As promised, I’m now returning to the subject of data QC issues, but will be addressing those types less esoteric in nature. That is, those data issues that occur and are relevant in our every day lives, as opposed to the people living on the moon!
In Branden’s post from Friday on response files, he suggests one method to QC its contents is by using SQLX. For the next few posts, I will look at the various ways a response file can be wrong, and how SQLX can make the user aware that these QC issues exist and require correction.
For the first example, let’s look at what happens when the Gain specified in the response file is wrong. Using SQLX, trace data is analyzed to create PSD’s, with a response file provided for each channel of a station. This response file is required in order to remove the instrument’s own “noise” from the signal, resulting in a frequency domain representation of pure ground motion only.
Under normal operating circumstances, a seismic station will record Earth noise falling within a nominal range of accepted noise levels. For some time now, this noise range, defined across a broad range of frequencies, is accepted to be the Peterson High and Low Noise Models (HNM and LNM). Specifically, LNM represents the absolute best possible scenario of low noise levels that a seismic station may record on planet Earth.
So when the Gain aspect of a response file is wrong, for example, what results is a vertical shift in the overall plot, either up or down, with the overall shape being retained. For the following plot, the Gain aspect of the Response File has been modified to be much too low. The resulting PSD PDF plot, then, falls well below LNM, which on planet Earth is simply impossible:
When providing a correct response file, having the right value for Gain, the resulting PSD PDF plot displays the same curve shape as above, but having overall noise levels falling nicely inside our reference baselines:
Future posts will look at what happens when the poles and zeroes of a response file are wrong, and what these problems look like in our handy PSD PDF plot.