The accuracy of determination of the moisture content of grain and seeds by estimation of their dielectric properties is well known to be dependent on the density of the samples. Where measurement involves the use of bulk samples of whole grain, as in a parallel plate or concentric cylinder capacitor, if the temperature is held constant, it is the bulk density which primarily influences the measured value of material of a given moisture content. A range of mechanical and electrical means of minimising this effect has been put forward by previous authors. The present paper reviews the electrically based solutions; these have been investigated and an analysis made of their effectiveness in minimising the density dependence of the measured dielectric values for static samples. A range of excitation frequencies from 500 kHz to 5 MHz was employed in laboratory experiments on hard winter wheat, Triticum aestivum L., varieties Mercia, Hereward, and Hussar, which were preconditioned to moisture contents from 10 to 22% wet basis. The bulk density of the samples was controlled between 665 and 873 kg/m(3) in a concentric cylinder capacitor. Of the models investigated for the correlation of the dielectric parameters of static samples of grain and moisture content, the most satisfactory in terms of the size of the errors involved, was the single frequency method based on the density-independence of the function [(epsilon' - 1)/epsilon '']. Non-linear regression analysis using data from varieties Mercia and Hereward yielded an equation capable of estimating wheat moisture content with a standard error of calibration of about 0.3 percentage point for moistures ranging from 11 to 22% at 1 MHz. The two-frequency, two-parameter method involving the measurement of the permittivity and the natural logarithm of the loss factor, also provided a result which is well within the performance expected of commercial moisture meters. (C) 1996 Silsoe Research Institute
