From: "Roger Lascelles"
References: <firstname.lastname@example.org> <5FIA9.631$1F5.email@example.com>
Subject: Re: Niosy A/D converter - an advantage ?
Date: Fri, 15 Nov 2002 11:24:33 +1100
X-Newsreader: Microsoft Outlook Express 5.50.4133.2400
"Kevin Aylward" wrote in message
> Its probably not really "noise" then, but "interference".
> For gausssian white noise, maybe, but in this case you probably have
> digital/power line garbage where this wont work.
That is the interesting part. The product has been sold for the last 4
years and gives rock steady readings on its 4 decimal digit readout. The
readings have an apparent resolution of 1 part in 9999 - with a 12 bit ADC.
So in a sense, it does "work" - by luck, perhaps.
The programmer says "It works, therefore it must be a correct design".
However, I am concerned that :
- it only "appears to work" and rigorous product testing might show
- a software revision might lead to a change in the type of noise and we
could put a lot of units out in the field with an accuracy problem.
- I suspect that the noise is coupled through the ground trace, and is at
least partly DC coupled, not just AC coupled. This means that a change in
program execution patterns could cause a DC shift wich would be fatal to the
extra resolution obtained.
The instrument has a 0.1% accuracy spec, but up to 0.01% resolution on some
If we clean up the digital noise, we will have to add a white noise dither
generator to get back our resolution. So we have to go forward carefully -
production continues at full speed !
Comments warmly invited.