Olive 1 Electronics Project

A network connected cigarette smoke detector for car rental companies

Andrew Morris
Nov 1 · 8 min read

This project was undertaken by two partners, Andrew Morris, a product developer and software engineer, and Tony Goodsell, an electronics engineer and prototyping expert.

Olive 1 was designed to be a mobile, network connected cigarette smoke detector for car rental companies. It enabled the detection of someone smoking in a rental car. On detection of such an event, it would send a notification over the mobile network to be logged by the car rental company’s system. As well as acting as a deterrent, it would enable the rental company to recoup any losses from the hirer on return of the vehicle.

The goal was to sell them to car rental companies and get them installed in all of the roughly ~5 million rental cars worldwide.

Electronics Hardware

Sensors

Version 1 was designed with a single CO sensor, however after testing it was found that this could give false positives in the presence of carbon monoxide from car exhaust gasses. Version 2 added an IR sensor that enabled the additional detection of heavy particulate smoke. The combination of both sensors enabled a much more accurate detection of cigarette smoke in the cabin of the vehicle.

RF Components

The Sierra wireless AirPrime WISMO218 GPRS wireless modem was chosen along with an embedded SIM that enabled connection to the mobile network and remote data transfer.

Microcontroller

The PIC 18F14K50 microcontroller was chosen. This controller had nice low power demands and all the input/output and CPU requirements that were needed.

Power

Version 1 relied on an external power supply which meant that it needed to be hardwired into the vehicle. Version 2 added a onboard lithium-ion battery. Along with power efficiency improvements made throughout the system, the battery capacity was large enough to enable long device operation between charges (>1 year). This enabled minimal operational logistics to deploy and operate the device across an entire rental company’s fleet.

Software

The core of the software was written in assembly for the embedded PIC microcontroller. The program was a small one running to about 1,500KLOC. A small amount of AT command code was also written to control the embedded RF module. The program was designed to send an alert notification to a REST API endpoint if the cigarette smoke detection algorithm determined a positive cigarette smoke event.

Detection Algorithm

The cigarette smoke detection algorithm needed to be smart enough to avoid false positives from other sources of pollution in the air. It was observed that a cigarette smoking event had a unique characteristic signature that could be used to identify it from other sources of pollution. This signature consisted of the unique combination and concentration of CO and particulate smoke over a specific time period. This model was used to write the algorithm. Machine learning could have been used in a subsequent version to make this model and algorithm even more accurate.

Power Efficiency

A key requirement of the system was to achieve a battery life, under typical usage, of >1 year between charges. A number of optimisations were made to the software to achieve this goal. The RF module was only powered up when a smoke detection alert needed to be sent. The behaviour of the infrared sensor was managed for best power consumption. The IR optics were selectively powered up based on input from the CO sensor. All these things helped maintain power usage to an absolute minimum.

Industrial Design

Version 1

The first version consisted of a main unit that was designed to be installed behind the dashboard of the car and connected directly to the cars power supply. The external CO sensor would then be installed in an appropriate position in the cabin and wired to and plugged into the main unit.

The industrial design goal of this version was to create a seamless, flat aluminium cube with one port for both power and sensor in.

Version 2

The design of version 2 was driven by the main goal of creating a completely self contained device. This meant adding an internal battery to remove the dependency on hard wiring to the vehicle’s power. It also meant adding the CO sensor inside the main housing. These improvements would enable the device to be easily installed in a more accessible location such as behind the rearview mirror. The constraints of this new version lent itself best to a low profile injection moulded plastic housing.

Manufacturing Processes

To achieve the industrial design goals of version 1, a number of manufacturing techniques were explored. Some of these included CNC milling, aluminium die casting, friction stir welding and aluminium thermal spraying. The industrial design goals of version 2 were a lot simpler, requiring straightforward plastic injection moulding.

Custom CO Sensor

Due to the size constraints of version 2 and the desire to embed the CO sensor within the main housing, it was necessary to design a custom CO sensor that was small enough to do this. No CO sensor available on the market was small enough.

Version 1

The first version was a redesign of a standard liquid electrolyte sensor into a more compact form. This was a test version as the final design still required a much smaller overall form.

Version 2

To reduce the size of the new version further, a radically different approach was required. It was clear that the only way to achieve this would be by using microfabrication. This also required a new solid state design. This version if successful would become the world’s smallest CO sensor and the world’s first solid state CO sensor.

Three main models were chosen to pursue; a conductor model, a dielectric model and a semiconductor model.

Microfabrication Procedure

The fabrication procedure involved a 15 step process to fabricate 44 sensors on one sillicon wafer. The process shown below focuses on one individual sensor through the fabrication process. The sensors would then be cut out of the silicon wafer.

The process involved may different steps such as photolithography, thin film deposition, atomic layer deposition and e-beam evaporation. Some layers in the sensor were only 1 atom thick.

Custom Battery

In order to maximise the battery life as much a possible, it was necessary to maximise the size of the battery and use as much internal space within version 2’s enclosure as possible. To do this it was necessary to design a custom battery. A standard Li-ion battery design was chosen, along with the various specific materials and chemistries, as shown below.

Welcome to a place where words matter. On Medium, smart voices and original ideas take center stage - with no ads in sight. Watch
Follow all the topics you care about, and we’ll deliver the best stories for you to your homepage and inbox. Explore
Get unlimited access to the best stories on Medium — and support writers while you’re at it. Just $5/month. Upgrade