The air/fuel ratio (AFR) is a key contributor to both the performance and emissions of an automotive engine. Its variation between cylinders - and between engine cycles - is of particular importance, especially during throttle transients.This paper explores the use of a fast flame ionisation detector (FFID) to quantify these rapid changes of in-cylinder composition in the vicinity of the spark gap. While this instrument actually measures fuel concentration, its results can be indicative of the AFR behaviour. Others have used the FFID for this purpose, but the planned test conditions placed special demands on the instrument. These made it prudent to explore the limits of its operating envelope and to validate the experimental technique. For in-cylinder sampling, the instrument must always be insensitive to the large pressure changes over the engine cycle. With the wide range of engine loads of interest here, this constraint becomes even more crucial. As the planned engine speed was among the highest reported for in-cylinder sampling with the FFID, response time becomes more of an issue. Moreover, sampling close to the spark gap severely limits the sampling time due to the short flame arrival time. Some implications of these constraints are discussed.A series of tests were conducted in a single-cylinder CFR engine to explore these issues. Simultaneous measurements at two sampling points were used to study the effect of FFID signal duration on the measured fuel concentration. Natural gas was used as the fuel, and care was taken to ensure a homogeneous mixture for the tests. Engine load, speed and spark timing were varied to create a range of conditions. Several configurations of the FFID were investigated in the quest for pressure insensitivity and adequate sample signal duration. Qualitative arguments and preliminary results of complementary computer modelling show the importance of early sample induction into the FFID.The results of validation tests using the final configuration are presented and show that the method can achieve a precision about the same as that for the FFID itself (∼2%). As this is much less than typical cylinder-to-cylinder and cycle-to-cycle fluctuations in the AFR at steady state and following throttle transients, the FFID should provide useful information about these fluctuations.