5

I understand that a capacitor can pass high frequencies and block low frequencies, because of its impedance properties.

Now, why do you need to put in series a resistor? Doesn't it already filter frequencies on its own?

For instance, when we use a decoupling capacitor, no resistor is needed.

Can somebody please clarify this?

Circuit fantasist
  • 13,593
  • 1
  • 17
  • 48
  • 1
    It may appear that "no resistor is needed" but actually, there is an impedance (resistive and / or inductive) formed in the power source wiring to where the decoupler is placed. – Andy aka Jan 06 '23 at 21:17
  • 5
    There are many ways and reasons for combining a resistor and capacitor. Maybe show some examples of circuits you don't understand, and we can explain them. – John D Jan 06 '23 at 21:29
  • In the \$s\$ domain the RC circuit is just a voltage divider. To operate you need two iimpedances. – copper.hat Jan 07 '23 at 07:27
  • 1
    @copper.hat, I would specify: "to operate **with voltages**, you need two impedances". That is why the output stages of amplifier circuits consist of two elements forming a "voltage divider". – Circuit fantasist Jan 07 '23 at 07:58

5 Answers5

8

schematic

simulate this circuit – Schematic created using CircuitLab

Figure 1. (a) All pass. (b) High pass.

I understand that a capacitor can pass high frequencies and block low frequencies, because of its impedance properties.

Alright, and so you're aware that the impedance decreases with increasing frequency.

Now, why do you need to put in series a resistor? Doesn't it already filter frequencies on its own?

If you want a high-pass filter then you need to create a frequency-sensitive voltage divider.

  • In Figure 1a C1 is feeding a buffer with very high input impedance. As a result there is no load on the capacitor and the right side will track the left side even to very low frequencies.
  • In Figure 1b C2 and R1 form a voltage (potential divider). C2 has low impedance at high frequencies so the divider effect will be very small. Let's say that R1 is 10 kΩ and C2 is 1 kΩ at the frequency of interest. The output voltage will be \$ \frac {10}{1+10} = 91\% \$ of the input voltage.
  • If we drop the frequency so that C2 now has an impedance of 10 kΩ then the output voltage will be \$ \frac {10}{10+10} = 50\% \$ of the input voltage.

If the input resistance of the buffer is suitable then you could omit R1 and still get the frequency sensitive voltage divider - a high-pass filter.

Transistor
  • 168,990
  • 12
  • 186
  • 385
  • 2
    I wouldn't call "a" all-pass. It most certainly blocks DC – Scott Seidman Jan 06 '23 at 21:48
  • 1
    @ScottSeidman Weird things happen when you go out to infinity. For any finite-period signal you can define an buffer input current small enough to pass that signal. (You may have difficulty building it.) – Matt S Jan 06 '23 at 21:53
  • 2
    @Scott Seidman, In my opinion, the capacitor in Fig. 1a does not block anything. First, it does not block the DC current because there is no current (it is "blocked" by the extremely high buffer input resistance). Second, it does not block the DC voltage since the voltage across it does not change (because there is no current flowing). – Circuit fantasist Jan 06 '23 at 22:31
  • 2
    @Circuitfantasist, the ideal capacitor has infinite impedance at DC. Replace the buffer with a 10 ohm resistor, and the impedance of the cap is still infinite, no? – Scott Seidman Jan 06 '23 at 22:53
  • 2
    @ScottSeidman, we all know that. The point I was making was just that if you apply a step change in level on the left side of C1 and leave it there then the right side of C1 will exhibit the same step change and due to the infinite input impedance of BUF1 it will never discharge. We all know that practical circuits don't have infinite input impedance and other effects such as bias currents will cause the voltage to drift. If you feel strongly about it then please address the points in your comments in your answer which seems to be discussing series resistance unlike this answer. – Transistor Jan 06 '23 at 23:07
  • @Transistor, in case (a), even if the current passes, the resulting voltage would be rather "DC" like, meaning it would kill the spikes of the high frequencies? Or at least, attenuate them significantly –  Jan 07 '23 at 13:59
  • @Charles34, no. The way I think of things is that the voltage across a capacitor will remain constant unless there is a discharge path. (Conversely the current through an inductor tends to remain constant.) Since there is no discharge path in BUF1's infinite impedance input there is nothing to discharge C1 so the voltage across it will remain constant so high frequencies will pass. In the practical circuit of 1b the low frequencies will be attenuated and high frequencies will not - the opposite of what you are suggesting. – Transistor Jan 07 '23 at 14:48
  • @Transistor, if there is no discharge path, then shouldn't the voltage of the capacitor build up until it reaches maximum and blocks all signals? –  Jan 07 '23 at 15:21
  • 1
    @Charles34, no. Imagine there is no charge on the capacitor and the left side is grounded. The voltage on the right-hand side is zero. Now connect the left side to +9 V. The right side will jump to +9 V and remain there. If you now superimpose an AC signal on the left side it will appear on the right. All the time (in this infinite input impedance example) the right side tracks the left side. – Transistor Jan 07 '23 at 16:33
  • @Transistor, I don't understand why both sides should have the same voltage, isn't there a voltage differential ? The capacitor I sucking up power as a resistor would, and left and right side of the resistor don't have the same voltages, if left side is 9V, and the resistor consumes 3V, then right side is 6V –  Jan 07 '23 at 16:54
  • Let us [continue this discussion in chat](https://chat.stackexchange.com/rooms/141881/discussion-between-transistor-and-charles34). – Transistor Jan 07 '23 at 16:57
4

If you are referring to for example a low pass filter, there needs to be a series resistor or inductor to form a "voltage divider". An ideal voltage source would otherwise charge the cap with high frequencies too, even if it's almost a short and the current required would be high.

