What type of simulation should I do for a 10MHz Oscillator?

[Hitotsu]

10 mhz oscillator

Hi guys/girls, thank you for looking at this thread.

I had designed a Ring oscillator that runs at 10MHz
I had simulate fast/slow corner and find that its within +-20%
I also done Monte carlo simulation

Is there any other type of simulation that i should be doing? Say for application in RFID.

Thank you for your help
Hitotsu

 

[erikl]

oscillator fs type

... Is there any other type of simulation that i should be doing?

sf / fs conditions simulated, too?
Self start under MC-mixed conditions, especially worst case process, min. op. voltage and low temperatures (min. op. temperature ... 0°C).
Self start with slowly rising op. voltage.

 

[Hitotsu]

oscillator monte carlo simulation

... Is there any other type of simulation that i should be doing?

sf / fs conditions simulated, too?
Self start under MC-mixed conditions, especially worst case process, min. op. voltage and low temperatures (min. op. temperature ... 0°C).
Self start with slowly rising op. voltage.

Thank you so much for telling me that. May I ask how do i simulate fs & sf? I do know for ff is just 'fast'

I don't really get the MC-mixed condition, whats is that and how can i simulate it? pls advice.

I did simulate the worst case (temp, vdd & res) though

 

[erikl]

Re: What type of simulation should I do for a 10MHz Oscillat

Thank you so much for telling me that. May I ask how do i simulate fs & sf? I do know for ff is just 'fast'

fs means fast pfets combined with slow nfets, sf the other way round (and I guess it's clear what ff means ;-) ). These are other possible "worst case" combinations which could occur on a processed wafer. If your PDK contains these models, it is wise to simulate with them.

I don't really get the MC-mixed condition, whats is that and how can i simulate it? pls advice.
I did simulate the worst case (temp, vdd & res) though

The problem with oscillators sometimes is, that they won't self-start under special condition combinations, somewhere in the 3-dimensional PVT space, which cannot be found by just simulating the extreme corners of this space (I once happened to have an oscillator design, which worked (self-started) at all simulated PVT corner conditions, but refused to do so at the condition combination {fs , highest (!) VDD, in the T-range -25 .. 0°C}. This was discovered during lab experiments on delivered chips, and could be verified by postLayout simulation under appropriately combined simulation conditions).

This is what I meant with MC-mixed conditions: A selection of different PVT simulation combinations apart from the corner conditions. Not necessarily selected by a Monte Carlo mechanism (actually I don't know if it is possible to select random PVT combinations by the MC method); such local PVT points (combinations) could also be selected intuitively.

If you have the time, just try some more points in the PVT space! With slowly ramping-up VDD.
Good luck! erikl

 

[Hitotsu]

Re: What type of simulation should I do for a 10MHz Oscillat

Thank you so much for telling me that. May I ask how do i simulate fs & sf? I do know for ff is just 'fast'

fs means fast pfets combined with slow nfets, sf the other way round (and I guess it's clear what ff means ;-) ). These are other possible "worst case" combinations which could occur on a processed wafer. If your PDK contains these models, it is wise to simulate with them.

I don't really get the MC-mixed condition, whats is that and how can i simulate it? pls advice.
I did simulate the worst case (temp, vdd & res) though

The problem with oscillators sometimes is, that they won't self-start under special condition combinations, somewhere in the 3-dimensional PVT space, which cannot be found by just simulating the extreme corners of this space (I once happened to have an oscillator design, which worked (self-started) at all simulated PVT corner conditions, but refused to do so at the condition combination {fs , highest (!) VDD, in the T-range -25 .. 0°C}. This was discovered during lab experiments on delivered chips, and could be verified by postLayout simulation under appropriately combined simulation conditions).

This is what I meant with MC-mixed conditions: A selection of different PVT simulation combinations apart from the corner conditions. Not necessarily selected by a Monte Carlo mechanism (actually I don't know if it is possible to select random PVT combinations by the MC method); such local PVT points (combinations) could also be selected intuitively.

If you have the time, just try some more points in the PVT space! With slowly ramping-up VDD.
Good luck! erikl

Hi Erik,

You have been most helpful & I really appreaciate that. I will try my best to find that special condition if it exists. My ring oscillator deviate alot at Monte Carlo analysis until i think i need a digital trim. Any fundamental design advice to make it more stable? Thanks for the help once again.

Best Regards
Joe

 

[erikl]

Re: What type of simulation should I do for a 10MHz Oscillat

Any fundamental design advice to make it more stable?
Best Regards
Joe

Hi Joe,
just these trivial recommendations: Try to compensate temperature dependencies, and stabilize VDD ! ;-)
There's no cure for process variations (apart from screening)
Cheers, erikl

 

[Hitotsu]

Re: What type of simulation should I do for a 10MHz Oscillat

Any fundamental design advice to make it more stable?
Best Regards
Joe

Hi Joe,
just these trivial recommendations: Try to compensate temperature dependencies, and stabilize VDD ! ;-)
There's no cure for process variations (apart from screening)
Cheers, erikl

Hi Erik

Thank for the advice, i'll think about it.

Best regards
Joe