Articles › November 2011 › Why don't we use smoothing on our measurements?
You may have noticed that the measurements and graphs here seem to be really rough and noisy compared to most other ones you've seen. This is because we do not use any smoothing for the majority of the data presented here. When smoothing is applied it does clean up the appearance and allow a clearer overall view of the data, however it can also obscure details and under-represent the magnitude of the results. The greater the smoothing applied, the more detail is lost.
Many measurements that you may have seen in other places will have 1/6th or even 1/3rd octave smoothing applied. One third octave smoothing is heavy handed and can obscure a lot of features in the data presented. It can turn a ragged looking response measurement into one that looks quite good. Some examples of the difference between raw data (or that with very little smoothing) and the same data with 1/3rd octave smoothing are presented here.
As can be seen in the measurements to the right, the difference is dramatic. The 1/3rd octave smoothed data looks much much better. The thing to remember is that both traces are representations of the exact same measurement, they are just presented differently. When looking at measurement data or attempting to compare it, make sure to note how much smoothing is being used. If an acoustic measurement looks very smooth and uniform it likely has some amount of smoothing on it.
We have chosen not to use smoothing on the measurements at Data-Bass because the purpose of the data presented here is to provide as much resolution, detail, and insight into the system behaviors during testing as possible. As a result, this does make the data look a little rougher and more difficult to compare in some cases.
We do at times apply smoothing in an effort to make the overall result easier to distinguish or to clean up a particularly rough looking result, but in those cases we will never use more than 1/12th octave smoothing (see image below). Smoothing of results has its place and is a useful tool but like any other it should be used with care.