In bypass caps the track to the device can be considered a series inductance. High frequency currents will flow the short distance from the capacitor to the device.

Ralph
  • 3,026
  • 1
  • 4
  • 22
3

There is no such thing as an ideal capacitor. There are a variety of models for a "real" capacitor, and the simplest of those is just an ideal capacitor in series with a small resistor.

So, there's at least a small resistor always present. The size of that small resistor is usually given on the data sheet for the family of capacitors you're using. Often, though, you don't care what that value is, because you're going to use your own resistor, which should be a few orders of magnitude bigger than the "effective series resistance" associated with the capacitor, and it's OK to ignore the ESR in many situations (warning: in some situations, the ESR can be critical).

Scott Seidman
  • 29,274
  • 4
  • 44
  • 109
  • 1
    Very useful practical explanations... But just to add that when we explain circuit ideas, principles, concepts... we use "ideal" elements... simply because the ideas are "ideal". In this case, the details get in the way of understanding the main thing... – Circuit fantasist Jan 06 '23 at 23:11
  • 1
    @Circuitfantasist it's pretty hard to understand bypass caps w/o analyzing the nonidealities (above and beyond the ESR), which is probably why so many think of them as something close to black magic. – Scott Seidman Jan 06 '23 at 23:16
  • I am not aware of this issue, I can only guess... – Circuit fantasist Jan 06 '23 at 23:21
  • 1
    @Circuitfantasist we used to have an expert post here fairly regularly about optimizing the selection of bypass caps to cover the appropriate frequencies for a given situation. He developed an online calculator. I wish I remember the user name. – Scott Seidman Jan 06 '23 at 23:25
3

I understand that a capacitor can pass high frequencies and block low frequencies, because of its impedance properties. Now, why do you need to put in series a resistor? Doesn't it already filter frequencies on its own?

The key is to understand the difference between an ideal voltage source in circuit theory and a real power supply in the real world.

Ideal Circuit

An ideal voltage source always keeps its output voltage the same by magic. No matter what is being connected across the ideal voltage source - resistors, capacitors, inductors, they have absolutely no effect on its output voltage. It cannot be "loaded down", even if it means the voltage source has to deliver millions of amperes to achieve this result. It's also why an ideal voltage source in circuit analysis cannot be short-circuited, or be connected to a different voltage source - it's essentially division by zero, an illegal move.

In ideal circuit theory, connecting a capacitor across an AC voltage source does absolutely nothing to its voltage, regardless of whether this voltage is signal or noise - an ideal source is an ideal source.

schematic

simulate this circuit – Schematic created using CircuitLab

Practical Circuit

A real circuit, like a battery or a power supply, always has a non-zero output resistance or reactance - it's the "hidden" R in RC. It's why, in a practical circuit, a capacitor can be connected directly across the power supply and works as a filter on its own. Because the existence of a resistor after the ideal voltage source, now its voltage changes depending on the load. The capacitor "loads down" the voltage from the power supply, thus creating a filter.

schematic

simulate this circuit

In other words, the large capacitor on the circuit board is only half of the filter, another hidden half is the internal resistance of the power supply itself. When they're combined, a voltage divider is created. In fact, a filter's performance can change dramatically under different source and load impedances.

Forgetting this fact can lead to practical problems - for example, many engineers learned it the hard way that a powerline EMI filter's noise attenuation won't be as good as its datasheet suggests, since it's measured with 50 Ω source and load in standard RF test instruments - a condition you will almost never find in a power supply!

So the bottom-line is, at the very least, R is here to enforce physical reality and to stop the math from breaking down.

To be even more realistic, the power supply is modeled as a voltage source followed by output resistance and wiring inductance. The capacitor is modeled as a ideal capacitor, with parasitic inductance and series resistance.

schematic

simulate this circuit

It is important to understand that decoupling is not the process of placing a capacitor adjacent to an IC to supply the transient switching current [...] rather it is the process of placing an L–C network adjacent to an IC to supply the transient switching current [...] All decoupling capacitors have inductance in series with them. Therefore, the decoupling network is a series resonant circuit.

  • Henry W. Ott - Electromagnetic Compatibility Engineering
比尔盖子
  • 6,064
  • 2
  • 24
  • 51
  • Very interesting... I feel like I wrote this nice story because it does not repeat well-known things about circuits but reveals the philosophy behind them... It gives me hope that in addition to concrete thinking people, there are also "circuit philosophers" here... – Circuit fantasist Jan 07 '23 at 18:43
1

Unlike the other "frequency" explanations, I will use the more intuitive "time" explanations.

V-to-I converter

The purpose of the resistor in an RC circuit is to "convert" the input voltage into output current; so it acts as a voltage-to-current converter (actually, the resistor does not convert but only determines the current according to Ohm's law).

I-to-V integrator

The single capacitor converts the constant input current into linearly changing output voltage; so it acts as a current-to-voltage integrator.

V-to-V integrator

The combination of the two cascaded converters converts the constant input voltage into exponentially changing output voltage; so it acts as a bad voltage-to-voltage integrator. It is interesting to see how we can make it perfect again...

Generalization

Besides in an RC circuit, we can see the same resistor in an R1R2 voltage divider, RD logarithmic converter, RLED voltage indicator, RBE transistor switch, RL differentiator, RA voltmeter made through an ammeter, etc. The resistor converts all these current-input devices into voltage-input devices.

You can follow the RC circuit evolution in detail in my other two answers:

Charging of capacitor in RC circuit

What is an integrator topology?

Circuit fantasist
  • 13,593
  • 1
  • 17
  • 48