[time-nuts] Fwd: Harmonics suppression in ring oscillators

Attila Kinali attila at kinali.ch
Tue Mar 17 17:19:09 EDT 2015

On Tue, 17 Mar 2015 07:52:07 -0700
Alex Pummer <alex at pcscons.com> wrote:

> Hi Attila, just think how all ring/delay line oscillators working: a 
> status change is traveling trough a delay, and after its arrival at the 
> next active component it will release a new status change, which will 
> travel and arrive at the next active component, and after once it will  
> get back to the "begin of the loop". Basically time which required to 
> travel between the active devices will determine the frequiency, and 
> --as it now obvious -- the status change can not release new status 
> change at the following active component before arriving to that next 
> active component , therefore self the oscillator can not run at higher 
> frequency that as it determinated by the delays. Of course depending on  
> the wave form at the output of the system you could see many other 
> spectral components

Actually, no. If you argue digitally, then you can prove that
any odd number of transitions is a valid state of the ring
oscillator. There is nothing that a "status change" will latch
on until it comes around again. E.g. consider a ring of 9 inverters.
The output state of each inverter could be:


I.e. we have 3 transitions in here, and oscillate at 2 times
the fundamental frequency. You can even show that something like


is valid, where the oscillator would oscillate at 4 times the
fundamental frequency.

If you argue analog (the circuit is better described in the analog
domain than in the digital), then you can have even higher modes,
as long as the Barkhausen Criterion is fullfiled, which are
infintely many, if you disregard the unity gain frequency.
Even if you take the limited gain bandwidth product into account,
the number of possible harmonics is quite high, especially for
longer ring oscillator chains.

As i wrote, for delay line oscillators, they usually use
frequency dependent components (aka filters) to select one
harmonic mode and suppress all others. But for ring oscillators,
i have not seen any description where any frequency dependent
component was used. All they talk about is a chain of inverters.

			Attila Kinali
< _av500_> phd is easy
< _av500_> getting dsl is hard

More information about the time-nuts mailing list