Combustion information from three combustion chamber geometries was analyzed: Pancake and horseshoe geometry on a single-cylinder research engine, and pentroof geometry in a turbocharged four-cylinder production engine. Four different fuels were used. In the horseshoe configuration, the cylinder pressure traces from the burnt gas and from the end-gas pocket were evaluated. It is shown that the characteristics of knock are to a large degree a function of the combustion chamber geometry and that they are influenced strongly by the transducer position. It is shown for pentroof geometry that the number of cycles required to properly describe the knock population is a function of the knock intensity. A large error potential is shown for samples smaller than about 100 - 200 consecutive cycles. Good agreement between knock description based on accelerometer data and based on pressure data was found. It is shown that fuel influences both the change in knock intensity with change in ignition timing and the relation between knock intensity by cylinder pressure and accelerometer due to changed frequency content.