Now that we have figured out the smallest signal we need to measure, let’s look at what the 16-bit ADCs of the MCP3901 can do for the energy measurement requirements. We need to look at more than resolution in order to determine if this is good enough for the application. If resolution were the only parameter to look at, we could simply calculate that over the input range of 1V, we would spread the 16-bits of resolution (65,536 codes) would give us a least significant bit, or LSB of 15.2µV. That sounds good, but does it give us our actual smallest detectable voltage?
Let’s look at some other parameters that could help us. One of these is distortion, which is typically characterized by the parameter THD or total harmonic distortion as introduced by the ADC that was not present in the original signal. Harmonic distortion results from spurious signals created from multiples of the fundamental frequency. It is specified as a ratio in dB of the highest harmonic spurious signal relative to the carrier frequency. The MCP3901 shows a specification of -104dBc THD. It would be smaller with a smaller input voltage. Similar devices from other suppliers typically have a THD of -90dBc or smaller. Naturally a larger value for THD is better, indicating a smaller value for the spurious harmonic signal.
Harmonic distortion is not the only parameter to look at to determining the accuracy of the ADCs. In order to include noise introduced by the ADCs, THD+Noise is used. A more common view of this parameter is the inverse of THD+Noise, called SINAD or signal in noise and distortion. The MCP3901 shows a SINAD of 91 dB. With a full scale input of 1V, the smallest detectable signal would be 28µV. Similar 16-bit resolution devices from other suppliers typically have a SINAD specification of 60-70 dB, indicating much higher noise. On the higher end of 70 dB, this would yield a smallest detectable signal of 316µV. This is a significant difference in accuracy between devices with the same resolution.
Looking at specifications other than resolution, it appears than specifications such as THD and SINAD can be very revealing with respect to the actual accuracy of the ADCs being used to make measurements. Looking at the specifications of the MCP3901 the 28µV detectable signal threshold with a 1V input signal is more than adequate for the 40µV application requirement. This is much better than what might be obtained from similar 16-bit resolution ADCs from other suppliers, which could not even meet the basic application requirements.
Related Links
