We use the preliminary results of a new survey of radio sources made using the Ryle Telescope at 15.2 GHz, to estimate the impact of foreground sources on cin-wave cosmic microwave background (CMB) images. This is the highest frequency survey that is relevant to the issue of radio source contamination in CMB experiments. The differential source count of the 66 sources found in 63 deg(2) is 80(S/Jy)(-2.0) Jy(-1) sr(-1), from approximate to 20 to approximate to 500 mJy. Extrapolating this to 34 GHz (where many cm-wave CMB experiments operate) gives an estimated temperature contribution of sources DeltaT(conf) = 9 muK in a CMB image, with a beam corresponding to multipole l approximate to 500. A means of source subtraction is evidently necessary, otherwise the signal-to-noise ratio in CMB images will be limited to 4 or 5, becoming worse at higher resolution. We compare the population of sources observed in this new survey to that predicted by extrapolation from lower frequency surveys, finding that source flux densities, and indeed the existence of many sources, cannot be determined by extrapolation.