3-4. Source impedance
Since the power supply and ground inside an electronic device are shared with various circuits, those can be easy paths for noise to flow out or enter into as shown in Fig. 3-4-1. In order to prevent noise conduction, a power supply filter is inserted as shown in Fig. 3-4-2(a). The effectiveness of the filter is expressed by insertion loss or S-parameters in the same way as the cases other than power supply.
At the same time, the power supply provides an electric current to the load circuit. A digital IC is connected as shown in Fig. 3-4-2(b) and changes the power supply current through its operation, and thereby noise is induced to the power supply, which in turn may interfere with the operation of the circuit itself. Hereinafter, this phenomenon is called as fluctuations in power supply voltage. The power supply filter is also required to have the effect of suppressing the fluctuations in power supply voltage.
In general, the effect of the filter to prevent noise conduction is not the same as the effect of suppressing the fluctuations in power supply voltage. The effect of suppressing the fluctuations in power supply voltage is expressed by source impedance. When the fluctuations in the power supply voltage flow out to the outside, noise will flow out as shown in Fig. 3-4-2(a). Although these two types of noise appear to be different, they are related to each other.
This section mainly describes the fluctuations in power supply voltage, and the source impedance with reference to digital circuit.
3-4-1. Fluctuations in power supply voltage
(1) Operation and source impedance of digital circuit
As described in Sections 2-3 of Chapter 2, a spike-like current occurs in the power supply and ground of a digital circuit in connection with the circuit operation. This current induces noise to the power supply fluctuating the power supply voltage and thus the circuit cannot provide stable operation. It also increases the likelihood of causing problems in terms of signal waveform and noise occurrence.
This function to prevent the fluctuations in power supply voltage is expressed by source impedance [Reference 5]
. Source impedance is one of the indices for power integrity and represents the impedance on the power supply side from the position where a digital IC (load) is connected (power terminal etc.) in Fig. 3-4-2(b).
(2) Influence of the fluctuations in power supply voltage
Fig. 3-4-3 shows an illustration that explains the influence on the noise of the entire device when the noise is induced to the power supply for a digital IC. The waveform of the IC power supply is shown in the center of the figure. As you can see the spike-like waveform, this is the noise that has been induced in connection with the operation of digital circuit. Here, the spike-like waveform is called fluctuations in power supply voltage. This effect interferes with the stable circuit operation and speed improvement, or spreads the noise to the power line or signal line, or deforms the signal waveform as shown in (1) to (4) in the figure. If the noise spreading to the power line is emitted from a cable, it will be a matter of noise regulations.
(3) Spectrum of power supply noise
The fluctuations in power supply voltage are originally from the current that flows at the rise and fall of the digital signal. Therefore, the noise in connection with the fluctuations in power supply voltage also has a discretely distributed spectrum just like the harmonics of the signal, if the circuit of the noise source is simple. Fig. 3-4-4 shows an example of experiment wherein the noise of the power supply for a digital IC that operates at 20MHz is emitted. The power supply voltage shows a spike at every 50ns (20MHz), and you can see that the noise spectrum is observed every 20MHz when this is emitted.
(4) Frequency characteristics of source impedance
In order to reduce the fluctuations in power supply voltage, the source impedance should be reduced. Since the impedance and voltage have a proportional relationship based on the Ohm's law, if the current that flows through the digital IC is constant, the voltage fluctuations become smaller as the source impedance is made smaller.
Fig. 3-4-5 shows an example of the measurement results of source impedance. Generally, the power supply is considered to be better to have smaller source impedance, which can provide a higher power supply performance and excellent noise reduction performance.
(5) Measurement of source impedance
Since the source impedance is very small, it is not easy to measure it. Fig. 3-4-5 shows results that were measured by a network analyzer. Since the value varies depending on the position of the measurement probe, it needs to be carefully measured at a defined position. It is usually measured between the power terminal and ground terminal of the IC (load). In order to improve the measurement accuracy excluding the effects of IC, you should temporarily take the IC away from the PCB and measure the impedance of the PCB side.
3-4-2. Decoupling capacitor
If the power supply circuit operates appropriately in accordance with the current of the load, the source impedance should ideally be zero. However, in the real world, the impedance increases gradually in the high frequency range above 10MHz as shown in Fig. 3-4-5, and in some cases it reaches as high as several tens of ohm.
(1) Decoupling capacitor
As shown in Fig. 3-4-6(a), since the wiring that connects between the power supply and load has an inductance and resistance, some impedance still appears from the load side even if the power supply operates ideally (0 ohm). Particularly in the high frequency range, the inductance in the wiring is the main cause of increasing the impedance.
In order to reduce the source impedance in the high frequency range, a capacitor is connected between the power supply and ground in proximity to the load as shown in Fig. 3-4-6(b). This capacitor is called decoupling capacitor, power supply bypass capacitor or simply bypass capacitor, etc.
