Pretty high - each location sample is hundreds of kilobytes if I remember correctly, although it was possible to trim that down if you knew there was a strong s

Pretty high - each location sample is hundreds of kilobytes if I remember correctly, although it was possible to trim that down if you knew there was a strong signal or you were happy to have some probability of an incorrect location. Annoyingly I can't seem to find the paper now.