It is of a vital importance for automotive industry to reduce the frictional losses in internal combustion engines and hence their fuel consumption and CO₂ emissions. The major contributors to this are the oil control ring (OCR) and cylinder liner interactions. These interactions are of complex multi-physics and multi-scale nature and many attempts have been made to improve the modelling and understanding of the phenomena involved. Even though continuous improvements have been made (see for ex. [1-3]), the form, waviness and roughness variation of real engineering surfaces still cause a large scatter in the results. The most of the models include measured liner surfaces, which are then filtered to capture the micro-effects of the roughness scale. However, by doing this, the contact conditions change (asperity contact distributes more evenly, see Fig. 1) such that the part of the boundary friction reduces and the part of the hydrodynamic friction increases accordingly for a given engine speed. This study focuses on one of these multi-scale problems, namely on how the 3D measurement type/size and filtering affects the predictions of friction between cylinder liner and OCR of truck engines. Two types of 3D liner surface measurements were investigated: (i) interference (with an objective of 2.5x, size of 2.5x3.3mm) and (ii) 4x8mm stylus measurement. The form was removed by fitting and subtracting a 2^nd order polynomial and the waviness was filtered out by using a robust Gaussian filter with three different cutoffs: 2.5mm, 0.8mm and 0.25mm. The friction behaviours (i.e. the Friction Mean Effective Pressure-FMEP) between the primary/ filtered liner surfaces and a perfectly flat ring surface were then simulated for different engine speeds. The relative errors of the OCR’s FMEPs with respect to the primary surface were evaluated and results discussed.