Peaks in PDN frequency-impedance profile

[matrixofdynamism]

The PDN is dominated by different parts of the PCB at different frequency ranges. This includes the bulk capacitors, decoupling capacitors, PCB plane capacitance, parasitic inductance and capacitance on the PCB, capacitors inside the IC packages (if they exist) that are closest to the die and thus effective at the highest frequency ranges.

The combination of all these gives us a curve that could look something like the attached image.

From what I know, we would like a flattened PDN profile. This means that the peaks must not be very sharp. There is also the concept of target impedance that I am trying to understand. The question here is, what problem do sharp peaks in impedance profile create? Why is this a problem at all that we need to flatten the profile? This could be done by adding some resistance possibly via controlled ESR capacitors but I am not sure about other methods.

[EEEnthusiast]

Sharp peaks are signs of resonance. If your current waveform has the same frequency content at which the PDN is resonant, then your voltage ripple can become very high. If the current waveform is periodic with any harmonic content falling within the peak, then the voltage will keep building up and reach a point where your circuit can malfunction. It is like a resonant circuit which gets periodic supply of energy which is in phase with its oscillation. Sharper the peak, higher the Q and lower the damping. This means increased peak voltage drops.

[matrixofdynamism]

If I see a graph that shows V-I characteristics, the gradient of the curve shall represent resistance (assuming V is on Y-axis). If there are peaks in this graph, I can see that at some places if we have AC signal, a small change in current shall lead to a change in voltage. This makes sense.

I am not able to relate an impedance-frequency profile with the concept of large voltage drops across a component. How is this link made mathematically?

[joeqsmith]

The question here is, what problem do sharp peaks in impedance profile create? Why is this a problem at all that we need to flatten the profile?

Emissions

There are lots of very good videos on this subject.   Some of the better ones I have watched were from Steve Sandler and Eric Bogatin.   

I attempted to use one of those $50 VNAs to make some PDN measurements and work some examples.   

[jmw]

I am not able to relate an impedance-frequency profile with the concept of large voltage drops across a component. How is this link made mathematically?

Ohm's law: \(V=I Z\)

The load attached to the PDN appears in series with the PDN's impedance and forms a voltage divider. Any transient current (noise pickup, surge demand from component) passing through the PDN's impedance causes a voltage drop across the PDN that is seen by the load.