Queston about pnoise of the charge pump
I don't understand why it has so large noise at 1MHz? When I change Fref, the position doesn't change.
something just looks wrong...
is there some other clock present in your circuit at 1 MHz?
Not sure what the units exactly mean, but 240 dB is a heck of a lot of dB. Maybe there is actually NO DATA at the lower frequencies, and the program is just trying to say "zero" at most of the frequencies, and you do have some analyzed noise at 1 MHz, and the program is showing the result at that one frequency. In other words, having the noise jumpr from 240 dB to 120 dB is like jumping from nothing to almost nothing, so it looks bigger on your graph than it really is. Maybe your noise sources are not uniform in frequency?
The other possibility, there is some sort of analog instability in the charge pump that makes it almost oscillate at 1 MHz, giving it a "theoretical" gain at 1 MHz that is very high.
There is only one clk, that is Fref.
The 1kHz noise current is about 1pA, but the 1MHz noise current is about 800nA. I find the main noise @1MHz comes from the digital circuits of PFD.
When I change Fref to 500kHz, the 1MHz noise is still there. But when I change Fref to 300kHz, the 1MHz noise dissappeard.
I agree with biff -- I hadn't looked at the scale (bad on me)
This is the noise current in pA. The noise current in low frequency is agree with my thought.
I've foud what the problem was.
When performing Pnoise analysis of a circuit that is mixing 1/f noise up to the carrier or its harmonics, you must be careful to place a cluster of frequencies near each harmonic to accurately resolve the noise peaks?but avoid putting frequency points precisely on the harmonics. If you do not follow these guidelines, then you run the risk of generating absurd noise totals because a very narrow noise peak artificially has its apparent width greatly magnified by a large frequency, and has it amplitude exagerated by placing a point precisely at the singularity. Choose the start and stop frequencies of the Pnoise analysis to avoid placing points precisely on the harmonics of the periodic drive signal, and add additional frequency points near the harmonics using the values parameter.