Current-Control Oscillator problem

[adscrz]

Current-Control Oscillator problem

I am designing a Current-Control Oscillator .
But it can not work in hspice,
when I set the initial value of n1, n2 in hspice:
.IC V(n1)= vcc V(n2)= 0

then it can work.
Without the initial value, it can not work.
I also find the reason why it can not work with hspice:
without the initial value, the n1 = n2 = 1 (high)
then the output of A24 or A25 will be low ( another will
be high).
So, if A24 output 0 (low) , C1 been charged by M4,
A25 must be output 1, C2 been discharged by M7.

But in hspice, if A24 output 0 (low) , C1 can not charge over Vref,
so the n1 always keep high, and the Oscillator can not work.
Though I find the reason, but I do not know why it happy,
Is the simulation software( hspice) problem or the
circuit problem ?
May I need a start-up circuit ?

BTW: When it work with the initial value,
C1, C2 charged current is: 176uA
discharged current is: 13.1uA
Frequency = 6.17M
Vdd= 3.3V Vref = 1.25V

**broken link removed**

 

[Btrend]

1.in my opinion, if u want the oscaillator to start oscallte, the phase dealy after the loop back should be 180°, and gain >=1. usually in ring oscillator, odd stage (180°) is used. if even stage (360°) is used, u should take more care in the phase delay after loop back.
2.in ur case, 8 stage is used.
and there exist a stable state when n1=n2=0,
cause in this case, the output of A13 & A14 is 1 ===> output of A24 & A25 are 0
===> C1top & C2top are high > Vref ===> n1=n2=0.
3. maybe insert a inverter to form 9 stage
4. in simulation, we usually skip initial condition, by using ramping VCC, try this.
5. At most of time, simulator is usually right if u provide it a correct model, and input stimulus.

 

[sunking]

modified the A13 and A14 for a more better structre.
such as added pull up Res for initial state

 

[watersky]

If OSC is simulated only,I think you can add the initial state. Once you connect osc with other. perhaps the initial state is not needed and osc can work.

 

[adscrz]

Thanks for all !
I will try and paste the result.