Is it necessary to make PLL loop filter to be programmable?

[zhangljz]

Hello everyone,

I am designing a classical PLL with a 2nd order loop filter, the reference clock frequency is 10MHz, VCO frequency is 40MHz ~ 350MHz, control voltage is from 0 to 1.2V.

I have a question about the loop filter. I use a poly resistor for the filter, but as we know the resistance of the poly resistor is varying regarding to process, temperature, etc. Do I need to make a programmable resistor, like several resistors in parallel switched by several signals, so I can program the resistance to have a roughly constant value.

This is my first time to design a PLL, I just want to make sure it is stable and locked.


Any advice is appreciated

 

[BigBoss]

It's feasible.If you need more precise Loop Filter specs. you can switch the resistors between then to find the optimum one..
Meanwhile...
How will you implement large caps. on your circuit ?? Will they be off-chip ?? If it is so, why don't you implement PLL Loop Filter off-chip too in order to get more flexibility??

 

[dick_freebird]

For a do-all design that covers a range of fref and N, you
(as a product designer) would need to support appropriate
range on the other aspects like loop filter corner.

For a fixed divide and single fRef and particular phase noise
expectation, a fixed filter with enough PVT margin ought to
do.

But you need to ask whether you know what you think you
know, or know what you don't.

 

[zhangljz]

Hello freebird,

Thanks for your reply.

I checked the phase margin of the PLL which is always above 50deg, when I change the resistor from 8K to 12K(ideally it is 10K) for different output frequencies.And the bandwidth also changes, but they are almost below 1/15 of the reference frequency.

For other parameters, like settle time, phase noise, I don't have specific requirements. Maybe I can just use one single resistor to make the design easier.

Is it OK?

 

[D.A.(Tony)Stewart]

for clean signals with type 1 mixer a 2nd order filter is fine with fast capture range and acceptable jitter.
If you want faster capture and lower jitter or hamonics, then a dual BW approach is sound. with a switched filter upon lock detect, making sure not to induce an offset.
Continuously variable is also possible. but 2 stage is usually adequate but not always for servo systems.

Type 1 can fail to capture but has better jitter reduction if BW is too low.

for type 2 mixer phase/f type 2nd order filter determines lock time and jitter easily.. same

For variable low SNR signal input to PLL, gain drops with SNR so Loop stability degrades.

For plant controls and complex PLL 5th order filters are used with complex phase equalization.

So Programmable yes , but define specs 1st.