[time-nuts] Switching regulator replacement for 7805
scott.j.stobbe at gmail.com
Sun Dec 4 20:45:24 EST 2016
Well a sigma-delta modulator in loose terms is an error amplifier around a
quantizer, so you get 1/loopgain rejection of quantization noise (in other
words the noise is shaped out in frequency). Resulting in a noise spectrum
that converges in 1/N versus 1/sqrt(N) for flat Gaussian noise, versus
lobes and nulls for a single sinusoid.
I totally agree that the first step is to reduce the switching residual
that is generated, even half a bond wire at say 1 nH is 13 mOhms at 2 MHz,
combine that with a power converter running 30% ripple current of a 1A
output is 300mA ripple current resulting in 4 mV just on half a bondwire.
Equivalently a 10 uF MLCC should be able to hit 10 mOhms at 2 MHz before
hitting its SRF.
On Sun, Dec 4, 2016 at 5:58 PM, Attila Kinali <attila at kinali.ch> wrote:
> On Sun, 4 Dec 2016 16:22:02 -0500
> Scott Stobbe <scott.j.stobbe at gmail.com> wrote:
> > If you wanted to be nutty you wouldn't go PWM at all, just like
> > fractional-N sythns don't just mash 2 divider values. You would
> > modulate your power stage. I don't know if you can buy one COTS, but
> > are plenty of papers on rolling your own.
> I guess you are refering to spread-spectrum techniques.
> Such DC/DC converters exist, but are usually those with high power
> ratings. IMHO it is also not worth the effort, as its main use is
> to meet EMI emission requirements. The only application that comes
> to my mind where spread-spectrum actually helps are high sensitiv
> radio receivers where every spur is a nuisance. For most other
> use, and time-nuts use in particular, it is much less useful.
> The noise energy is not gone. It is still there, just spread over
> a large bandwidth. In time measurement applications, noise is
> integrated over time _and_ frequency. Thus even if the noise is
> spread over a large bandwidth, the energy will still contribute
> to the uncertainty and degrade the ADEV. It will be just harder
> to identify as the peak is now much smaller and wanders in frequency.
> It is much better to the design such, that as little as possible
> of the switching energy leaks out of the DC/DC converter and filter
> out the rest.
> Depending on the application, another possible application is to
> sync up the DC/DC converter to the "main" clock source. This makes
> the switching noise then coherent to the system, which either makes
> it average out completely, or possible to filter it out in the digital
> domain using a deep notch-filter in receiver applications.
> Attila Kinali
> Malek's Law:
> Any simple idea will be worded in the most complicated way.
> time-nuts mailing list -- time-nuts at febo.com
> To unsubscribe, go to https://www.febo.com/cgi-bin/
> and follow the instructions there.
More information about the time-nuts