The aims of this work were to investigate (i) whether soundscape perceptual dimensions are correctly reproduced by ambisonic loudspeaker playback, (ii) whether soundscape dimensional analysis is robust to changes of location and from the field to laboratory playback, and (iii) whether a simple soundscape synthesis can be used to interactively design a soundscape. The first two aims were addressed by an experiment which attempted to repeat the dimensional analysis made by Kang (Kang, J. (2007), Urban Sound Environment. London: Taylor and Francis). Kang used semantic differential scales to conduct an in-situ survey of two urban soundscapes in Sheffield, UK. He used factor analysis to derive four perceptual dimensions from the responses. The present work repeated this approach, but the fifteen participants were judging ambisonic recordings of four soundscapes in Manchester, UK. The present work found very similar dimensions to Kang but with more variance explained: relaxation/calmness (41%), dynamics/vibrancy (10%), communication (7%) and spatiality (7%). The dimensions from the two studies load onto the semantic scales in a similar way. These results indicate that an ambisonic reproduction of a soundscape gives similar results to field experiments, though with more variance explained. They also show that dimensional analysis of soundscape response is robust enough to produce similar results for different locations in different cities. To investigate the third aim, the ambisonic reproduction was extended to a system which allowed independent interactive control of sixteen foreground sounds set in an ambisonic background soundscape. Eight participants were able to use this system to successfully design a soundscape that expressed their intentions. It was found that the designed soundscapes seemed to be based more on participant expectation of typical urban soundscapes than on their preference for individual sounds. These results suggest that a more sophisticated soundscape synthesiser might be suitable for real design problems.