EMRFD Message Archive 4529

Message Date From Subject
4529 2010-04-16 15:27:28 Tim Voltage regulator noise = phase noise
I've read on the time-nuts list in the past several years, about a subject of great discussion here, low phase noise oscillators. The time-nuts usually go even further and talk about not just the oscillator, but the whole chain including the distribution amps.

One of many ingredients in an oscillator that seems to influence the phase noise is as mundane as the power supply. Just looking at typical 7805 or 7808 regulators spec sheets, I see noise (10 to 100kHz) is specced at typical 50uV or so. I contrast this with the classic LM723 which allows the reference to be filtered... the datasheet lists 2.5uV noise for this configuration.

I will have to rig up some A/B tests and hear or measure if I can tell the difference between a J310 type LO with a 7808 regulator vs a much lower noise LM723 regulator.

Any thoughts? "Been there done that"? I don't have fancy pants spectrum analyzers but it looks like the bench approaches in EMRFD will be quite doable - the real limiting issue may be how clean of a reference oscillator I can make. Is a HP 10811 gonna be about as good as any crystal oscillator I can tack together with microprocessor crystals?

Been looking at old 70's era databooks lately and the LM723 is really really sweet. I mostly just pop in a 3-terminal regulator out of convenience but the LM723 really isn't that much harder to use.

Tim.
4530 2010-04-17 03:17:04 Johan H. Bodin Re: Voltage regulator noise = phase noise
Tim,

see http://www.wenzel.com/documents/finesse.html

There are many other interesting ideas on this site as well, click
library at http://www.wenzel.com/.

73
Johan SM6LKM

Tim wrote:
> I've read on the time-nuts list in the past several years, about a subject of great discussion here, low phase noise oscillators. The time-nuts usually go even further and talk about not just the oscillator, but the whole chain including the distribution amps.
>
> One of many ingredients in an oscillator that seems to influence the phase noise is as mundane as the power supply. Just looking at typical 7805 or 7808 regulators spec sheets, I see noise (10 to 100kHz) is specced at typical 50uV or so. I contrast this with the classic LM723 which allows the reference to be filtered... the datasheet lists 2.5uV noise for this configuration.
>
> I will have to rig up some A/B tests and hear or measure if I can tell the difference between a J310 type LO with a 7808 regulator vs a much lower noise LM723 regulator.
>
> Any thoughts? "Been there done that"? I don't have fancy pants spectrum analyzers but it looks like the bench approaches in EMRFD will be quite doable - the real limiting issue may be how clean of a reference oscillator I can make. Is a HP 10811 gonna be about as good as any crystal oscillator I can tack together with microprocessor crystals?
>
> Been looking at old 70's era databooks lately and the LM723 is really really sweet. I mostly just pop in a 3-terminal regulator out of convenience but the LM723 really isn't that much harder to use.
>
> Tim.
4531 2010-04-17 15:25:43 ehydra Re: Voltage regulator noise = phase noise
This is the only real substantial paper I found in years:
http://ehydra.dyndns.info/NG/AN491-PowerSupplyRejectionforLowJitterClocks.pdf

from silabs.

- Henry



Tim schrieb:
> I've read on the time-nuts list in the past several years, about a
> subject of great discussion here, low phase noise oscillators. The
> time-nuts usually go even further and talk about not just the
> oscillator, but the whole chain including the distribution amps.
>
> One of many ingredients in an oscillator that seems to influence the
> phase noise is as mundane as the power supply. Just looking at
> typical 7805 or 7808 regulators spec sheets, I see noise (10 to
> 100kHz) is specced at typical 50uV or so. I contrast this with the
> classic LM723 which allows the reference to be filtered... the
> datasheet lists 2.5uV noise for this configuration.
>
> I will have to rig up some A/B tests and hear or measure if I can
> tell the difference between a J310 type LO with a 7808 regulator vs a
> much lower noise LM723 regulator.
>
> Any thoughts? "Been there done that"? I don't have fancy pants
> spectrum analyzers but it looks like the bench approaches in EMRFD
> will be quite doable - the real limiting issue may be how clean of a
> reference oscillator I can make. Is a HP 10811 gonna be about as good
> as any crystal oscillator I can tack together with microprocessor
> crystals?
>
> Been looking at old 70's era databooks lately and the LM723 is really
> really sweet. I mostly just pop in a 3-terminal regulator out of
> convenience but the LM723 really isn't that much harder to use.
>