Nonvolatile Electronics' NVS5B15 sensors were employed for this project.[NVE, 1994] NVE is at present the only commercial vendor of GMR sensors, although other electronics companies (for example, Honeywell and Motorola) are expected to offer GMR products in the next few years. GMR sensors detect a single vector component of an applied field, like Hall generators or pulsed inductive detectors, so three orthogonal sensors banks will be necessary for full 3D imaging capability. A single 2D array of sensors was selected for this demonstration where alternate sensors have orthogonal axes of sensitivity. The checkerboard layout of the 5x5 array of sensors is illustrated in Figure 2. In order to simplify interpretation of the magnetic images at the end of the report, the outputs from the sensors with vertical and horizontal axes of sensitivity are displayed separately, as illustrated on the right side of the drawing. The performance of the 5x5 sensor array has been compared with a 3x3 array (not shown) where all the elements have the same vertical axis of sensitivity. The 3x3 array uses the same printed circuit board layout as the 5x5 so that the effect of varying the sensor spacing by a factor of two could be determined.
A schematic of the imaging system is shown in Figure 3. The sensor arrays were interfaced with an electronics chassis that contained a 15V and 5V power supply. The outputs of the sensors were connected to a National Instruments 64-channel data acquisition card which was installed in a Pentium PC. National Instruments' LabView software was used to acquire the images that follow. The images are unprocessed beyond resizing and adjustment of the grayscale for printability. Since there are only 25 sensors per image, the data files are only 400 bytes (25 sensors x 16 bits per sensor) in size. Each image is an average of 1000 readouts of the full array during a 10-second period (acquisition rate = 100 Hz), although there is no reason that data could not be acquired much more rapidly (10-100 kHz). 1000 readouts of the array was decidedly overkill; images were not degraded by the averaging of smaller data sets.
A variety of ferrous objects were imaged. These included tools, bolts, nails, rebar and permanent magnets. All objects were imaged in their remanent magnetic state (i.e., no external applied field) except where otherwise specifically noted. Before an image was acquired, the no-object output of all the sensors was obtained using the PC. This background signal represents a combination of offsets in the sensors, the sensors' response to the earth's field (no magnetic shielding was used) and their response to magnetic objects in the laboratory where the data was acquired, e.g. rebar in the floor. This background signal was saved to a file and then subtracted from subsequent data. Objects to be imaged were placed typically 1.5 cm above the sensor array on a lexan stand. The falloff of the signal from the array with separation was studied by stacking firebricks between the array and the ferrous object. Larger objects such as rebar could be detected at a meter separation (signal:background ratio of 2:1) although there was no real image at that separation with the 12cm-square array used for this demonstration.