Home Up Contents

Measure Once, Think Twice
Home Up

 

Home
DC or not DC...
Potentially Positive
Measure Once, Think Twice
Reading Between the Lines
Something Super
Want to see my Etchings?
The Gestation of a Regulator
The Sulzer Circuit
The Jung Circuit

Measuring Microvolts

The noise output of a linear power supply is dependant upon a number of factors, primarily: -
bullet Noise in the Voltage Reference
bullet Voltage / Current Noise of the error amplifier
bullet Line noise / line rejection
bullet Noise gain of the circuit
bullet Dynamic noise from circuit load demands
bullet Other regulator circuit element noise contributions

In order to be able to quantify the improvements made to the regulator circuits it is necessary to be able to resolve measurements in the order of single-digit nV / rt Hz levels.

This requires some serious attention to detail - a circuit with a gain of several hundred, ultra-low input noise and excellent common-mode rejection is necessary, in order to ensure that stray fields do not dominate the measurements made.

In short an instrumentation amplifier was built, of the classic three op-amp style, using AD797's in order to achieve the lowest noise figure possible. The circuit was run from two 9V batteries, in order to prevent any earthing problems, or mains-generated noise contaminating the results.

The output of this amplifier is then sent to a PC sound card and a PC running some suitable FFT Spectrum Analysis Software. I originally used a SoundBlaster 'Live' card, a reasonably low noise card, but which can only be reliably used at a 48kHz sample rate. Any other rate selection results in on-the-fly sample rate conversion, which not only sounds dreadful, but measures dreadful too.

Latterly I have changed to an M-Audio 'Audiophile 24/96' card, capable of a much lower noise floor and wider bandwidth. It's also a great sounding card too, and I use it a lot for recording radio broadcasts straight to the PC.

The resultant tests were exactly as expected - the Sulzer was significantly quieter than the POOGE design, and it's measured noise was sufficiently far above the noise floor, that improvements to noise could be quantified.
bullet Lesson 3 - Learn what to measure

Audio is a subject where measurement is often espoused as being meaningless, and that all that matters is how the result sounds. I have sympathy with that view, if it sounds good, it is good but being an engineer I like to know why - I'm sorry it's just the way I am ;)

Simplistic steady state noise changes were easily measurable, but the actual sonic results of various changes did not always correlate with the 'improvement' made.

It was time for a visit to the Blue's Clues thinking chair, for a little deliberation.

Deep Thought - The Results

 

[ Home ] [ Up ] [ News ] [ Links ] [ Andy's Audio Journeys ] [ Projects ] [ Andy's Guestbook ] [ Press Releases ] [ The Products Page ]

Send mail to webmaster@alw-audio.co.uk with questions or comments about this web site.
Copyright © 2003 ALW Audio
Last modified: Saturday July 19, 2003 09:06 +0100