[time-nuts] Effect of EFC noise on phase noise
scott.j.stobbe at gmail.com
Mon Aug 1 16:50:07 EDT 2016
The broadband thermal noise at a circuit point with a cap is always kT/c
On Monday, 1 August 2016, Bob Camp <kb8tq at n1k.org> wrote:
> If you wire up all the possible circuits and check them all out … the
> answer is that big C / small R wins. Big R gets you into resistor noise
> and stray pickup.
> > This duplicates the problems encountered when trying to quantify low
> > frequency noise from a voltage reference; it is difficult to make an
> > low frequency high pass filter with lower noise than the lowest noise
> > references and the capacitor is the problem.
> > In Linear Technology Application Note 124, Jim Williams discusses the
> > problems with electrolytic capacitors for this type of application. I
> > have read that you *can* get away with aluminum electrolytics if you
> > grade them for low leakage and low noise. The dielectric absorption
> > is also a problem unless you can wait hours for best performance.
> > What about the alternative of buffering the signal with a low noise
> > low input bias current operational amplifier so that a large film
> > capacitor can be used instead? Is the low frequency noise of a good
> > operational amplifier still too much? What about a chopper stabilized
> > amplifier without suitable output filter?
> > On Mon, 1 Aug 2016 11:46:51 -0400, you wrote:
> >> Hi
> > .. until you discover that you picked the *wrong* capacitor manufacturer
> and you have
> >> more noise from leakage in the cap than you did to start out with :)
> In general “big C and
> >> small R” is the better solution than “big R and small C”.
> >> The pesky part is that with electrolytic caps, the whole “noise
> current” thing changes as
> >> the voltage moves around. You go to measure things and by the time the
> gear is set up,
> >> the noise has dropped. Turn it all off, come back the next day and it’s
> noisy again.
> >> An even more subtle issue can be capacitor temperature coefficient on
> really long Tau filters. If C
> >> changes (due to temperature fluxuations) faster than the settling time
> of the filter, you get noise. Charge
> >> is the same so delta C gives delta V.
> >> I *wish* I could tell you that was all purely theoretical.
> Unfortunately it’s based on empirical data
> >> collected in the “how could I be so stupid” fashion.
> >> Bob
> >>> On Aug 1, 2016, at 11:21 AM, KA2WEU--- via time-nuts <
> >>> A good filter in the cable is highly recommended, 5 KOhm & 1000 uF
> >>> many things
> > _______________________________________________
> > To unsubscribe, go to
> > and follow the instructions there.
> To unsubscribe, go to
> and follow the instructions there.
More information about the time-nuts