ENOB is essentially a measure of the noise and distortion
at a specified frequency.
> If I have a 14 bit ADC with an ENOB of 12bits, can I simply drop two
> bits from each and every sample ?
That's a real bad idea. Even though the two LSBs may be noisy, there
is almost certainly still data there, and you want to use it. For
example, the ENOB may be a function of frequency, and there may
be frequencies of interest that have a relatively high ENOB. So
if you dropped two bits based on one or two frequencies, you'd
be missing out on better performance at other frequencies.
> If so, have I further increased my quantization noise ?
> What are the benefits and penalties to this ?
The penalty is a loss of information. If you're using an FPGA or other
digital hardware to process the signal, then the benefit is less
hardware. If you're using a DSP or GPP, then I see no benefit since
practically all have data path widths of multiples of eight.
Why do you want to drop the bits? Why not just use them?
--
% Randy Yates % "Ticket to the moon, flight leaves here today
%% Fuquay-Varina, NC % from Satellite 2"
%%% 919-577-9882 % 'Ticket To The Moon'
%%%% <yates@ieee.org> % *Time*, Electric Light Orchestra
http://home.earthlink.net/~yatescr
Reply by Fred Marshall●November 18, 20062006-11-18
<yazz35@gmail.com> wrote in message
news:1163829895.699510.176890@m7g2000cwm.googlegroups.com...
> If I have a 14 bit ADC with an ENOB of 12bits, can I simply drop two
> bits from each and every sample ?
>
> If so, have I further increased my quantization noise ?
> What are the benefits and penalties to this ?
I think you will find that ENOB is a function of how much of full scale is
being used. Consider this: a notional 14-bit A/D for which the input never
toggles the 2 most significant bits. Then it looks like a 12-bit A/D for
all practical purposes - with the noise floor to go along with it.
This implies throwing out the MSBs instead of the LSBs doesn't it? Are you
willing to do that? Is it justified on the data?
Fred
Reply by Jeff Caunter●November 18, 20062006-11-18
>If I have a 14 bit ADC with an ENOB of 12bits, can I simply drop two
>bits from each and every sample ?
>
>If so, have I further increased my quantization noise ?
>What are the benefits and penalties to this ?
ENOB is a subject in itself, and requires some reading to fully understand
its implication to each application.
In effect, ENOB is indicating how each individual sample's accuracy is
affected by noise and distortion. Some of these undesirable effects can be
reduced by various forms of integration, revealing more of the true
bit-significance of the converter. If you drop the LSBs of the converter
output, then you lose this ability.
Try taking a block of data quantised with the 14-bit ADC, FFT it, then do
the same with the same data, but reduced to 12-bits. You will see they are
not the same.
As I said, this is not a subject that can be explained in just a few
words.
Jeff
Reply by yazz...@gmail.com●November 18, 20062006-11-18
If I have a 14 bit ADC with an ENOB of 12bits, can I simply drop two
bits from each and every sample ?
If so, have I further increased my quantization noise ?
What are the benefits and penalties to this ?