Smart Sensing with Adaptive Analog Circuits

How does Smart Sensing with Adaptive Analog Circuits relate to Analog and Power Systems?

In many power systems, the use of several types of sensors (e.g., Resistance Temperature Detectors, Thermistors, Thermocouples) is required for monitoring the system. These sensors generally have curves that relate, for example, resistance to temperature. These curves coupled with instrument precision do not always provide the most accurate output. Couple these sensors with Smart Sensing with Adaptive Analog Circuit and the output can be significantly improved.

This is by no means the only use of this technology it could also be used to take sensor input in and output control information for driving processes.

Can Smart Sensing with Adaptive Analog Circuits branch out to other disciplines?

Of course! One major application that these circuits can be applied to is Embedded Systems where computer processors use input and output devices/sensors to interact with the environment. Due to wireless technology, networks can be formed with sensor devices to monitor habitats, pollutants, earthquakes, and many more[3]. This data is important, therefore it needs to be relayed expediently and accurately. Having the Embedded Systems application makes mobility of these devices more feasible and cost effective for any user.

Adaptive circuit also allow for self-sustaining circuits [4], where the consumption of these power systems has the ability to prolong the battery life of these self-sustaining circuits. In the use of these, it allows designers to use these event-driven detectors to set and manipulate the analog systems to adapt to the given circumstances at the time of design and after the application of the overall system. This is used in systems mostly in sensing applications for sound and chemical information that is to be collected by scientists and engineers for uses mentioned in the previous paragraph.

What Challenges Does Smart Sensing with Adaptive Analog Circuits face today?

The challenge they faced in reference [1] was the selection of an appropriate learning algorithm. They settled on a neural network as opposed to a gradient descent algorithm (e.g., Logistic Regression). They stated that this method was selected because the weighting learned from the training data was a better fit for the validation data. They also imply that a great deal of simulation was required in software prior to final implementation in hardware.

One of the biggest challenges with any machine learning approach is getting large quantities of labeled data for training and validation. Utilizing large quantities of data will yield better results against testing data. After the base programming of these systems, the use of adaptive analog systems will adjust smartly and in real time to the incoming data sets. To do this though, it takes large amounts of learning or previously encountered sets to have the system to react accordingly in a predictable manner.

Another issue with obtaining large amount of data is the analog-to-digital (ADC) rounding errors [6]. As the data is coming from an analog source and there isn't efficient analog memory yet, all the data is stored into digital devices. This error decreases as your sampling resolution gets higher, but until there is no need to store the information digitally there will always be an additional source of error from the ADC.

What are some useful resources to learn more about Smart Sensing with Adaptive Analog Circuits?

[1] Zatorre, Guillermo & Medrano-Marqués, Nicolás & Celma, Santiago & Martín-del-Brío, Bonifacio & Bono-Nuez, Antonio. (2005). Smart Sensing with Adaptive Analog Circuits. 3512. 463-470. 10.1007/11494669_57. (https://www.researchgate.net/publication/221582936_Smart_Sensing_with_Adaptive_Analog_Circuits)

[2] Green, Rob. (2018). “How Nest Thermostat Learns Your Temperature Preferences?” https://roboauthority.com/how-nest-learning-thermostat-work/

[3] Zatorre, G., Medrano, N., Sanz, M., Aldea, C., Calvo, B., & Celma, S. (2009). "Digitally Programmable Analogue Circuits for Sensor Conditioning Systems". https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3297160/

[4] P. Mayer, M. Magno and L. Benini, "Self-Sustaining Acoustic Sensor With Programmable Pattern Recognition for Underwater Monitoring," in IEEE Transactions on Instrumentation and Measurement, vol. 68, no. 7, pp. 2346-2355, July 2019, doi: 10.1109/TIM.2018.2890187.

[5] A. A. Makinwa, K., Baschirotto, A. and Harpe, P., 2018. Low-Power Analog Techniques, Sensors For Mobile Devices, And Energy. Cham Springer.

[6] https://en.wikipedia.org/wiki/Quantization_(signal_processing)