of 0.5 V to 0.75 V
Ultra Fast For high voltage output (> 12 V) applications
ES2x 20 ns recovery time, 50 V to 200 V reverse voltage
BYG22x 35 ns recovery time, 50 V to 200 V reverse voltage
Description Comments
A common choice for DC-DC Flyback converters. Low ESR and high
Tantalum capacitance values. Require significant voltage derating (30–50 %).
Short circuit failure mode.
T491 & T495 series, TE-L
Kemet and Panasonic, respectively
series
Super low ESR. Expensive, but coming down. Require no voltage
Aluminum Organic
derating. Open circuit failure mode.
A700 series, 6PTB series Kemet and United Chemicon, respectively
Very low ESR, but usually low capacitance also. Low ESR can cause
Ceramic
loop stability problems.
A good fit for most applications, but expensive. Not as reliable as AO
SMD Electrolytic
or ceramic. No loop stability problems.
Tantalum capacitors are most commonly used. SMD Electrolytics may provide a good cost versus
reliability compromise--if space allows--until the cost of Aluminum Organic capacitors comes down
further. Ceramics can be used, but loop compensation must be given careful attention to ensure
stable operation.
Figure 8. A typical TL431 circuit’s resistor divider & frequency compensation components
First, it is imperative that no power handling components get so hot that they fail, or cause other
components or circuits to operate incorrectly, even during abnormal conditions. Second, the
designer needs to verify that the peak drain voltage does not exceed the maximum voltage rating of
the MOSFET switch. This usually occurs when the converter is overloaded while operating at the
highest input voltage (57 V), just before it loses regulation of the output and goes into self-protection
mode. Third, the designer must also verify that the drain current waveform does not show any
indication of transformer saturation (decrease in inductance near the peak of the primary current
ramp), especially at start-up and high line overloads. Practical dc-dc converter design
A practical dc-dc converter design example
Figure 9 is the circuit diagram of a typical PD converter. The PoE front-end identifies the PD as a
Class 2 device. The DC-DC converter is a Flyback that uses a Zener diode clamp (VR3) on the
MOSFET drain-node. It uses a 20 V, 4 A, standard Schottky diode (D1) for its output rectifier, and
tantalum output capacitors (C7–C9). The frequency compensated (R11, C13) TL431 (U3) drives
feedback through the opto-coupler (U2). Lastly, a Y capacitor and damping resistor (C4 and R7)
attenuate common-mode EMI noise.
The unpowered CAT-5 cable plugs into the input (lower) jack (J1), and the injected 48 V power is
placed on the unused (Alternative B) pair of wires that connect to the output (upper) jack (J2). This
maintains the current midspan PSE convention, and prevents the 48 V from being connected to the
unpowered switch or router. This solution does not include the detection and classification functions
specified for a PSE by 802.3af, but that does not matter, since it is a temporary solution that only
powers a single port.
Figure 10. The circuit diagram of a single port, midspan ( 21.7 W) PSE
References:
1. IEEE 802.3af Part 3: Carrier Sense Multiple Access with Collision Detection (CSMA/CD) Access
Method and Physical Layer Specifications Amendment: Data Terminal Equipment (DTE) Power via
Media Dependent Interface (MDI).
2. Application Note AN-31: DPA-Switch DC-DC Forward Converter Design Guide, April, 2003, Power
Integrations.
3. Application Note AN-32: TOPSwitch-GX Flyback Design Methodology, December, 2002, Power
Integrations.
4. DPA-Switch Family datasheet:, July 2005, Power Integrations.
5. Engineering Prototype Report (EPR)–86: 6.6 W, Multi-Class Powered Device (PD) for Power over
Ethernet (PoE) Using DPA423G, February 2006, Power Integrations.
6. Design Idea DI-70: DPA-Switch PoE Detection and Classification (Class 0) Interface Circuit,
November 2005, Power Integrations. 7. Design Idea DI-88: DPA-Switch PoE Detection and
Classification (Class 0-3) Interface Circuit, November 2005, Power Integrations.
8. Design Idea DI-101: DPA-Switch Under-Voltage with Wide Hysteresis, November 2005, Power
Integrations.