Electronic – Bypass capacitors needed in low-frequency digital logic systems

You want to reduce the number of corners and straighten up some of your components.

1. The trace from R1 heading downwards wants moving a fraction left to remove the kink as it comes out of the resistor.

2. C2 can be moved up, and the two parallel traces from the negative pin can be combined into one.

3. D1 can move a fraction right to get rid of that little 90 degree bend as the trace from the 4561 pin 3 enters it.

I have marked them on your picture:

Some other thoughts:

• You may also want to have your corners at 45 degrees, not 90 degrees (mitre). I'm not entirely certain why people do this, but I for one think it looks nicer. You could of course break with all convention and use "Any Direction" lines, but I think that looks amateurish.

• You have a lot of wasted board space. Board costs money. The smaller your layout, the less board you use, the cheaper it costs.

• If you're thinking of etching this yourself, you may want to think about fattening up your traces a little to ensure they transfer properly and etch well.

Oh, and most importantly:

• Where are your inputs and outputs? I see the core of the circuit, but no Vcc, no GND, and no HIN/LIN/SD to the chip.

http://www.ti.com/lit/an/scba007a/scba007a.pdf

You'll see the big capacitor referred to a "bank" or "bulk" capacitors. The smaller ones are of course also "bypass" capacitors. The basic idea is that, in the real world, the parasitics of a capacitor aren't ideal. Your "bank" capacitor will help for transient power draw (changes in real current change) but, due to real world issues, if RF noise (EMI) gets on the line, the smaller bypass capacitor will let that noise short to ground before it gets to your IC. Additionally, both of these capacitors will be helping to suppress switching transients as well as improving intercircuit isolation.

Even though the physics is the same, the terminology is altered to their function. The "bank" capacitors "provide" a little extra charge (like a charge bank). The "bypass" ones allow the noise to bypass your IC without harming the signal. "Smoothing" capacitors reduce power supply ripple. "Decoupling" capacitors isolate two parts of a circuit.

So, in practice, you put a bank cap next to a bypass cap and there's your 10uF and 0.1uF. But two is just arbitrary. You have some RF on your board? Might need a 1nF cap, too.

A simple example of realworld impedance can be seen in this picture. An ideal cap would just be a large downward slope forever. However, smaller caps are better at higher frequencies in the real world. So, you stack TWO (or THREE, or HOWEVER MANY) next to each other to get the lowest total impedance.

I have, however, read dissenting opinions on this, saying that the self resonance between the two actually creates a HIGH impedance at certain frequencies and should be avoided, but that's for another question.