|
Magnetics Business & Technology Magazine |
|
CAE Tools Take On Magnetic
Design Challenge
|
|
By: Lawrence Meares
Figure 2: Transformer Design Screen provides trade off of more than 20 different design parameters. The software automatically flags design constraint violations. Flyback Magnetic Design Example Our first step is to determine the appropriate specifications for the
device. When flyback regulators operate in the discontinuous conduction
mode (DCM) using peak current sensing, they have excellent power line
noise rejection and the turn-on switching stress is small. Noise rejection
comes from the fact that the power delivered is 1/2*Lp*Ip2*F, which is
not dependant on line voltage, except at start-up when the regulator transitions
through the continuous conduction mode (CCM). Each switching cycle in
DCM begins with no energy stored in the flyback inductor so that it is
not necessary to commutate the rectifier diode; hence, reverse recovery
generated losses are not present. During the "on" time of the switch, S1, energy is stored in
the inductor. During its "off" time, energy is released to the
output. The solution of the equations shown in Figure 3 requires calculations
that are tedious and error prone. The tool we are using simplifies the
work by providing an SMPS wizard that converts the power supply requirements
into the data required for the flyback transformer design. The SMPS wizard
inputs, and the resultant spreadsheet entries, are shown next in Figure
4. The last geometry is the one that the program finally settled on. The total time to calculate full design specifications is under 1 second, thus encouraging the user to explore many different options. Using the Litz wire recommendation could reduce the increased AC resistance in the secondary; however, Litz wire is costly in manufacturing. The program default is to choose up to 2 strands per winding. These strands are considered to be planar; that is, they aren't twisted. Using multiple strands gives an effect somewhat like using Litz wire, but without incurring the added manufacturing cost. Changing the Max Strands constraint from 2 to 4 resulted in an improved design. Before accepting the design, a field solution will provide greater accuracy and will account for gap field induced eddy currents. To see these losses, the "apply using fields button" is pressed. The temperatures rise changes very little (about 0.5 Deg C.) and the winding stack shown in Figure 6 appears to be acceptable.
The primary is next to the gap and the field configuration shown is for
the secondary conducting with no primary current. The highest intensity
field is black so you can see the region just above the gap where the
gap-induced eddy current losses are highest. The field simulation used
here is not an FEA, but a special solution of Maxwell's equations first
described by Bennet and Larson[1] in 1940. Later on Dowel[2] and others
produced similar results, but with more complex formulations. The trick
is to mathematically replace each round wire turn with a rectangular wire
that is rotated in the field so that the boundary conditions of the solution
are satisfied. Once that is accomplished the Fr or ratio of AC/DC resistance
can be calculated for each turn. Removing the effective surface winding
that surrounds the core reveals a gap field winding that simulates gap-induced
fields. The whole process takes about a second to compute for this design.
That's too long for the thousands of trials that lead up to the design
creation; however, it's much faster than an FEA and the accuracy is acceptable.
The finished design summary is shown in Figure 7. Accuracy in magnetic models is overwhelmingly determined by manufacturing process controls. The wire can be stretched by as much as ½ a gauge from tension in the winding apparatus. The Insulation can be compressed considerably, affecting electrical and thermal properties and variations in the thermal path to the ambient environment. These factors can cause the predicted temperature rise to vary by an order of magnitude. While each of these parameters can be controlled in the design spreadsheet, it's up to the user to input the correct data for the manufacturing process. The need for a prototype cannot be eliminated if this data is inaccurate. One of the characteristics of high frequency designs using regular magnet wire (Heavy Formvar in this case) is that the window will be under filled to achieve the lowest overall power dissipation. That's because increasing the size of the wire becomes counter productive due to rising AC resistance. However, the requirement to keep the windings away from the air-gap mitigates this effect. At this point, one could explore the design further by splitting windings
or trying different materials. A winding sheet and design summary report
can be printed out or copied to Microsoft Excel.
Figure 10, Simulation results, Wizard predictions in parenthesis
Backgrounder:
|