[SOLVED] Ring oscillator with even number of inverters

[iVenky]

A ring oscillator has odd number of inverters and each inverter is followed by a load capacitor. The total phase shift due to inverters is 180. The capacitors contribute another 180 phase shift and hence we have oscillations in ring oscillator.

Now if we have even number of inverters. We have 0 degree phase shift because of the inverters. If the capacitors also provide 360 degree phase shift at some particular frequency then even even number of inverters will oscillate, right?

Can even number of inverters oscillate?

Thanks in advance

 

[v_c]

I think you must have odd number of inverters otherwise the oscillations will not be sustained. This is because the output of the last inverter is the input of the first inverter.

 

[iVenky]

Here's one circuit that has two inverters and can still produce oscillations.

https://www.falstad.com/circuit/e-inv-osc.html

I would be really happy if you could explain me how to find out the frequency at which it oscillates.

Thanks in advance.

 

[jpanhalt]

There are many application notes on ring oscillators. Here are just two of my favorites:

https://www.fairchildsemi.com/an/AN/AN-118.pdf
https://www.onsemi.com/pub_link/Collateral/AND8067-D.PDF

The Fairchiild note describes the two inverter design, but doesn't give a general formula for calculating its frequency. Perhaps that is because it will not oscillate under all conditions. The Onsemi note does give a formula that applies when R2>>R1. For R2=10R1, f= 1/2.3R1C1

John

 

[v_c]

I guess under the "right conditions" any feedback system can be designed to oscillate. Sometimes, poorly designed power supplies look more like oscillators than power supplies. I think the advantage of the odd number of inverters is that it is probably less stringent on the conditions for sustained oscillations.

 

[FvM]

I don't see that your new thread is adding new aspects to the topic of "ring oscillators with even number of stages". You'll also notice that this is a kind of serial topic at edaboard and has been answered more than once, just look at the similar threads list...

https://www.edaboard.com/threads/240882/#post1030854

The only "new" point is the reference to the two inverter relaxation oscillator in post #3.

As explained in your previous thread, the (only) problem of the N*2 ring oscillator is the missing DC bias, respectively it's latching behaviour. This point is solved in the relaxation oscillator by breaking the DC loop and adding a feedback resistor.

In more simple words: It's no ring oscillator.

 

[iVenky]

I don't see that your new thread is adding new aspects to the topic of "ring oscillators with even number of stages". You'll also notice that this is a kind of serial topic at edaboard and has been answered more than once, just look at the similar threads list...

https://www.edaboard.com/threads/240882/#post1030854

The only "new" point is the reference to the two inverter relaxation oscillator in post #3.

As explained in your previous thread, the (only) problem of the N*2 ring oscillator is the missing DC bias, respectively it's latching behaviour. This point is solved in the relaxation oscillator by breaking the DC loop and adding a feedback resistor.

In more simple words: It's no ring oscillator.

Actually I wanted to follow this discussion in that thread but I couldn't find that thread. That's the reason why I started it again. I have added one more thing this time. I read in Sedra book that the parasitic capacitances are responsible for the additional phase shift need for Barkhouson's criteria.

---------- Post added at 22:07 ---------- Previous post was at 21:53 ----------

I don't see that your new thread is adding new aspects to the topic of "ring oscillators with even number of stages". You'll also notice that this is a kind of serial topic at edaboard and has been answered more than once, just look at the similar threads list...

https://www.edaboard.com/threads/240882/#post1030854

The only "new" point is the reference to the two inverter relaxation oscillator in post #3.

As explained in your previous thread, the (only) problem of the N*2 ring oscillator is the missing DC bias, respectively it's latching behaviour. This point is solved in the relaxation oscillator by breaking the DC loop and adding a feedback resistor.

In more simple words: It's no ring oscillator.

Thanks in advance.

---------- Post added at 22:51 ---------- Previous post was at 22:07 ----------

I think I am getting that configuration.
Just tell me if my interpretation of that circuit is correct---

Assuming that initially the input to the first inverter is at "low" then the output of the first inverter should be "high" and the capacitor charges to that voltage as its other terminal is at "low" or "ground". So now once it charges to that voltage then input of the first inverter becomes "high" and the output of the first inverter becomes "low" and the capacitor discharges and thus it produces the oscillations. This is similar to the relaxation oscillator.
There are few questions that I have
What do you mean by "latching behavior"?
This inverter oscillator seems to be entirely different from the normal oscillator. This makes me to ask one more question. How can you possibly explain the Barkhousan's criteria that take places in astable multivibrators? Ahh I am confused !!

 

[FvM]

Actually I to wanted follow this discussion in that thread but I couldn't find that thread.

I see. I often need to search the forum for previous threads. I this case, I only need to browse your recent posts...

The barkhausen criterion applies only to linear oscillators. LvW has already mentioned it in your previous thread. There must be however a positive feedback with gain > 1 to get an oscillation of ascending magnitude. But with settled oscillation, the involved waveforms are far from harmonic oscillations, only square and triangular (exponential, strictly spoken) waves that can't be analyzed in terms of phase and magnitude.

The circuit operation is explained in detail in jpanhalt's link. You can calculate the exact differential equations of the RC charge/discharge if you want to know more.

One more latching of N*2 inverter chains. They are simply working as a digital latch. Even if you manage to inject a pulse train into the chain, it falls to all '1' or all '0' after a short time.

 

[iVenky]

Okay. I am going to conclude this discussion. I searched about Barkhauson's criteria in Wikipedia and it clearly says that Barkhauson's criteria is applicable only for Linear circuits. So it cannot be applied to astable multivibrator.