I was reading the sticky post about bypass caps, trying to understand their purpose:

The reason for them is "stiffening" of the power supply. The wires on a breadboard, extraneous bus lines on a solderless breadboard, etc all create parasitic resistance and inductance. This resistance and inductance (impedance) of the board resists current flowing from the power supply to the microcontroller or logic IC.

Here is where the bypass capacitor acts like a mini power supply to "hold up" the voltage until the current overcomes the impedance of the power supply lines to feed the IC.

I'm still not totally clear on why it is needed. So there is some impedance in the circuit. Why is that detrimental to the circuit? What is the danger here?

 

The danger is that if the amount of current drawn by the circuit changes, so will the supply voltage(s). This is generally undesirable: it may cause badly degraded performance. In the case of circuits capable of amplifying it may lead to a situation called oscillation. What can happen is that a change in current leads to a change of voltage, which leads to a change of current...and so on indefinitely.

Various other kinds of faulty operation are possible, such as poor frequency responses, bad pulse shapes, unwanted signal paths between different circuits sharing a power supply, bit errors in digital circuits.

I can assure you that decoupling capacitors have been necessary since the days of grandfather's valve (tube) radio. They are still necessary in modern digital equipment, and lack of or failure of these components is a common source of malfunctions.

 

So there is some impedance in the circuit. Why is that detrimental to the circuit? What is the danger here?

It can be detrimental to a circuit that has switching components and needs to provide current for e.g. power mosfets. A static circuit with constant voltage and current levels surely wouldn't have problem with some impedance caused by a trace's inductance, an dynamic circuit does.

Imagine a voltage source. What would you expect this source's internal resistance to be? As low as possible, right.

Now imagine you have a voltage source, an inductance (traces) in series and then a circuit. Now you try to pass a rapidly changing current through the trace. You can't. The trace's inductance is going to limit the rate of change of current that flows through it. If you pass a rapidly changing current through the trace you will a voltage drop on it, right? (we are talking about ns here) It's an inductor.

This can lead to unpredictable behaviour in circuit components and it also distributes the switching noise all over the place, possibly negatively influencing other circuits.

That's why you put the BP cap as near as possible to the component that draws current from it.

 

Digital chips, when they switch (and it is their job after all) generate spikes in the power supply. The bypass capacitors "absorb" these spikes, preventing them from affecting other components. Without the capacitors, these spikes can be interpreted as a digital signal. This is one application.

 

OK, so if I understand correctly, they act like a buffer, and smooth out the voltage spikes/drops. And the reason it must be NP is because the signal is oscillating, right?

So, the 0.1uF cap does that? It just seems like such a small capacity. Just out of curiousity, would a larger cap work just as well for that purpose, or would their be other side-effects?

Thanks
Mossen

 

Unfortunately, perfect pure capacitors are impossible to manufacture: they inevitably have parasitic inductance and resistance, so more is not necessarily better. 100nF is commonly used for a lot of applications to things like moderate speed logic circuits, but some circuits may need much higher capacitance. In such cases, more than one one capacitor may be needed to meet the requirement while still maintaining a low parasitic inductance and resistance. This is not without its pitfalls, as multiple capacitors connected by inductive wiring or PCB traces may lead to resonances.

Decoupling capacitors are not necessarily NP. In fact, the larger values of decoupling capacitors are often polarised electrolytic types.

 

As Adjuster said, nothing is perfect. This is a good (but not perfect) solution to a very common problem.

Many of my projects I don't bother with the bypass caps. If I have problems I can always add them later.