A new determination of the numerical value of the gravitational constant G was carried out by means of an apparatus consisting of a fibreless torsion balance which was carried by the buoyancy of a floater in a liquid; the angular deflection of the balance was compensated with the aid of a control loop and an electrostatic torque transmitter; the reacting bodies were cylinders. The result obtained for G differs from the CODATA value by about +6 × 10-3 in relative value. No dependence on the materials or on the distance between the cylinders was found. The apparatus is described, the methods of evaluation and the results are reported, the possible sources of errors are discussed.

balance and test cylinders of about 1 kg (tungsten) orThe numerical value of the Newtonian gravitationai constant has been determmed us-cyhdrically shaped masses and a static torsion balance with electrostatic compensation of the gravitational induced forces. The experimental set-up is described and the results are announced there is a surpris&m deviation fiom the CCDATA value.We have finjshed an experiment for determining the value of the Newtonian gravitational constant G and, in addition, to make tests for the inverse square law with small distances ana for a supposed dependence on the materials. The appmtus was an improved version of that used in an experiment we described inThe basic element of our experiment is a static torsion balance w i t h a beam lengrh of 420 mm, combined with a dekce to compensate the gravitational torques by electrostatic forces.The torsion fiber usual applied has the functions of bearing. the balance and givh~g a restricting torque; in our device it is replaced with both a mercury buoyancy beand an electrcstatic system similar to \Pa..zU's quadrant eizctromerer. AS msasurz rix gravitational induced torques by this electrometer, we have to d b r a t e it; the calibration procedure is based on measurements of the variance of capacitances with the angle of deflection of the balance. An advantage of tlus method is &e simple way of reahzmg without hturbances of the zxperiment.The electrometer forms with a laser interferometer and with an electronic circuit an automatic control system, that keeps the balance in a k e d argqular position {see Fig. 1).-4s sources of the agavitatio~ forces we use cylinders with a mass of about 120 g and 50 mm diameter (made from glass c e m i c ) as the main piuts of the 120 g @ass ceramic), which are movzd in the tvironment of the balance. The cylinders are arranged with coincidmg axes, one at each side of the balance masses, so that at all times one is ;r, a near position, thz other one in a far position.The essenaai procedure of the measurements consists in an automaticdy controlled moving of the four test masses, each between two rest positions, by air pressure i n suitable shaped cavities, and in r e c o r m the resulting control voltages. The center-to-center distance between the two rest positions of one mass is about 105 mm in all cases, the distance to the balance masses is varied in various runs within a range of 37 mm to lY%m~ in #,Y near positions respectively i42 mm to 302 mm in the far positions.The main quantities we have to determine are the control voltages for the compensation. In order to evaluate the resuhng G we must measure many other quantities in addition, especially the masses and sizes of the test bodies and the torsion balance, the &stances and spacings, the a m voltages of the electrometer and the capacitances and angles with the cdbration procedure. So we derive advantages of the special laboratories of the PTB.The experiment has the following results:Within the sensitivity oE our method we cannot state any deviation of Newton's ...