Measurement Techniques and Error Sources for Low-Power Nanotechnology and Other Sensitive Devices II

External noise sources[1] are often interference signals generated by motors, computer monitors, or other electronic devices. This noise can be controlled by shielding and filtering, or by removing or turning off the noise source. These noise[2] sources are usually at the power line frequency, so when making lock-in measurements, the test frequency should not be multiples or fractions of 60Hz (50Hz). This is achieved by making each measurement an integer multiple of the power line period in the DC reversal technique.
 
Thermal EMFs are generated when different parts of the circuit are at different temperatures and when conductors of different materials come into contact with each other. Thermal EMFs can be reduced by keeping all wires at the same temperature and using copper-to-copper connections where possible. Since it is not possible to guarantee that every part of the circuit is made of copper (the test object itself is usually not copper), measurement methods such as lock-in amplifiers and DC reversal techniques are often used to reduce thermal noise.
 
Test lead resistance[3] can also introduce errors in the measured resistance. To prevent lead resistance from affecting the measurement accuracy, the four-wire (Kelvin) method should be used for measurement.
 
1/f noise[4] is used to describe any noise that increases in amplitude at low frequencies. Noise of this nature can be seen in components, test circuits, and test instruments. Environmental factors such as temperature and humidity can cause this noise, or chemical processes in components such as "aging" and "drift" on the label can cause this noise. 1/f noise can be observed as current, voltage, temperature, or resistance changes.
 
In the above discussion, we will focus on 1/f voltage noise in measurement systems. This is because the interference to the test object or components in the test circuit is mainly this noise. For example, carbon film resistors mainly show 1/f resistance errors of 0.01% to 0.3%, while this resistance error is 1/10 of that for metal film and wirewound resistors, and the resistance error of semiconductors is somewhere in between.