Designers of networking equipment often request a model of the transformer and common-mode-choke module used in 10-/100Base-T applications. With that model, they can simulate the module's performance with the associated driver IC and other related components. Before building physical prototypes, the circuit can then be optimized. This guarantees that the final design will comply with the requirements of the chosen networking protocol.
In some cases, simulation results lead to changes in the design of the driver IC. But more often, they request that the magnetics designer modify the transformer and/or choke module to achieve the desired performance.
A typical model, which can be used for PSpice analysis, consists of two sections (Fig. 1). When Pulse customers request this model, it's supplied with parameters based on measured values, rather than the specified limits of the device. The values shown in the diagram are for a typical transformer module.
To properly apply this model, designers should understand how its various parts are derived. They also need to know how those parts affect the output waveform and the EMI noise entering the transformer. Although the model treats the transformer and common-mode choke separately, those two sections do interact. When they're combined, the parasitic elements of the common-mode choke enhance the transformer's performance.
The source impedances (RA and RB) represent the input impedance from the driver device. They're equal for each half of the differential-circuit operation. To be a critically damped circuit, the sum of RA and RB must equal the reflected load impedance. If they aren't equal, the return loss of the transformer will be affected.
The primary winding resistances (RP1 and RP2) are calculated by: wire resistance per inch × number of turns × winding length per turn around the core. This parameter is minimal. It affects the amount of insertion loss within the passband frequencies.
The leakage inductances of the primary (LLP1 and LLP2) and secondary (LLS1 and LLS2) are the winding-wire inductances for each half of both windings. For the best common-mode rejection and EMI, every winding half must be balanced with its other half.
When designing 10/100Base-T trans-former and common-mode-choke magnetics, the ultimate goal is to minimize leakage. This parameter drastically affects return loss at the higher frequencies (Fig. 2).
Low leakage inductance produces a return loss with a resonant dip. This dip allows the return loss to be greater at the higher frequencies. As the leakage inductance increases, that return loss becomes the rounded curve. It then has less return loss at the same frequency.
A shunt resistance (RCX) stands for the parallel-equivalent core loss across the transformer's primary side. This resistance, which represents the eddy-current losses in the core, increases with frequency. It has the greatest effect on the midband insertion loss and the pulse-backswing-voltage amplitude, better known as the flyback voltage. This voltage varies by the square of the number of turns on the core.
The distributed capacitances (CDP and CDS) are the turn-to-turn capacitances between wires of the same winding. When the secondary one, CDS, is reflected back to the primary side, that side multiplies its value by the inverse square of the turns ratio. CDS then becomes parallel with the primary distributed capacitance, CDP. This total capacitance is seen as a shunt capacitance by the driver chip.
Increasing distributed capacitance can improve the return loss of the transformer up to a point of diminishing returns. Exceeding this value will produce a deterioration of that loss (Fig. 3).
Lying between the primary and secondary windings of the transformer is the interwinding or coupling capacitance (CWX1 and CWX2). Such coupling capacitance impacts the high-frequency roll-off of the transformer. The higher the capacitance is, the lower the cutoff frequency will become. This parameter will affect return loss as well (Fig. 4).
The transformer inductance contains two components (L1A and L1B) on the primary side. Because the winding's inductance varies by the square of the turns, each half provides 25% of the total winding inductance:
L1A = L1/(N1A +N1B)2, or
L1/4 = L1A = L1B
where L1 is the total primary inductance, N1A is the number of turns on L1A, and N1B is the number of turns on L1B. The turns are on the same core. If L1A and L1B were on two separate cores, the inductance of the two windings added together would be two times that of each winding.
The secondary inductance is calculated by multiplying the primary turns by the turns ratio. For a 1:2 step-up transformer, the secondary has two times as many turns as the primary. Because they're on the same core, it boasts four times the inductance of the primary.
L2 = (2)2 × L1
where L2 is the total secondary winding inductance.
The primary inductance offers the greatest influence on the low-frequency roll-off. There has to be enough inductance to pass the low-frequency pulses, such as the link pulse. Meeting these requirements does demand an assurance that a sufficient amount of inductance will be left in the coil after a certain amount of bias current passes through the transformer.
Situated in tandem with the transformer, the common-mode choke provides high impedance to attenuate the EMI noise in the operating frequency range. That impedance is derived from the choke's inductance and core-loss resistance. The correct material must be chosen to achieve the core-loss characteristics and permeability needed to provide the right number of turns for the suppression.
The winding resistances (RCM1 and RCM2) have very little influence on the passband of the module's insertion loss. Again, these values represent the wire resistance of the winding as described for the transformer.
The choke's leakage inductances (LLC1 and LLC2) appear in series with the transformer leakage inductance. Their greatest effect can be seen in the deterioration of the return loss at the higher frequencies, as well as the slowing of the output waveform's rise time (Fig. 5). Leakage inductance should be kept to a minimum. However, the methods used to reduce it increase the coupling capacitance of the windings (CWC).
That coupling capacitance—the winding-to-winding capacitance of the common-mode choke—has its greatest impact on return loss. It appears in parallel with the distributed capacitance of the transformer. As it grows, the return loss at the higher frequencies improves to a point. Then, in the region of the resonant point, it starts to deteriorate again.
The distributed capacitances (CDC1 and CDC2) have very little effect on the waveform. Their greatest influence is seen in the stopband, where they tend to improve noise attenuation at higher frequencies.
In the core material, the parallel-equivalent core-loss resistances (RCA and RCB) represent eddy-current losses. The impedance produced by these losses peaks at a frequency determined by the number of turns on the core. The parallel-equivalent resistance and inductance both vary by the square of the turns used.
The distributed capacitance also has an effect on the impedance at higher frequencies. As a result, there is a point at which too many turns will eventually cause a problem at high frequencies, decreasing the amount of attenuation.
The primary inductances of the coil (LCM1 and LCM2) are determined by the equation:
L = permeability × N2 × core constant
where L is the open-circuit inductance, N is the number of turns, and the core constant is the core's cross-sectional area divided by its mean path length. For a ferrite material, permeability decreases with frequency, having its highest value at frequencies below 1 to 2 MHz. Above that point, permeability deteriorates with frequency until only the parasitic parameter of leakage inductance remains.
For the best return loss, the load resistances (RL1 and RL2) must match the source resistances based upon the turns ratio of the transformers. High return loss means no signal reflections back to the primary.
All of the components in this model have some direct impact on the desired waveform, as well as some indirect effects through their interactions. Naturally, some factors exert greater influence than others. Determining the component values that will achieve the best waveform is essentially a "balancing act," which encompasses not only transformer and/or choke design, but ultimately many aspects of system-level design.