(2) Absorption of the fluctuations in power supply voltage
Decoupling capacitor works as a temporal reservoir for electricity and absorbs the change in the load current, preventing the fluctuations in power supply voltage and noise occurrence. Since it is placed in proximity to the load, the influence from the impedance of wiring will be reduced. From the viewpoint of impedance, this operation means that the source impedance appears to be reduced.
However, even if the decoupling capacitor is used, a small wire still remains as shown in Fig. 3-4-6(b) and causes some impedance. Therefore, the capacitor should be placed in a manner to make this section as short as possible.
(3) Noise confinement effect
From the viewpoint of noise suppression, you may think that a decoupling capacitor confines the high frequency current generated in the power supply for load inside the section between the load and decoupling capacitor preventing the noise from spreading to the power line far away. Therefore, the decoupling capacitor is an important component not only for stabilizing the circuit operation but also for preventing noise occurrence. In order for more effectively preventing noise from spreading, a ferrite bead may be added as shown in Fig. 3-4-2(a), or a capacitor with a superior noise reduction performance such as a three-terminal capacitor may be used.
(4) Verification of the effectiveness of decoupling capacitor
Fig. 3-4-7 shows the change in the fluctuations in power supply voltage when a decoupling capacitor is used for the test circuit in Fig. 3-4-4. By attaching the capacitor, you can see that the voltage fluctuation margin has been reduced from 0.48V to 0.10V as well as reducing the noise emission by 10dB at the same time.
Fig. 3-4-8 shows a case of using a higher performance three-terminal capacitor. In comparison to the case of using common MLCC, the fluctuation margin of the power supply voltage has been reduced as well as significantly suppressing the noise emission. This is because the three-terminal capacitor has an advantageous structure particularly for noise reduction. Three-terminal capacitors will be further described in Chapter 6.
3-4-3. Loop impedance
(1) Frequency range of source impedance
The source impedance shown in Fig. 3-4-5, in fact shows an example of achieving an extremely low impedance with use of a plurality of decoupling capacitors. You can see that the frequency characteristics can be divided into 3 regions as shown in Fig. 3-4-9.
(2) Who is in charge of the low frequency range?
The relatively flat region  below 1MHz is a region where the output impedance of the power supply module is observed. If the decoupling capacitor is not used, the impedance would start increasing from a relatively low frequency as indicated by a dotted line in the figure. This is due to the output characteristics of the power supply module and/or the effect of the inductance in the wiring.
If a decoupling capacitor is used, the impedance of this high frequency range can be suppressed.
(3) Who is in charge of the high frequency range?
The regions in the relatively high frequency range indicated by (2) and (3) in Fig. 3-4-9 are the regions where the impedance of this decoupling capacitor is observed. Region (2) is a range where the capacitor has capacitive impedance and is controllable by increasing/decreasing the electrostatic capacity to some degree. Region (3) is a rage where the capacitor has inductive impedance. To further reduce the impedance of this region, you need to reduce the ESL of the decoupling capacitor, or the inductance of the wiring for attaching the capacitor.
(4) Loop impedance
The inductance of the wiring is composed of the patterns and vias that connect between the load IC and decoupling capacitor as an example shown in Fig. 3-4-10. The total inductance is derived by adding the sum of the entire current loop that goes through these components, to the ESL of the capacitor. Fig. 3-4-11 shows a representation of equivalent circuit.
The impedance of the current loop created by a decoupling capacitor may be called loop impedance. The loop impedance in Region (3) shown in Fig. 3-4-9 is caused by the inductance that is mainly derived from the wiring and capacitor.
In order to reduce the loop impedance in the high frequency range, the inductance needs to be reduced. That is to say, when the target value of loop impedance is Z Target
(ohm), the frequency is
(Hz), and the total inductance is L Loop
(H), the following formula is given:
For example, if you need to reduce the loop impedance to 1 ohm or less at 100MHz, the total inductance needs to be approx 1.6nH or less. This is an extremely small value.
(5) Elements of loop impedance
Since an actual circuit may have a wire branched off or may have a plurality of capacitors, you cannot think as simple as Fig. 3-4-10 and Fig. 3-4-11. However, this model is useful as a concept to break down the loop impedance into elements. In order to efficiently minimize the loop impedance, inductances that account for a large part of the total need to be reduced.
3-4-4. How to minimize loop impedance
In order to minimize the loop impedance in the high frequency range, both of the ESL of the capacitor and the inductance of the wiring should be reduced. If you can design skillfully, it is possible to reduce the total inductance to about several nH for a double-sided board or to 1nH or less for a multilayer board. The example in Fig. 3-4-9 indicates that it is about 0.3nH.
(1) Use a capacitor with small ESL
Among others, the ESL of each capacitor (in case of MLCC) is about 0.5nH, which takes up a large part of the total inductance. To reduce this, low ESL capacitors are effective and will be described in Chapter 6. The low ESL capacitors are also described in the Murata website.
Example of 3-terminal Capacitor Structure
(2) Reduce wiring inductance
To reduce the inductance in wiring and via, it should be "thickened and shortened." For example, capacitors and vias should be arranged in a manner to reduce the area of the current loop shown in Fig. 3-4-10. Furthermore, the pattern should be as wide as possible. Placing a capacitor right below the IC (on the other side of substrate) and thinning the substrate are often effective to make the current loop small.
