[time-nuts] How to measure regulator noise?
donmer at woosh.co.nz
Tue Jan 8 19:53:41 EST 2008
Hi Bruce and Henk,
Your point about Bandwidth is a valid one, and as you say, it depends on
The 400H has corner frequencies of about 10Hz, and 4Mhz, so the noise BW is
a bit wider than this [Y/N?]. This setup can`t measure noise at very low
Frequencies, or Frequencies much greater than the [nominal]
4Mhz, because the noise frequency components simply can`t get through.
For my purposes the setup I described is adequate - I mainly want to compare
one regulator with another [although it would be nice to be able to do so
over the frequencies
0Hz to infinity]. I suggest that if a regulator proves itself quieter over a
restricted BW such as I described, it is *probably* quieter at other
Take that Bruce! ;-)..............................Don.
----- Original Message -----
From: "Bruce Griffiths" <bruce.griffiths at xtra.co.nz>
To: "Discussion of precise time and frequency measurement"
<time-nuts at febo.com>
Sent: Wednesday, January 09, 2008 9:26 AM
Subject: Re: [time-nuts] How to measure regulator noise?
> Henk ten Pierick wrote:
>> On Jan 8, 2008, at 12:50, Don Collie wrote:
>>> I`d just hang an AC millivoltmeter[or microvoltmeter] across the
>>> I use my H/P 400H, which will give readings down to about 50uV. If
>>> regulator produces less noise than this [say a 723, with
>>> 2uV], then you`ll need a more sensitive meter.
>> Electronic voltmeters or microvoltmeters have a noise bandwidth which
>> is larger than the bandwidth on the front panel. The front panel
>> bandwidth is related to accuracy and is not the -3dB bandwidth. The
>> noise bandwidth is nearly always range setting dependent and can vary
>> very much. I have seen a factor of four in noise bandwidth between
>> adjacent range settings. The only way to have a good indication on
>> noise is in a known and constant bandwidth. The noise bandwidth of a
>> filter is not the same as the 3dB bandwidth and dependent of the
>> filter order and shape. It is always more.
> Thus if one is to make useful comparative measurements with an AC
> voltmeter, an external filter which has a significantly narrower
> bandpass than that of the AC voltmeter is useful.
> A spectrum analyser with the capability of averaging cross power spectra
> has the added advantage (over an AC voltmeter) of being able to make
> meaningful measurements of noise below the noise of its input amplifiers
> (or the noise of external preamplifiers). Since its difficult to build a
> preamp with noise much below 10nV/rtHz whilst ensuring that the preamp
> input will survive worst case transients etc when connected to a power
> supply of 10V or more, such a capability is useful for measuring the
> noise of ultra low noise regulators which may have high frequency noise
> of 20nV/rtHz or less.
> For example the preamp used in Linear technologies AN83 is virtually
> guaranteed to be damaged by connecting its input to a powered up 20V
> Off course, connecting it to the power supply before powering up the
> supply will (if the regulator output slew rate is sufficiently low)
> allow it to be used to measure the noise of higher voltage supplies (at
> least if higher voltage 330uF OSCON caps were available). However
> accidents/mistakes do happen (as do faulty regulators) and eventually
> the preamp input opamp will be destroyed.
> Using standard electrolytics for the input coupling capacitor for
> testing higher voltage supplies isnt particularly useful as they have
> increased leakage current and associated noise. Schemes such as using a
> relatively high resistance in series with the preamp input which is
> shorted out after the coupling capacitor has charged are not foolproof
> and eventually a mistake will lead to destruction of the preamp input
> For circuit schematics of preamps with input protection schemes that
> allow power supplies with outputs greater than 5V to be tested see:
> time-nuts mailing list -- time-nuts at febo.com
> To unsubscribe, go to
> and follow the instructions there.
More information about the time-nuts