[time-nuts] Harmonics suppression in ring oscillators
Hal Murray
hmurray at megapathdsl.net
Fri Mar 20 03:01:54 EDT 2015
> Hmm... maybe the assumption that all edges walk around at the same speed is
> wrong?
It's really really hard to make things like edges travel at exactly the same
speed. If it isn't exact, then one will eventually catch up with another and
self destruct.
The signal integrity wizards discuss eye patterns for multi gigabit serial
links. They now divide jitter into two parts: random and data-dependent. If
you have a long string of 0s as compared to a single 0 between 1s, the data
line will have a chance to get closer to a solid low. Starting from closer
to 0 takes slightly longer to make a transition. You can see it in the eye
diagram.
Does anybody have a scope on a ring oscillator? Is the signal symmetric? If
not, that says that the H-L transition travels at a different speed than the
L-H transition.
Actually, just looking at the prop times on a gate mignt be good enough. The
ring is just a handy signal generator.
It would be fun to make a ring with no inverters, inject a pulse, and watch
to see how long it lasts. I'll bet there is matastability type math that
depends on the width of the pulse. If you get the width exactly right it
will last a long time. Too long and it settles to all 1s. Too short and it
settles to all 0s.
--
These are my opinions. I hate spam.
More information about the time-nuts
mailing list