The Designer's Guide Community Forum

The Designer's Guide Community Forum
https://designers-guide.org/forum/YaBB.pl
Design >> Mixed-Signal Design >> Ring osc start up
https://designers-guide.org/forum/YaBB.pl?num=1266370083

Message started by jugemu1234 on Feb 16^th, 2010, 5:28pm

Title: Ring osc start up
Post by jugemu1234 on Feb 16^th, 2010, 5:28pm

Hi,
Im designing 4 stage ring oscillator. Each delay cell consists of 2 cross-coupled inverters.
It can oscillate in any corners but the time to start oscillation is uncertain. I believe this has annoyed not only me, so can you suggest me to get this oscillate properly on start-up?

Assumption; current starved type ring oscillator, enable signal is on much earlier than current is actually supplied.

Title: Re: Ring osc start up
Post by thechopper on Feb 16^th, 2010, 7:10pm

jugemu1234 wrote on Feb 16^th, 2010, 5:28pm:

Assumption; current starved type ring oscillator, enable signal is on much earlier than current is actually supplied.

Hi,

I think the problem is the enable signal is on earlier than the bias current.
You should change the generation of the enabling signal based on the bias current you use for suppling the oscillator.

Best
Tosei

Title: Re: Ring osc start up
Post by jugemu1234 on Feb 16^th, 2010, 7:22pm

Hi,

I understand your point. But nonetheless, it does not assure it oscillates right after specific time point, does it? If it becomes like inp=inn=op=on=..., which is so rare but possible, it might to take uncertain time to oscillate.

Title: Re: Ring osc start up
Post by Berti on Feb 16^th, 2010, 11:48pm

Is this not just a simulation problem? If you have enough gain in your delay stages then oscillation will always start due to noise and mismatch. In simulation however, this is not always the case. Do you use initial conditions to ensure that oscillation starts in simulations?

Regards

Title: Re: Ring osc start up
Post by jugemu1234 on Feb 16^th, 2010, 11:56pm

Hi,

In simulation, yes, I can observe oscillation no mater what. As you pointed out, some fluctuation was amplified.
However, what I think problematic is it depends on something unpredictable, somethign uncertain. For example, ss process at -40, gain becomes small enough probably.

Therefore, I think its better to have sort of start up to provide the seed of oscillation to ensure it oscillates in certain time.

Thanks,

Title: Re: Ring osc start up
Post by loose-electron on Feb 21^st, 2010, 2:03pm

If you really want this - do a set of switches that forces one stage inputs to opposite states -

generally its not needed, but if it gives you better feelings about it on power up, go for it.

Title: Re: Ring osc start up
Post by Mayank on Jun 3^rd, 2010, 6:08am

Quote:

If you really want this - do a set of switches that forces one stage inputs to opposite states -

generally its not needed, but if it gives you better feelings about it on power up, go for it.

In some chips at 40nm & below, silicon results showed that oscillation on internal oscillator nodes die down too slowly sometimes.
Hence, I generally do that in all oscillators, but i power down all nodes to GND.
That works too, because cross-coupled inverters force opposite movements on psuedo-differential nodes at power-up.

--
Mayank.