long ring oscillators td delay

Hi,

I am trying to design a basic differtial ring oscillator for a fixed frequency. As i read from the text books the frequency of oscillation is (1/2*n*td), where td is the delay per unit cell.

I was simulating the ring oscillator in spectre but the oscillation frequency is different from the formula, i would have not bothered if the frequency difference is small. But there is a 2 to 3 factor difference. I wasnt able to reson it out the huge difference. Have any people experienced such issues, if so, please help me to resolve it.

Thanks in adavnce,
Prakash.

well spectre also takes into picture the wire delay, that is not shown in the formula.

Also see the technology used.

Smaller technogies have greater wire delays

i can help with filter designs, if you need a waveguide filter i could give you an hint.

F

HI,

Presently I have used the differentiel ring oscillator configuration as shown in Razavi book, but the frequency is very different from the theroritical calculations. As said earlier the 20% deviation is acceptable but more than that, I am not able to understand the reason.

Prakash.

how do you decide the delay time of each stage? i think the td should be decided concerning the real load effect!

The formula takes into account just a small number of variables to demonstrate the oscillation effect. However, during simulation, spectre takes into account many additional things like transister (inverter) sizing, parasitics, wire delay, transmission line related effects, loading etc.

You are better off trusting spectre. don;t forget to do the post layout simulations with realistic loads on the output to get the results that are closest to the actual performance.