M. Járó - L. Költő szerk.: Archaeometrical research in Hungary (Budapest, 1988)

Prospecting - SŐRÉS László: Geophysical measurements at the site of a Roman homestead at Balácapuszta

must be noted, however, that the directions indicated represent only one possible version. The collinear plots of vertical and horizontal direction in the figure do not show wall directions but the directions of the measuring profiles. A different linking of the anomalies is also possible; a factor that must be taken into consideration. The wall direction of area no. 2 was determined on the basis of the anisotrop}

. but this should be regarded as having an informative character only. The anomalies of area no. 3 can be correlated definitely. The gate-like building no. V cannot be reconstructed, but its presence is also justified by the maximum of the resistivity map. Appendix Filter planning and filter examination Geophysical profiles are extremely noisy, and interpretation in their original form would be very difficult. Based on the experience gained so far, a 3-point convolution filter provides a suitable result for filtering out the random noises. The coefficients of the applied smoothing filter are as follows: 0.25; 0.5; 0.25. It is more complicated to prepare such a filter, which enhances the effects presumably caused by walls or wall type buildings, from the resistivity map providing a blurred picture. The filtering procedure we applied corresponds to the cross correlation of the measffred profiles and the ones received from the theoretical model. A 7-point filter was made, which is the equivalent of the curve of Figure 2 normalized and transformed into zero sum. Elements of the resultant 7-point filter are as follows: -0.14; -0.17; 0.13; 0.39; 0.13; -0.17; -0.14. Let us examine some transfer characteristics of the filter. Since it is a high-pass filter of zero sum, it filters the frequencies near zero very efficiently. However, on the leading and trailing edges of a square wave it provides maxima. The number of filter elements is finite, which results in distortion causing the "swinging in" of the edges of the wall anomalies. The negative peaks of the filtered profiles are consequences of swinging in. In the case of "Dirac" pulse-like peaks, the function transforms itself; thus, as far as these peaks are concerned, it behaves as a low-pass filter. Detailed examination of the resolution of the filter was performed on models prepared by means of the theoretical curve. The result can be seen in Fig 4. The resistivity profiles were obtained by the convolution of the equivalent of zero sum of the theoretical curve and the "Dirac" impulse-series replacing the walls (p), thus ensuring the superposition of the effect of the individual walls. After that, smoothing filtering and then high-pass filtering followed; the result of this procedure is shown by the curves marked with "f". Smoothing filtering was applied because this is how the processing of field curves is performed - for the sake of noise suppression —, and the question is how the effects concerned appear in the measuring material. It can be seen from Fig 4 that, if the distance of the walls is 1-2 m, the filter cannot separate the anomalies merging into one another. With this wall distance the filter marks out the place of the outer walls (see curve 4), if their distance is smaller than the wavelength of the filter — as an elementary wave -, then only one maximum appears (curve 3). In such cases, it cannot be determined whether there is only one wall or more. Should the distance of the walls be greater than 3 m (curves 1 and 5),