Nahh don't split the path. You're making stubs. Just put the diodes and terminator in the same place, one straight route. The diodes have much higher impedances (for most of a cycle, anyway), their capacitance is dominant.
Don't bother with feedthru capacitors, just use pairs of small chip capacitors flanking the diode. Keep distance from signal to diode to capacitor as short as possible. You can safely place them close enough that their pads are nearly touching, leaving a minimum width soldermask sliver between them (~0.1mm).
If you're feeling really pedantic, use two pairs of flanking capacitors, with a ferrite bead (~100 ohm at 100MHz?) between them. Stitch the ground plane modestly in the area, isolating RF signal from the DC side.
This can all be one board, no need for two.
Nice part about 2-layer 50 ohm trace, the width is so big you can basically put all three pads across the trace and still have soldermask between them. Without widening the trace at all...
At worst, I suppose a tee coil peaking configuration could be used, but that wouldn't be needed until like ~GHz. Also would be hard to do on a PCB; and most of all, wouldn't be easy to realize a stable frequency response due to the C(V) dependence of junction capacitance. So it wouldn't be able to provide the ~double bandwidth that is possible from that configuration in a linear circuit, the worst-case improvement would be more modest. Anyway, there are <1pF RF diodes available to cover that range, without questionable hacks..
Tim
Home
Help
Search
Login
Register
