Four different fuel blends have been fired in a 28 MWel BFB. Wood pellets (test 0) were not problematic for about ten years, contrary to a mixture of demolition wood, wood cuttings, compost overflow, paper sludge and roadside grass (test 1) which caused excessive fouling at a superheater bundle after already a few weeks. Two week tests with the test 2 blend (wood chips and cuttings, compost overflow and paper sludge) and test 3 blend (wood chips, demolition wood and wood cuttings) showed that, based on air cooled probe tests, fouling was slightly higher with test 3 compared to test 2. The bulk of the deposits consisted of a mixture of CaSO4 and K2SO4 (with minor amounts of Na2SO4). The occurrence of alkali sulfates is explained by a combination of 4 different deposition mechanisms. Gaseous alkali chlorides that condense, either as chlorides or sulfates and form aerosols, that deposit, especially initially, by thermophoresis (1), and, in the more mature deposit, by eddy diffusion and Brownian diffusion (2). The gaseous alkalis can also condense directly on the tube surface or on, or in the deposit (3). Gaseous alkalis can also condense on larger particle surfaces that deposit by inertial impaction (4). Whether gaseous alkalis have sulfatized during transport or after deposition, or both, could not be determined. CaSO4 is formed by sulfation of Ca-containing particles (not being silicates or slag) that cause the deposits to be sintered modestly (tests 2 and 3) or to a large extent (test 1, matured deposit). A molten K-Na-SO4-Cl salt is thermodynamically stable at the temperature range of tube surface +40 K to local flue gas temperature, and may also have contributed to sintering when reactive Ca is present locally in only low concentrations. Thermodynamic calculations predict fouling tendency as test 3 > test 2 > test 1 > test 0. The total amount of molten salt seemed a better predictor for fouling than the amount of salt as a percentage of total ash. (C) 2014 Elsevier Ltd. All rights reserved.