The low impedance drive of the gate output structure drives the crystal back into the input impedance of the gate. This impedance is quite low due to the feedback resistor around the inverter. (510 ohms shown in the circuits that Mahmood posted). This will almost certainly overdrive the crystal.

Overdriving will cause aging because the crystal material is pushed beyond its natural response. You can see this effect very distinctly if you build a circuit like Mahmood showed versus one of the type like was at the top of the web page that I posted. In the inverter circuit design you can see on the scope that the wave shape that the inverter forces through the crystal is far from sinusoid. The natural response of the crystal itself would be sinusoid. On the other hand the transistor type curcuit, provided it is biased properly, will permit the crystals response to dictate the oscillation without overdrive abd on the scope you will see a very nice clean sinusoid.

I know that many will say that "I've see these dual inverter circuits for years" and I agree that you are right, you see them designed this way everywhere. Why? it is because digital engineers have designed the circuits. In many a case I have seen digital engineers design the circuits with the two inverters and say "Hey it works!". However they had no clue as to why an experienced RF engineer or oscillator engineer would have designed the circuit using the transistor or FET type circuit.

At one computer company I worked with in Southern California they had an 8085 computer system that was using a 2 inverter type oscillator with the crystal. That design used 330 ohm feed back resistors if I recall. Anyway, the product was experiencing eratic behavior after the computers were in the field for a period of time. We traced it to early failure of the crystals. Replacing with a Colpitts transistor circuit solved the problem. At that time the crystal frequency was 2.4576 MHz.

At another company a two inverter design was built into a product for a 32.768 KHz crystal. The product needed accurately produced 32.768 KHz frequency because it was divided to a 1 second reference pulse. We noticed that the supposed 1 second reference had three main problems. First it would not always start up oscillating. Second the frequency would not stabilize but keep shifting as the product was left powered up. Third the 1 second reference would not be the same measured value each day the product was powered up. We fixed these problems by replacing the dual inverter oscillator circuit with a transistor type Colpitts type circuit.

Michael Karas