(3) Parallel arrangement of capacitors and vias
When a plurality of vias and capacitors are used in parallel, the impedance can be reduced.
Since the inductance of wiring and via is very small and the mutual inductance is also involved, it is hard to get a simple estimation. Therefore, electromagnetic simulators are used for estimating such loop impedance. For your reference, Fig 3-4-12 indicates a general range of the inductance. However, the inductance can vary several times depending on the shape of the wiring. Furthermore, even if it is only a length of 1mm, it can cause an inductance of about 0.5nH, which is not negligible.
(4) Be aware of antiresonance
If two or more capacitors are used, you need to consider the resonance between capacitors. Generally, if you put capacitors with a different self-resonant frequency are connected in parallel with each other, antiresonance causes a frequency with high impedance (will be described in Chapter 6).
In addition, you also need to consider electrostatic capacity in the high frequency range above 100MHz apart from the wiring inductance. Furthermore, the resonance of the power source plane and the influence of IC package also become significant in the high frequency range. Electromagnetic simulators are also used for considering such complex elements.
3-4-5. Difference between source impedance and noise suppression
As described above, another important purpose for using a filter for the power supply is to block noise going in and out. Usually, this filter is composed of a capacitor and inductor, which form a low-pass filter. Fig. 3-4-2(b) shows a configuration of a typical filter for power supply. (The functions and configuration of filter will be further described in Chapter 3)
Although both capacitor and inductor are effective for noise reduction, these work in a different way in terms of suppressing the source impedance. Fig. 3-4-13 shows a case of T type filter in which the capacitor works for decreasing the impedance while the inductor works for increasing it. Therefore, in case of using a capacitor, if using a higher performance capacitor as shown in Fig. 3-4-7 and Fig. 3-4-8, you often have better results in suppressing the fluctuations in the power supply voltage and reducing emitted noise. However, if using a inductor, you need to be aware that you might increase the fluctuations in power supply voltage even though you can reduce the noise. Therefore, if using a inductor for the noise suppression of IC power terminal, it should be positioned at (b) instead of positioning at (a) in Fig. 3-4-13(a). In addition, the electrostatic capacity of the coupled capacitor should be sufficiently large.
3-4-6. Take noise measures on the noise paths
Even though the voltage fluctuations have been suppressed by reducing the source impedance with a decoupling capacitor, you might not be able to see a sufficient effectiveness in terms of noise reduction effect. Fig. 3-4-14 shows an example that simulates such a case in the previous test.
The positions of 2 capacitors (a) and (b) are shown in Fig. 3-4-14 ((a) is the same as the one with MLCC in Fig. 3-4-7). Both ICs have a decoupling capacitor at the position of 6mm from the power terminal and thus the loop impedance is considered to be the same. The fluctuations in power supply voltage are also within the same range. However, (b) emits 10dB more noise than (a) does.
The reason for this difference is that (a) uses the capacitor on the path that conducts the noise, while (b) uses the capacitor outside the noise path (between the IC and the antenna that emits noise). Therefore, the filter needs to be attached along the noise path to eliminate noise.
3-4-7. What to do if noise path is unknown
As shown in the test in Fig. 3-4-14, if you know where the noise path is in advance, it is easy to attach the capacitor at the position (a). However, noise paths are usually unknown. In some cases, a noise path exists on both sections of the wiring as shown in Fig. 3-4-15(a). In such a case, where is the best position to put the capacitor?
In this case, you can put the capacitor on both sections of the wiring as shown in Fig. 3-4-15(b) to confine the noise Although this technique requires more capacitors, it allows preclusively reducing the risk of noise interference as well as reducing the loop impedance due to the connection of the capacitors in parallel with the wiring on the left and right sides.
Alternatively, you can connect the power line after the capacitor as shown in Fig. 3-4-15(c). This method is effective for improving both source impedance and noise reduction, it is not enough to completely reduce the noise.
The method that provides the best performance is to connect the power line via a low ESL capacitor such as a three-terminal capacitor as shown in Fig. 3-4-15(d). You can expect the effectiveness on both source impedance and noise reduction.
In case of using a power plane for a multilayer board, it is advantageous for suppressing the source impedance due to its small wiring inductance. However, directly connecting the power terminal to the power plane makes it difficult to narrow down the path of the spreading noise and is thus disadvantageous for preventing the noise from flowing out. The method of connecting the source power supply (power plane) via a capacitor as shown in Fig. 3-4-15(c) and Fig. 3-4-15(d) is also applicable to multilayer boards and improves the noise suppression effect.
“3-4. Source impedance” - Key points
Source impedance is one of the indices for power integrity
Lower source impedance is desirable
A low source impedance suppresses the fluctuations in power supply voltage
It is effective for stable circuit operation, signal integrity and noise reduction
Use decoupling capacitors effectively for reducing the source impedance
Wiring design is also important in addition to capacitors