It has been my experience with large measurement systems over the past 15 years or so (that long? where did the 1900s go?) that the networks occurring between the phenomena to be measured and the digital data to be analyzed rarely even approach the level of being ignored – the concept itself often does not even reach the level of consciousness required to be ignored. The ADC output is considered to be an exact representation of the phenomena itself – and the main thought of ADCs is “Can we get more bits?”.
This seems to be a recent development; I can see such change when I compare project documents from decades ago – up to about 1990 – with project documents and stated goals created since that time. Unfortunately for me, I was trained during the earlier days and keep trying to apply the old to the new … there’s a satisfaction in developing a piece of accurate measurement equipment on time and under budget. But that’s me; I like to bounce electrons around on the head of a pin. Analog dimmers are of more interest than ON/OFF switches – and digital dimmers never seem to allow that fine touch of control a good linear system provides.
But this is now: It is the scientists and software developers that get the nice offices, clean facilities, and funding – the electronics personnel are tucked off in a back room someplace, pulled out to acknowledged existence when need be; but put away out of sight when not. I can only suppose this is because the electronics people too often point out flaws in project suppositions.
For example, I was once informed that the Nyquist limit did not need to be applied – the required changes to the project would upset the model, create too many problems in implementation, and an algorithm could be developed to correct for any resultant “errors”. At a different time on another project, all looked well – except when the physical opamp was changed to a different part number. No one considered the effect of gain-bandwidth product. It was assumed the gain remained as designed regardless of which opamp was used.
“You can’t do it that way” is not an acceptable comment once the model has been developed. “You shouldn’t do it that way” simply opens a can of worms of methodology. Projects fail when they need not as a result. I know of one that was recently shut down for “no progress” when one possible fix was suggested 5 years earlier.
“Too many cooks spoil the broth” is no longer pertinent – too many cooks helps spread the blame … and “progress” is funded while success requires a new round of proposal writing.