Studies have shown that when the PCB power and ground planes are closely spaced they provide a low-impedance path for the switching currents in a Power Distribution Network (PDN) [1,2]. The power- and ground-plane pairs also impact the effectiveness of the decoupling capacitors. In this study, we investigate several PCB configurations in which we vary both the distance between the power and ground planes, as well as the placement and values of the decoupling capacitors.
In Part I and Part II (to appear in the next issue), we investigate two four-layer PCBs with power and ground plane pairs spaced 3 mils (Case A) and 30 mils (Case B) apart, respectively. In Part I, the boards are populated with four capacitors of the same value, at various distances from the measurement point. In Part II, we use 30 mil PCBs with multiple capacitors of the same value, as well as the capacitors with the values decades apart. The distance of the capacitors from the measurement point does not vary.
In Part I, we measure the impedance of the two boards for the following cases:
- Case 1 – Bare boards, no decoupling capacitors
- Case 2 – Four 1 nF caps at a short distance (1 inch) from the measurement point
- Case 3 – Four 1 nF caps at a medium distance (1.5 inch) from the measurement point
- Case 4 – Four 1 nF caps at a far distance (2 inches) from the measurement point
Figure 1 shows the two four-layer PCB stackups investigated in this study.
Figure 2 shows the details of the capacitor locations on a PCB [3].
Figure 3 summarizes the PCB configurations investigated in this study.
Measurement Results
Figure 4 shows the measurement setup used in this study.
Case 1A vs Case 1B: 3 mil-board vs 30 mil-board
No decoupling capacitors
Figure 5 shows the impedance measurement for Case 1A and Case 1B.
Observations: The impedance of the 3-mil PCB is lower than the impedance of the 30-mil PCB, over the entire frequency range, except for the small region around the resonant frequency of 282 MHz corresponding to the 30‑mil board. At 200 MHz the impedance of the 3-mil PCB is 25.4 dB (60.2 – 34.8 = 25.4 dB) lower than the impedance of the 30-mil PCB. The 3-mil PCB resonates at 228 MHz.
Case 1A vs Case 2A: 3 mil PCB
No caps vs. caps at close distance
Figure 6 shows the impedance measurements for Case 1A and Case 2A.
Observations: Addition of the capacitors (close to the measurement point) introduces a resonance at 105 MHz and an antiresonance at 138.5 MHz. In Region 1 (1 MHz – 122 MHz) Case 2A-PCB has a lower impedance (4.2 dB difference at 50 MHz). In Region 2 (122 MHz – 235 MHZ) Case 1A-PCB has a lower impedance. In Region 3 (235 MHz – 306 MHz) Case2A-PCB has a lower impedance again. Beyond the frequency of 306 MHz the capacitors have no impact on the impedance.
Case 1B vs Case 2B: 30 mil PCB
No caps vs. caps at close distance
Figure 7 shows the impedance measurements for Case 1B and Case 2B.
Observations: In Region 1 (1 MHz – 164 MHz) Case 2B-PCB has a lower impedance (17.6 dB difference at 50 MHz). In Region 2 (164 MHz – 331 MHZ) Case 1B-PCB has a lower impedance. In Region 3 (331 MHz – 496 MHz) Case2B-PCB has a lower impedance again. Beyond the frequency of 496 MHz the capacitors have no impact on the impedance.
Case 2A vs Case 3A: 3 mil PCB
Caps at a Close Distance vs. Caps at a Medium Distance
Figure 8 shows the impedance measurements for Case 2A and Case 3A.
Observations: On a 3-mil board, moving the capacitors further away (from 1 inch to 1.5 inch) from the measurement point has no impact on the board impedance.
Let’s see if the same holds for the 30-mil board.
Case 2B vs Case 3B: 30 mil PCB
Caps at a close distance vs. caps at a medium distance
Figure 9 shows the impedance measurements for Case 2B and Case 3B.
Observations: On a 30-mil board, moving the capacitors further away (from 1 inch to 1.5 inch) from the measurement point shifts the resonant and antiresonant frequencies as marked. Thus, moving the capacitors does impact the impedance but the effect is relatively small (there was virtually no effect for the 3-mil board).
Next, let’s move the capacitors even further away (2 inches) from the measurement point. Let’s investigate the 3-mil board first.
Case 2A vs Case 4A: 3 mil PCB
Caps at a close distance vs. caps at a far distance
Figure 10 shows the impedance measurements for Case 2A and Case 4A.
Observations: On a 3-mil board, moving the capacitors even further away (from 1 inch to 2 inches) from the measurement point still has no impact on the impedance.
Let’s see if the same holds for the 30-mil board.
Case 2B vs Case 4B: 30 mil PCB
Caps at a close distance vs. caps at a far distance
Figure 11 shows the impedance measurements for Case 2B and Case 4B.
Observations: On a 30-mil board, moving the capacitors even further away (from 1 inch to 2 inches) from the measurement point has a large impact (17.5 dB difference at 50 MHz) on the impedance. This result should look familiar (compare Figure 11 with Figure 7). Or better yet, let’s compare Case 1B (no capacitors) to Case 4B (capacitors 2 inches away). This is shown next.
Case 1B vs Case 4B: 30 mil PCB
No caps vs. caps at a far distance
Figure 12 shows the impedance measurements for Case 1B and Case 4B.
Observations: On a 30-mil board, the capacitors placed 2 inches away from the measurement point virtually have no impact on the board impedance! The impedance curve for the bare board with no capacitors is the same as the one with the capacitors 2 inches away from the measurement point.
When the capacitors were placed at 1 inch, or 1.5 inch away from the measurement point, they had large impact on lowering the board impedance. When placed too far from the measurement point (2 inches in this case) they have no impact!
Recall: On a 3-mil board adding the capacitors at a close distance had a relatively small impact (Figure 6);
moving them 1.5 inch away (Figure 8), or 2 inches away (Figure 10) had the same small impact. That is, the location of the capacitors did not matter, as long as they were up two 2 inches away.
On a 30 mil-board adding the capacitors at a close distance had a large impact (Figure 7); moving them 1.5 inch away had a similar large impact (Figure 9), but moving them 2 inches away (Figures 11 and 12) had no impact on the board impedance. That is, placing the capacitors too far (2 inches) renders them useless as they provide no benefit!
References
- Hubing, T– Effective Strategies for Choosing and Locating PCBs Decoupling Capacitors, Trans. IEEE EMC Symposium, pp. 632-637, 1995
- Hubing, Van Doren, Drewniak, and Wilhelm – An Experimental Investigation of 4-Layer PCB Decoupling, Trans. IEEE EMC Symposium, pp. 308-312, 1995
- Piper, S., Teune, J. – Practical Aspects of Embedded Capacitance for Printed Circuit Board Power Distribution Networks, 2015 IEEE EMC Symposium, Santa Clara, CA.
