Magnetic cores of transformers, shunt reactors, electric motors, generators, rotating machines etc. exhibit very complex flux distributions. As it is well known, the variations of the local inductions, but also fluxes in normal and transverse direction tend to influence the core losses and the magnetostriction in a very significant way. For improvements and optimizations of the magnetic cores, aiming lower losses and audible noise, the evaluations of the local induction values are an inevitable task. Furthermore, the correct detection of the induction out of the rolling direction for assessment of the rotational magnetization is of a great significance for industry, in particular for transformer core manufacturers. Established methods of local induction measurement, such as drilling holes in the laminations and inserting pick up coils, are extremely laborious as well as destructive for the cores. In the current work, a novel tangential induction sensor was manufactured and is presented. With this sensor, non-destructive measurements, not only on the surface, but also in core interior should be possible. Contrary to the traditional tangential H-sensors, the current sensor possess a very thin nanocrystalline ribbon as a 'dummy' core. Around the ribbon a thin wire ( - 50 m) is wounded. Placed in a location of a laminated core with unknown induction B, the sensor is magnetized with Bs. In most likely case, both inductions deviate from each other, on one hand side due to the different permeabilities and on the other due to existing demagnetizing field. However, using a priori calibrated function B(Bs), the unknown induction can be evaluated. A very important part of the concept of the tangential induction sensor is the calibration of the sensor. The sensor can be calibrated in an Epstein frame, or in a region of laminated core of well known induction. However, in the current work, measurements not only in the rolling direction, but also for two other important directions were performed for assessment of rotational magnetization. Therefore, an individual sensor was not sufficient. For that aim, in the current work, a set of sensors was used placed on a thin kapton foil, attached to the surface of a hexagonal grain oriented sample. The sensors were placed in rolling direction, in transverse direction (TD) and in 30 angle to the RD. The calibration was performed by means of a Vienna Rotational Single Sheet Tester. As shown in a previous work, the resolution of the nanocrystalline sensors depend strongly on the used geometry. For that purpose, three sensor sets with different lengths and widths of the nanocrystalline ribbons are prepared, tested and compared with each other. In an individual set, all three sensors are of the same dimensions. The longest set of sensors tend to be not useful at all, even for low induction values B. Due to the high length of the sensor, the demagnetizing effect tend to be very weak, and it reaches saturation for B > 0.6 T, due to the extremely high initial permeability values of the nanocrystalline material (r - 100000) . The sensor set of middle length exhibits excellent resolution for low induction values B, but is not effective for a magnetisation above 1.1 T due to saturation effects. On the other side, the widest and shortest set of sensors tend to exhibit the best results, good resolution for high B, and sufficient resolution for low B. The presented sets of sensors were tested in a 3-phase, 3-limb transformer core. The sensor shows good results of the measured induction for the rolling direction. However, it was not very effective for the transverse direction, mainly due to saturation effects. For the interpretation of the achieved results, numerical simulations are presented. They are performed by a novel MACC (Multi-directionally non-linear magnetic equivalent circuit calculation) method, developed by our group.