If you look at the spec. for those 24-bit ADC, you will find that the input
frequency range is geared for audio frequency. Meaning it usually will not
detect frequencies down to DC. This is not acceptable for NMR experiment
unless you can convince your user not to put signal at on-resonance. It also
won't go up to very high frequency, but that is less an issue. The second
issue is the pipeline delay, that is the time between the signal feeds the
ADC until the digital signal comes out from the ADC. Internally, the 24-bit
ADC has digital filter, usually with fixed frequency bandwidth, in there and
it takes time for the filter to calculate the result. But this is not an
important issue. The third one is actually more serious, can you make your
input noise so small to make 24-bit ADC worthwhile? Suppose the ADC takes
10V input, it will make noise 10V/2^24 which is less than 1 micro V to make
using 24-bit ADC worthwhile. It is more a gimmick from the ADC manufacture
to think that 24-bit ADC is that useful. Of course, if you have one of those
ultra-low temperature input from your probe up to ADC, go for it by all
means. The key here is not only the cryo-probe, it is also all the
amplifiers in your signal acquisition pathway.
Over-sampling isn't a way to solve your dynamic range problem,
signal-averaging is. One small correction on that 256 oversampling to gain 8
bit resolution. It should be 4^8 which is 65536 oversampling. 256 times
oversampling is actually relatively simple given today's 10MHz 16 bit ADC.
Just my 2 cents.
Albert
on 7/25/02 6:05 AM, Alan Kenwright at a.m.kenwright_at_durham.ac.uk wrote:
> I really just wanted to open a bit of a discussion here.
>
> Recently we have been doing some work on our 500's looking for very
> small amounts of impurities in samples by running proton spectra.
> Typically we use as much oversampling as we can and still some of the
> impurities are near the limits of detection. I know there must be other
> people out there interested in doing the same thing (particularly in
> pharmaceuticals), so the question I want to raise is whether there is a
> good reason why we should still be using 16-bit ADC's and oversampling
> when 24-bit ADC's are now readily available? (A 24-bit/96 kHz stereo
> soundcard for a PC costs $US500 or less, which doesn't seem much in the
> context of a $US500k spectrometer). I think I'm right in saying that to
> achieve the same accuracy/dynamic range using oversampling you need an
> oversampling factor of 256, which is not realistic.
>
> I realize that there are other aspects to DSP, but actually extending
> the dynamic range at very little cost seems to me the kind of thing we
> should be looking for in terms of development, particularly if you start
> looking at things like cryo-probes at high field (I wish!). Is it
> simply that the spectrometer manufacturers don't feel this is a
> priority? Do you feel it's important, or is this really a minority
> interest?
>
> --
>
> Alan.
>
> Alan M Kenwright
> Senior Research Officer (NMR)
> Department of Chemistry
> University of Durham
> Durham DH1 3LE UK
>
>
>
Received on Mon Jul 29 2002 - 09:04:32 MST