Re: Equivalent time-and-bit-depth sampling?


Hi David,
In dithering the signal still only makes the signal traverse roughly one
least significant bit of headroom - in my case it would traverse the whole
headroom between the captured voltage (+overshoot) and zero. That is the
property I thought would bring an advantage. In addition it does so
"slowly", i.e. nearly all of this traversal happens at very low slew rates
compared to the slew rate of the high frequency components of Gaussian
noise, which in turn means you are not limited by the performance of the
ADC near Nyquist, and don't need to worry about aliasing as much. I believe
this could be another advantage.

On Wed, 3 May 2017 18:56 David @DWH [TekScopes], <
TekScopes@...> wrote:

On Wed, 03 May 2017 15:07:43 +0000, you wrote:

I assumed that the device holding the charge would be non-linear? On the
other hand the filter can be linear.
Sample and holds can be pretty linear or at least linear enough that
they do not hold back the ADC.

It is my hunch the resolution of sampling just one constant voltage is
going to be less than the resolution obtained by sampling a dynamic,
changing, signal, which depends on the charge.
This is what I meant by there is no need for this sort of complexity.

Almost all ADCs and especially fast ADCs have enough inherent noise
that nothing needs to be done at the input to dither the signal
allowing productive oversampling. If the ADC is quiet enough, and
some are like the ones used in the Tektronix 2232, then adding a
little bit of gaussian noise is easy enough; for instance noise could
be added to the reference input.

Posted by: David <@DWH>


Yahoo Groups Links

Join to automatically receive all group messages.