Why capacitors are used with crystal oscillator?

          There are two kinds of crystal oscillators. One operates at what is called the "series resonance" of the crystal. This resonance is the frequency at which the (AC) impedance between the pins of the crystal is almost zero. The frequency is independent of how much capacitance happens to be in parallel with the crystal - its inside the oscillator and part of the circuit board, etc. But, even frequency that the oscillator runs at.

          The other kind of oscillator oscillates at "parallel resonance"of the crystal. At this frequency, the impedance from pin to pin of the crystal is almost infinite. This frequency depends on how much capacitance is connected in parallel with the crystal. This parallel capacitance is called "load capacitance". Generic signal-inverter oscillator is this kind of oscillator.

          The common oscillator connection is for the crystal to be connected from the inverter output to the input. And, there is a capacitor at each end of the crystal to ground. The NET load capacitance is SERIES equivalent value of those two capacitors.PLUS stray capacitance from the circuit board and the guts of the oscillator. Suppose that the crystal is rated for 22pF load capacitance. The stray capacitance is about 7pF. So, that leave 15pF to be made up from discrete external capacitors. If the external capacitors are equal, then their equivalent is half of their individual value. Thus, in this case, we would want a pair of 30pF capacitors.

          It should be made clear that the same crystal exhibits BOTH series and parallel resonances. If a crystal is ground so that it oscillates at 8.000MHz in oscillator that runs in series resonance,this crystal is called an "8.000MHz series resonant" crystal. But this same crystal, with maybe 22pF of parallel capacitance, might have a parallel resonance of 8.1MHz. Same crystal, different oscillator circuit. This SAME crystal could, if the manufacturer wanted, be sold as an 8.1MHz/22pF parallel resonance crystal.

• Most microprocessor oscillators run in parallel resonance mode.

Why single ended trace require impedance of 50 Ohm?

Three factors which is highly influence PCB trace impedance calculation

1.crosstalk: dramatically various with nearest ref plane.Cutting height by half reduces the crosstalk by 4 times.

2.The height of the trace above the nearest ground plane: It should be minimum. Less height means less radiation.

3.Less height means less radiation..it 'll reduce the impedance: which is less susceptible to capacitive loading.

• As per this three consideration PCB need to be thin as possible as to reduce distance between reference plan and signal plan.So we can not compromise with height of trace.What stops you from pressing height of trace down to zero is the fact that most chips can not comfortably drive impedance less than about 50 Ohm.

  EXCEPTIONAL: The old National BTL family drives 17 Ohm 

                             Rambus which drives 27 Ohm  

• Now impedance is inversely proportional to cross sectional area of trace.We have another parameter width of trace to maintain impedance of trace which is also not possible because all high speed boards are highly dense board.

• It not wise to use always 50 Ohm trace because an application like NMOS 8080, which works on low operating frequency(ex. 100 kHz) doesnt have Electromagnetic Interference and that type of application can't drive 50 Ohm.For this kind of processor you should use thinest and heighest impedance line to minimize operating power. 

• As  per mechanical view increasing impedance require decreasing in width of trace.The tiny lithography that high impedance trace require becomes difficult to fabricate.Where as 50 Ohm trace gives much wider trace to be manufacture.

As per above conditions we should use 50 Ohm impedance of single ended trace. Other differential pair traces have impedance other than 50 Ohm like USB having impedance of 90 Ohm!!!. 

.