The Color of Money:
Magnetic sensing systems are good at identifying counterfeit currency and other negotiable documents. System accuracy depends on using small, sensitive magnetic sensing devices—such as giant magnetoresistive sensors.
Carl H. Smith and Robert W. Schneider, NVE Corp.
The magnetic fields from these particles are smaller than Earth’s magnetic field (in contrast to the fields from stripes on credit cards, which are considerably larger than Earth’s field). The small fields from these particles, however, produce signatures that, when read by magnetic sensors, can be used to identify the denominations of currency presented to point-of-sale devices, such as vending and change machines.
To construct a magnetic sensing system that can reliably meet the requirements of typical currency identification applications, you must follow good practices to manage circuit noise and magnetic biasing. Sensitive giant magnetoresistive (GMR) sensors are particularly useful in handling these elements for systems that identify currency and checks.
Typical Magnetic Media Detection Applications
Another application in which small magnetic fields are detected is the reading of magnetic ink character recognition (MICR) numbers. The stylized MICR numbers produce a unique magnetic signature when documents in which they appear (e.g., checks) are sorted at high speeds. An example of the output from a low-field GMR bridge sensor (i.e., a NVE AA002-02) suitable for this application is shown in Figure 1.
Dealing with Noise
Inherent Noise. The sensor and the sensing system, like all current-carrying conductors, produce inherent noise. Inherent noise can include such things as sensor and amplifier offset, thermal noise, and 1/f noise.
Thermal noise is associated with random thermal motions at an atomic level. Because the noise is uniform with frequency, the noise voltage in a given bandwidth is proportional to the square root of the resistance, temperature, and bandwidth. To minimize thermal noise, you can limit the bandwidth to the frequencies of the magnetic signal of interest and use small resistors. The resistance of the sensing resistor itself may be constrained by power and amplification considerations.
Resulting from point-to-point fluctuations of the current in the conductor, 1/f noise increases at low frequency and often dominates below 100 Hz. At the lowest frequencies, it’s not easily distinguishable from drift. Whereas thermal noise is independent of current and exists even without current, 1/f noise grows as current increases. Bandwidth limitation, especially on the low-frequency end, will reduce 1/f noise. As with any random noise source, averaging a repetitive signal will increase the SNR by the square root of the number of signals averaged.
Transmitted Noise. Sources of this type of noise include any voltages picked up by the circuit and any magnetic signals picked up by the sensor that are not part of the desired magnetic signature. Any time-varying magnetic field will not only produce a signal in the magnetic sensor but will also induce a voltage in any circuit loop. To minimize the inductive pickup, you must follow good circuit practices, such as minimizing potential circuit loops and placing amplification as close to the sensor as possible.
You can usually find stray 50–60 Hz magnetic fields in any industrial location. The increasing use of computers and other equip.ment with rectifier-fed capacitor-input power supplies results in nonsinusoidal currents that produce magnetic fields at harmonics of 50–60 Hz. Any moving or rotating magnetic material in equipment produces a time-varying magnetic field at frequencies characteristic of its rotational period. Transmitted magnetic noise sources are best minimized by filtering and by using magnetic shielding.
Instrumentation amplifiers are a good choice for use with low-field GMR sensors. When combined with an operational amplifier, you can easily achieve gains of several thousand. And you can incorporate into the circuit high-pass and low-pass filters formed from passive components to limit noise and to avoid saturation of the amplifiers by any offset or by DC magnetic signals, such as the Earth’s magnetic field. If 50/60 Hz noise is large enough to cause difficulties, a notch filter can be added.
Small effects (e.g., the magnetization of electrical components) can cause additional offsets when using high gain. Most surface-mount resistors have ferromagnetic nickel plating on their ends, and most battery casings are ferromagnetic. When in doubt, try picking up the component in question with a permanent magnet.
The Art of Magnetic Biasing
The simplest way of biasing is to pass the object to be magnetized over a permanent magnet and then transport it to the vicinity of the sensor. This works well in currency detection and reading MICR numbers on checks. The articles to be read are moved by a transport mechanism and passed over the magnetic sensor one at a time. A permanent magnet must be placed at some point upstream remote enough not to saturate the magnetic sensor. The bills or checks will then have their particles in a reproducible magnetic state when they reach the magnetic sensor.
Figure 2 shows the results of a magnetic trace from right to left across the right half of a new U.S. $20 bill.
The upright sections of the letters in the word TWENTY cause the multiple peaks in the center. The border surrounding the portrait causes the large peak on the right, and the frame surrounding the bill, the peak on the left.
The small size of GMR sensors offers the possibility of making closely spaced arrays of sensors to image a larger area rather than just obtaining a signature from a single trace along or across the material, as shown in Figure 2. Magnetic sensor arrays are used to achieve a magnetic image, which can then be used to obtain additional information encoded in the document or object (e.g., currency denomination).
Carl H. Smith is Senior Physicist and Robert W. Schneider is Director of Marketing at NVE Corp., 11409 Valley View Rd., Eden Prairie, MN 55344; 952-829-9217, fax 952-996-1600, email@example.com.