13 June 2012

Signal to Noise Ratio

My last blog explained the difference between single-ended and differential measurements. To follow-up, I want to discuss another key concept in the quest to make accurate and consistent measurements: signal to noise ratio (SNR). I will also reiterate the importance of a differential measurement.

Signal to noise ratio is the ratio of the signal voltage to the noise voltage:

SNR = Vout/Vnoise

SNR is usually averaged, and a good SNR is dependent on the measurement being made and desired accuracy of the measurement.

Noise is usually specified as the root mean square amplitude of a voltage (VRMS). Noise can come from multiple sources, including the datalogger, machines, radio frequencies, and power supply lines. The noise inherent in the datalogger can be less than 1 mVRMS, or even 1 µVRMS. Machines, like motors, could introduce noise with amplitudes greater than 1 VRMS. Some dataloggers, such as those made by Campbell Scientific, filter out the power supply noise by filtering out any noise with a frequency of 50/60 Hz, but the machine and radio frequency noise can be more variable, so it can’t be filtered as easily. Sensors with large signals have more “wiggle” room, whereas sensors with µV-scale signals, such as thermocouples and Apogee SI-100 series sensors, will be much more affected by noise.

Noise can be partially filtered out when differential measurements are made because it will cancel itself out. A single-ended measurement doesn’t have the same luxury. Always grounding the shield wire and always using twisted pair wire can also help mitigate noise. The drain wire is connected to a metal shielding around the twisted pair wire and acts as a Faraday cage blocking electric fields. Twisted pair wire is designed to reduce noise by balancing the signal with an equal and opposite signal, which is then cancelled out by taking a differential measurement. A datalogger designed for low voltage, differential measurements can also help make all the difference when it comes to SNR.

Below is a table describing the best way to check for noise on different sensors from Apogee Instruments.

Sensor Method
SI-100 Series Cover the aperture with a small piece of foil and wait for the foil temperature to equilibrate with the sensor temperature. The target temperature should match the sensor body temperature and be stable to within 0.02 C.
SP, SQ, SU Series

MP, MQ, MU Series
Cover the sensor with a cap or dark cloth, so that no light reaches the detector. The reading should be zero and stable to within 1 mV.



Skif Smith

Electrical Engineer

No comments:

Post a Comment