This approach will not work in all cases but did for me. The noise was due to the temperature sensor behaving strangely so all I really needed to do was ignore a bunch of data. My samples were 1 second apart originally and when plotted looked like this...
That wasn't ideal for me. To address this I created a new column and put the following formula into it, then filled that down all the way to match the number of data rows I had.
What this gives me is a counter that goes from 0-59 and then resets again and cycles like this all the way down my data.
After I had this additional column, I selected it and used the filter button to make it a filtered column.
Pressing the filter icon on the column let me specify which values I wanted to see, so I selected 0, 15, 30 and 45. This roughly translated to 4 samples per minute now, instead of 60 i.e. a sample 15s apart. Effectively this reduced the resolution of my data but also had the side effect of filtering out lots of the noise.
So now I was left with a subset of my original data...
Creating a chart from this data meant the chart would only display the data that was not filtered out! With my subset of data the chart looked much cleaner!
There are statistically better ways of doing this kind of filtering, this one is just quick and easy and worked in my case. I wouldn't use it in all situations.
I've also uploaded my original data set in case anyone is interested: ps4_fan_data_set.zip [22.76 Kb] (199 downloads)