I have a bunch of data taken from a CR10X and CR23X that I have discovered to be improperly grounded. Has this ever happened to anyone? What implications does this have on the quality of my data and what can be done to correct the data?
You can be sure that those of us who post on the site have all seen similar problems. Even with a perfect installation, things can go wrong later. Wired break or get cut. Vibration causes screws to get loose and render a connection intermittent.
The logger manuals are a great source for information on how to collect clean data, proper grounding and shielding practices. Not all loggers handle the signal and power grounds the same so pay attention to that as well.
Without knowing more about your set up, I can only make some general statements.
1. If your sensors were wired and measurements taken differentially, then there may not be a problem.
2. If your sensors were wired and measurements taken single ended, the problem may be small or large depending on the environment of the logger and sensors.
a. If your sensor return was wired to the AG terminal, then you still may be okay. If not, but the sensor is at the same ground potential as the battery (such as on an automotive vehicle) it may still be okay as the AG and the battery ground are tied together. See figure 14.7-1 in the manual. This will be less of a problem if nothing on the logger is taking much current.
b. If it is just the chassis ground that was missing and the system was in an electrically quite environment, the data should still be okay. If it was in a noisy environment, but you used a long integration time, the data may still be okay.
c. If the logger was not grounded but you can determine the offset of the logger to ground, you could apply that value as a correction factor.
The signal level of the sensors is important too. If the signals were in the multi-volt range, the signal to noise ratio may high enough that the errors can be ignored. However, if the signals were in millivolt range, the problem is of more concern.
And lastly, look over your data closely. Look for the impossible reading. I have seen people graph air temperature data that had noise induced readings of 100C. And I have seen people report vacuum readings below perfect vacuum.