Designing and Testing a noncontact temperature breathing sensor

A project log for NeuroCuddl

Biofeedback based Virtual Reality system for mental wellness

Vignesh RavichandranVignesh Ravichandran 08/19/2018 at 15:510 Comments

In this log I want to expand on what I posted last time regarding regarding breathing sensing. While I documented most methods available and the ones we tried, I want to talk here about our trials and tribulations in building our respiration sensor using a noncontact temperature sensor. 

We started off by using a standard MLX90614 noncontact temperature sensor which gave us good initial results. 

The sensor was reliable but a bit large for it's functionality. Melexsis, the company which makes these amazing sensors had a very compact SMD sensor (MLX90632) as a replacement to the MLX90614. This sensor according to their material is meant to be used in even high accuracy applications and offers 0.1 Celcius resolution and increased reliability. This sensor like it's bigger brother talked over the I2C bus.

Sparkfun used to make a handy breakout board for it too for people are interested. The size and accuracy made us choose this sensor to use in our project. We quickly made an equally teeny breakout board for this sensor with 0.1" headers.

After a quick order on oshpark, we got the first board of this project.

We got to assembling this sensor using my hot air gun using some solder paste.

Despite having some issues originally in getting the solder under the sensor to reflow a bit of kapton and preheating the board solved the problems. 

For testing I had this setup with the sparkfun library for MLX90632 and used the serial plotter tool to check how this sensor stacks up.  Further to remove any issues due to environmental factors I normalize each value of temperature with the IC temperature to pick up only external changes that I was interested in.

While sensor is able to get the trends the sampling rate was a bit too low (2Hz) even though a fast mode exists in the documentation there is no reference to register values for enabling this. For starters this result looks good but there warrants an exploration into various noncontact sensors.

Next build log we'll talk about some Initial sketches and design constraints, form factor requirements of NeuroCuddl.

Anyways, Happy Hacking folks.