Wideband decoupling switching load (DC-GHz range)

[The_Dutchman]

Hello all,

For a wideband load I'm trying to design the power supply decoupling. The load pulls about 200mA DC current from my power supply with about 100mA of amplitude peak current at 3V supply. This peak current is at frequencies from about DC to 2GHz (multi-carrier signal) but for now I'm simulating with single-tone sine over the mentioned frequency range. I want to keep this DC supply as clean as possible.

I'm thinking about using the ADA4870 which is a low-noise unity gain stable power opamp. I will use it in a +1 gain configuration, with my reference DC voltage voltage so the output tracks the set voltage.
However the bandwidth of the OPAMP is limited to about 70MHz and also the output impedance increases from there on. So the opamp cannot supply high frequency load currents so they have to come from the bypass capacitors. These bypass capacitors also have the ESL and ESR parasitics which limit the frequency response so I have to put multiple sizes in parallel. However capacitors with high self resonance frequency come at the cost of low capacity which have to little charge to supply the currents.

I'm looking for answers that can put me in the right direction regarding calculations of how I should tackle/calculate this combined problem. Can you put me in the right direction on how to calculate how much capacitance I need taking into account the frequency limits of the capacitors? Or any other suggestions?

Thanks!

 

[FvM]

However the bandwidth of the OPAMP is limited to about 70MHz and also the output impedance increases from there on.

A more appropriate description would say, the output impedance is continuously rising beyond 1 MHz. And not explicitely specified in the datasheet, but obvious by circuit analysis basics, mostly inductive. Bypassing the output with a capacitor without damping (e.g. series resistors) causes resonances and possibly instability.

Allowing a certain amount of real impedance for your wideband source would greatly ease the design.

 

[chuckey]

Even if your power supply had zero ohms output impedance at all frequencies, the wiring between it and the multifrequency amp will have its own impedance and resonances. So depending on the wiring length, it would be best to decouple properly as possible at the load end, so just keeping the power supply to deal with ripple at less the 10 KHZ.
Frank