LAN magnetics design procedure
The design procedure starts with the determination of what is needed for the electrical and mechanical requirements from the customer and the IEEE specifications.
The mechanical requirements are dictated by the customer’s motherboard real estate area and board separations. This determines the physical size of the connector that the magnetic components occupy. The pin positions also control the useable area for the magnetics within the package. All this has to be addressed to best meet the electrical and EMI specifications of IEEE and the customer, especially near end crosstalk (NEXT) and common mode rejection.
It is desirable to have a margin of about 0.050 inch distance from the buffered coil surface to the inside walls of the connector package. These constraints limit the size of the magnetics to be used.
All printed circuit boards used in the connector should have short run traces of low resistivity for the best EMI performance. The input to output trace placement should be carefully observed to avoid cross talk and common mode signal problems.
It is extremely important that the placement of the isolation capacitor and its traces be positioned such that the isolation voltage from the input to output of the transformer is maintained per the IEEE standards. Position of the wires to the magnetic coils in relation to the capacitor is important to both voltage isolation and EMI performance.
The transformer electrical design should have a “worst-case analysis” performed, including low temperature requirements to achieve the minimum inductance at the lowest temperature and DC bias conditions of the ferrite material being used.
After choosing a core size that will fit in the package, the number of turns is calculated to achieve the correct amount of minimum inductance on the core.
The optimum wire size that will allow both windings on the core to fit the ID in one layer is then calculated. The nearness of the wire to the core limits the loss of flux lines due to winding coupling. This will achieve the best leakage inductance other than twisting the wire for better coupling between the windings.
With these wires on the core, the maximum coil outside diameter and height with any buffer applied to the coil is then calculated.
The coil parasitic parameters that this winding produces are also calculated. These would be: leakage inductance, coupling capacitance, distributed capacitance and DC winding resistance. These can be calculated using the formulas listed in the following chapters, the formula glossary or the use of the spreadsheet on the CD.
The pulse waveform template is now addressed. Calculate the rise time, the low frequency pulse top droop, the fall time, the backswing and the recovery time. The fall time can be estimated to be the same as the rise time in most cases. These can be calculated using the formulas listed in the following chapters, the formula glossary or the use of the spreadsheet on the CD.
Now these answers are checked against the requirements of IEEE and the customer. If they do not conform to them, then there have to be adjustments made. These can be as simple as changing the percentage cover on the core, twisting the wire, changing the wire gauge or perhaps the wire insulation thickness.
The common mode choke is then designed to provide the maximum of attenuation for CMR in the frequency bandwidth of 30 to 1000 Mhz. Most of the noise problems will be observed at the harmonics of the clock frequency.
It would be very helpful and time saving if data for both impedance and CMR attenuation were available for various turns and wire gauges for several Nickel-Zinc ferrite materials. This data could be integrated into a spreadsheet so that quick calculations could be run arriving at a coil solution that could be wound and tested for optimum performance. If a certain frequency is known to be a problem for EMI, then the impedance of the core should be at a peak at or above this frequency. The reason that a point above this frequency is chosen is due to the dielectric effect of the buffering material. The impedance peak will shift downward in frequency after the coating.
Once a CM choke is chosen, then it is combined with the transformer and the pair is then tested for final CMR attenuation. When it is close to what is needed, the coils are coated with the buffer material and retested to see if the compensated dielectric change varied the frequency of peak attenuation to the problem frequency.
When the final design is achieved, then the coated coil sets are wound and assembled into the connector and tested for Insertion Loss, Return Loss, Common Mode Rejection and Near End Crosstalk (NEXT). The results are then checked against the IEEE requirements. Limiting the wire lengths between the transformer and the CMI choke and adjusting the coil orientation may vary the crosstalk and the common mode rejection.
When these specifications are achieved, then EMI is tested over the bandwidth both in an anechoic chamber and in an outside range using both 3 and 10-meter antennas. All these steps are shown on the following flow charts with the referenced chapters of formulas and examples.