In this work, we show how a distributed sensing network consisting of very low-cost nodes can also be used to locate radio transmitters without prior knowledge of which waveform is used. This information can aid in increasing location awareness among cognitive radios, as well as provide assistance in locating offending transmitters. The low accuracy of the internal clocks of these low-cost receivers as well as the geographical distribution of the nodes result in significant challenges regarding the synchronization of the receivers in order to position the source with adequate accuracy. In this article, we synchronize the nodes to an arbitrary modulated RF signal, after which we calculate estimated time differences of arrival to an unknown transmitter. We describe the implementation as well as give results on measurement accuracy in various scenarios using a prototype network of nodes spread out in the city of Turku, Finland. In the individual distance measurements of receiver pairs, the errors in distances vary between 30 and 900 m, depending on channel conditions.