Any experience in design 1uA 32.786KHz crystal oscillator?

[dwayne22]

Re: Any experience in design 1uA 32.786KHz crystal oscillato

The simulator's algorithm can actually act to 'damp' the oscillation. The Gear integration algorithm will do this. I suspect that numerical roundoff errors might also. Trapezoidal rule integration routines are under-damped, which can cause oscillators to work in simulation which won't in real life.

So, transient analysis is not only time-consuming, but doesn't help a whole lot.

That's why using an accurate loop gain (AC analysis) method is important. In oscillators, the loop gain assumptions can be violated, so the simple method is insufficient. Worse, it usually overestimates the gain!

The Middlebrook method is accurate and has worked for me ...

Also, AC analysis is fast, so doing a bunch of corners in no problem.

 

[offer12]

Re: Any experience in design 1uA 32.786KHz crystal oscillato

Thanks for this answer it helps.
could u tell me what u consider as 'good' marginal gain from the AC loop analysis? again with tran simulation I get oscillation even with gain<0db..
My Goal here is to reduce power, according to my spec it should be lower than 2uA (0.18 process), as I c it, one traoff is btw the loop gain and the power.
thanks again, offer