trobled with the system parameter design of fractional-pll

[jasonxilion]

we wanna design a fractional pll with ring oscillator.
the input frequency is from 5M to 50M, and we set the bandwith to be 50khz, the segma-delta modulator is 3th order. the current ICP is varying with the N which varies from 1 to 128

the troble we meet is by calculation the Cs in the LPF is more than 1uF!, if we want the phase margin more than 50

the other problem is the variation of bandwith is very big according the variation of N. we can't ensure that the system be stable at worst case (taking corner into account)

those are really key issues in pll design. anyone discuss this topic with me?

 

[aProgrammer]

if N varies from 1 to 128,
the loop gain also varies from 1x to 128x,

so it's a problem,

I suggest you to use a VCO
with a 25M to 50M range,
and with dividers by 2,
you could reach 5M.
(with one divider you cover
12.5M to 25M,
and with 2 dividers, 6.25M to 12.5M
...)

 

[jasonxilion]

Re: trobled with the system parameter design of fractional-p

aProgrammer:
thx for ur suggestion
but the spec requires N varies from 1 to 128. we tried to vary Icp with N, but it also bring a lot of problems.
i am really troubled

 

[RFDave]

Re: trobled with the system parameter design of fractional-p

Design your VCO so that you can adjust the VCO Gain as N varies. That way, you can keep you loop gain more constant that with a 1:128 ratio in N

Dave
www.keystoneradio.com

 

[jasonxilion]

Re: trobled with the system parameter design of fractional-p

the bandwith variation is just one of my problems.
the huge area requred by the cap in the lpf also makes me headache. since the bandwith is set to be 50hz, the cap is as big as nF!

can active filter solve this problem? what is the disadvantage of active filter? we never do this kind of filer

 

[ranier]

Re: trobled with the system parameter design of fractional-p

if input frequency is 5MHz , 500k is enough bandwidth to operate .
why do u use 50 hz ? i think it is a quite low value.