How to build a start-up circuit for ring oscillator?

[Sezi]

Hello,
I'm implementing a ring oscillator at 10MHz with cmos inverters and transmission gates in between which are controlled by control voltage. But when i simulate in spectre, i need to set initial point otherwise all inputs and outputs of inverters stay in mid-rail. how can i build up a start-up circuit for this vco? or do i need one? would it jump to some rail in real life with noise and other effects?

 

[avimit]

Re: Ring Oscillator

Yes, It would jump to some logic at power up on real silicon. You may also design the one of the inverters in inverter chain, such that the NMOS and PMOS are not perfectly balenced. i.e one either he PMOS is large or the NMOS is large, so that at power up one of them dominates.
kr,
Aviral Mittal.

Added after 2 minutes:

Forgot to mention one thing: Be careful, ring osillators are very very sensitive to process/temperature, so your 10MHz oscillator may not be a 10MHz oscillator after it has been fabricated.

 

[Sezi]

Re: Ring Oscillator

thank you for the reply. what do you suggest to prevent variations after fabrication? i don't have tight phase noise specs and don't have enoguh time to get into tuned osc. so i thought this would be ideal solution..

 

[lihui002]

Ring Oscillator

you don'nt need use start-up circuit.
At first, you will sure your oscillator loop is the positive feedback, and the gain of this loop >1(normally, it will > 2 each stage), and this oscillator will be OK.

Added after 2 minutes:

The circuit will balance at mid rail when in simulation, the reason is it is idea when simulation.
In fact, a liitle noise will induce the oscillator to oscillate because it is positive feedback.

 

[avimit]

Re: Ring Oscillator

As far as variations in frequency is concerned, ring oscillator cannot be guranteed to work at right frequency on silicon, and that is the reason ring oscillators are not used in silicon where you need exact or even near to ecact frequency.
Kind Regards,
Aviral Mittal.

 

[Sezi]

Re: Ring Oscillator

hmm ok but then what would you suggest?

 

[avimit]

Re: Ring Oscillator

Well,
It depends upon the application.
And I am sorry, I am not an expert on it. But as I have put a ring oscilltor in on of the SoCs I was working on, so I knew about it. We had put a ring oscillator to know the silicon behaviour.
Kr,
Aviral Mittal

 

[crystalguy]

Re: Ring Oscillator

Of course the main problem (and advantage) with the simulation is the fact that a ring oscillator has a Q of 1 or less than one (Q= energy in/energy out). The reason being that there is virtually no charge storage from cycle to cycle. The result of this is that the proposed oscillator is delightfully dependant on the device models and the tail current control mechanism. If the phase delay through each inverter device is something other than 180 degrees at the frequency of interest than proceed with care. In the interest of completeness, be sure that the number of inverters is an odd integer. If you are still struggling with this it might be helpful for us to know by what method of simulation you have arrived at your results.

 

[Fom]

Re: Ring Oscillator

Another thing: use conservative Spectre option when you simulate oscillator.

 

[Sezi]

Re: Ring Oscillator

I have been simulating only the vco itself with a control voltage input. the number of inverters is 5 and each of them have 700fF load capacitance at the output node because frequency of interest is 10MHz and i can't achieve it with only parasitic capacitances. the frequency control is done by varying the resistance of a CMOS transmission gate placed between each inverter pair. i think it's useful to minimize the transient simulation step and make a conservative simulation. also i want to mention that when i placed the vco in the pll, i don't need an initial condition, it starts itself. so i think the problem is solved, or was there any?
thank you all guys

 

[leonlin]

Re: Ring Oscillator

The circuit will balance at mid rail when in simulation, the reason is it is idea when simulation.
In fact, a liitle noise will induce the oscillator to oscillate because it is positive feedback

 

[dasong]

Re: Ring Oscillator

i design a pll used the 3rd ring oscillator,i think the ring oscillator don,t need the start up circuit.but i have a problem the ring VCO's noise charcher is poor ,how can imporive the noise character.

 

[Sezi]

Re: Ring Oscillator

Hi dasong,
How do you make a noise analysis? is it pnoise? and what's a good noise performance and what's a bad noise performance? thanks in advance..

 

[bill_sun]

Ring Oscillator

u need not a start-up circuit.it's just a simulation problem.

Added after 3 minutes:

yes ,u can set an initial voltage at some nodes.