Differences in Roughness: Rz vs. Ra

As this poster says, there is no real relationship between the roughness measurements. Typically, they are used by manufacturers (or researchers) to define somewhat different roughness characteristics (peak heights, peak to peak distances, frequency of valleys beyond a certain depth, etc). A few years ago, there were reports circulating of another technique that involved calculation of a roughness parameter from roughness trace data using a "chaos theory" recursion analysis. It was projected to have great advantages to defining surface roughness. I tried it a few times, but never saw any advantage to it beyond use of Ra values. To my point, with regard to roughness parameters, use whatever parameter works best to describe the surface for it's intended use.

An important roughness parameter for the surface "quality" is the Abbott curve but it is not used as often as it would be recommendable.

Ra is not of any use for surface quality. It is only the center average of some data points. If you need more valleys than peaks in the surface Ra has no peak or valley information. An Ra plot can be inverted and the number is the same but there are now more peaks than valleys. Rz takes five samples of peak & valley information and is about 7 times more sensitive than Ra for process control. If you need more functional surface information you must use other paramters. (Abbot-Firestone, etc.)

Some very good info here: http://www.taylor-hobson.com/faq.php