I have a point layer with soil values. How can I get a raster layer with different classes of values, since there are areas without point sampling, what´s the best way to have a regular cover of the total area, given the sampling points??
Without knowing much about the data try this approach:
In QGIS go to Raster > Analysis > Grid
Select the point file as input and name an output file.
Select algorithm as Inverse distance to power. Just as a starting point try setting min points to 5 and max to 20. Start with large radius say 30 or 40 for both and then slowly decrease as you observe results. Leave other settings as default.
Once map is processed right click on the layer and go to Properties > Style and select Render type as Single band Psuedocolor. Select a color ramp then click Classify and Apply.
Now this assumes you are using a UTM type projection with map units in metres.
IDW in QGIS/GDAL is a good start but with soil sampling you may want to try Kriging the data. University of Sydney have software Vesper which is designed for this. It has a fairly steep learning curve though. More information here: http://sydney.edu.au/agriculture/pal/software/vesper.shtml
First you must have an opinion to the resolution of your data. How fare away do the soil data have a spatial relation. You can then interpolate between the lines. Your soil data resolution will give you a hint for how large you wants the grid cells in the raster. A good interpolation method on soil data (geophysical or geochemical) is IDW since it smoothens you data slightly washing out some of the uncertainties in machinery and laboratory analysis. In QGIS menu Raster > Interpolation pick output interpolation method to IDW and experiment with the settings.