I would agree with Kleinstein on his comments. Ferrite beads between drivers and transformers as well as direct capacitive loading of the transformer secondary sides are probably not what you want.
With all that said, there are a couple things that are helping you here:
1. The LT1533 allows you adjustable slew rate to help control what is happening in terms of switching edges. Slowing the edges is a hit to efficiency, but it helps in controlling the high frequency that is so hard to clean out of a supply line.
2. Filtering can help prevent signal from getting to the LT3045 to begin with. Parasitic effects of components must be considered however.
3. The LT3045 gives excellent supply rejection as long as attention is paid to layout and an understanding of the coupling possibilities is understood.
The LT1533 gives better noise performance with the slower slew rate because you bring the switching edge down to a frequency that is easier to filter/reject. Switching edges of 1nsec are over 1GHz frequency content, while slowing that to 1usec now puts things in the MHz range. Higher frequency means higher energy content and easier transmission and coupling of the signal.
When it comes to adding ferrite beads and capacitors as a way to limit noise/signal that might reach the LT3045, one should understand what capacitors and inductors look like across frequency. Each component has parasitic factors that change the resultant impedance across frequency: capacitors have series resistance and inductance, inductors have series resistance and parallel capacitance. What this mean is that at some high frequency, capacitors will look like inductors and inductors will look like capacitors.
The way an LC filter works is it is an impedance divider; at low frequencies, the inductor is low impedance and the capacitor is high impedance, signal passes through. At the resonant frequency, these impedances are equal and the signal through it is 50%. Beyond the resonant frequency, the capacitor is lower impedance than the inductor and low amounts of signal get through; imagine it as a resistor divider where the values change across frequency. The issue at high enough frequencies is making sure the impedances are correct, and parasitic effects reverse things. Standard inductors look capacitive at high frequencies and the capacitors look inductive; the impedance divider is going the wrong way! A ferrite bead works well because it doesn't have the parasitic capacitance and it is high impedance at the frequency of concern (switching edge frequency). You want to couple that with a capacitor that has a low impedance at that frequency, which may be 100's of pF up to the uF range depending on switch edge transition speeds. MLCC's are nice because of their low series resistance.
As for the layout, this is what turns out to be the most critical aspect of using the LT3045. You need to think about where your AC currents will flow. The reason for the strips of copper to and from the LT3045 is to force the AC current to return directly under the feed. This minimizes the loop area and the field created by the loop. At the same time, the output device, load, output capacitor, SET pin capacitor, etc. all create minor loops except that these are orthogonal to the input loop. This severely reduces the coupling coefficient between the input and output.
Because of your slow switching edges and distance, you may get away without shielding. Again, higher frequency edges are harder to control because of parasitic effects, with a slow enough switch transition you get away from where those effects come into play.
Home
Help
Search
Login
Register
