Dear all,

Imagine there's a 'classic' PLL (3-state PFD/CP, LC oscillator, integer N diivder) which locks nicely.
Let's assume the Kvco of the LC oscillator equals +100MHz/V, meaning the VCO frequency and the VCO tune voltage
are related by a positive derivative (~when the tune voltage increases, the VCO frequency will also increase).

Now assume I leave the whole PLL the same, but negate Kvco to -100MHz/V, meaning that now the VCO frequency will decrease when the tune voltage increases.

My question now is: would the PLL still lock? I would say no, since when e.g. the PFD/CP is giving 'UP' pulses to increases the VCO output frequency, the PLL will now actually do the opposite, namely lowering the VCO frequency.

Therefore, I would say the PLL is unstable and will divert to a condition where the LF voltage is clamped to one of the supply rails.

The problem now is however that I have a nearly-ideal Simulink model which proofs me wrong (PLL still locks when I negate the VCO gain).

Does anyone have any comments on this?

Kind Regards

Peter

Hello Peter,

some time ago I have made the same observations.
However, after change of the KVCO sign the PLL will never lock in the linear model (with phases as inputs). That is quite clear since – in this case – we will have a positive feedback with saturation.
But the nonlinear simulation model of the PLL (in my case with a multiplier PD) works properly for both signs.
However, the begin of the pull-in process will be different for both signs.
If the VCO frequency is lower than the reference (and KVCO positive) the pull-in process starts with an increase of the VCO frequency (with a positive control voltage).
In the other case (KVCO negative) the pull-in process first shows a decrease of the VCO frequency (with a negative control voltage).
But in both cases there is a similar beat frequency which slowly changes the VCO towards the reference frequency.
By the way, I like VISSIM for control system simulations.

Lutz    

Dear all,

I have indeed found a stupid bug in my Simulink model. Theory/reasoning and my model are now in sync, as they should be.

Kind Regards

Peter