A low-cost and open-source olfactometer to precisely deliver single odours and odour mixtures

This paper presents a low-cost, fully open-source olfactometer that enables precise delivery of single odours and mixtures without requiring specialized facilities or skills, thereby democratizing access to controlled olfactory research across both animal and human studies.

The Gist

Imagine you are trying to teach a dog a new trick, or perhaps study how a human remembers a specific smell from their childhood. To do this scientifically, you need to be able to "turn on" a smell instantly, keep it there for exactly as long as you want, and then "turn it off" immediately so the next smell doesn't get mixed up.

This is surprisingly hard to do. Unlike a light switch or a speaker that can start and stop instantly, smells are like invisible ghosts. They float, they stick to walls, and they linger. If you try to blow a scent at a subject, you often end up with a messy cloud that takes forever to clear, making it impossible to know exactly when the smell started or stopped.

The Problem:
Scientists have been studying smell (olfaction) less than sight or hearing, partly because the machines needed to control smells are either:

1. Too expensive (costing tens of thousands of dollars).
2. Too rigid (you can't change how they work because the software is locked).
3. Too complicated (requiring a PhD in engineering to build).

The Solution: The "Lego" Smell Machine
Doyle and his team at the University of Cambridge built a new smell-delivery machine that solves all these problems. Think of it as a customizable, low-cost "Lego set" for smells.

Here is how it works, using simple analogies:

1. The Air Highway (The Plumbing)

Imagine a busy highway with three lanes:

• The Constant Lane: A stream of clean, fresh air is always flowing to the subject's nose. This is the "baseline."
• The Switch Lane: A valve acts like a traffic cop. When no smell is needed, it keeps the traffic flowing on the "clean air" path.
• The Flavor Lane: When a smell is needed, the traffic cop instantly redirects the air through a specific "flavor station."

2. The Flavor Stations (The Modules)

The machine has four "stations" (modules), like four different soda fountains.

• Each station has two bottles: one with pure water (solvent) and one with soda syrup (the actual smell).
• The Magic Trick: If you want to serve just "Cola," the machine opens the Cola bottle. But here's the clever part: if you want to serve "Cola" and "Lemon," the machine doesn't just open two bottles and stop. It opens the Cola bottle and a "water" bottle for the Lemon station.
• Why? This keeps the total amount of air pressure hitting the nose exactly the same, whether you are delivering one smell or a complex mix of four. It ensures the subject isn't reacting to a sudden blast of air, but only to the smell.

3. The Brain (The Electronics)

Instead of a fancy, expensive computer, this machine uses an Arduino (a tiny, cheap microcontroller you can buy at a hobby store).

• It's like the conductor of an orchestra. The computer sends a simple signal: "Open Valve A for 2 seconds, then close it."
• Because the software is open-source (free and shared), any scientist can tweak the code to make the machine do exactly what they need, whether it's a 1-second flash of scent or a 5-minute continuous cloud.

4. The Proof: Does it work?

The team didn't just build it; they tested it rigorously:

• The Robot Test: They used a super-sensitive electronic nose (a "mini-PID") to measure the smell. The machine delivered smells with the precision of a Swiss watch. It could turn a smell on in a quarter of a second and turn it off just as fast.
• The Mouse Test: They put mice in a box. The mice learned to ignore a boring smell (solvent) but immediately perked up and investigated when a new, interesting smell (methyl tiglate) was introduced. This proved the machine could deliver distinct, detectable scents.
• The Human Test: They asked humans to wear a mask and press a button when they smelled something. Humans could detect the smell quickly. Even when they mixed the target smell with "distractor" smells (like trying to find a needle in a haystack of scents), the machine delivered the mix perfectly, and humans performed exactly as science predicts they should.

Why This Matters

This machine is a game-changer because it democratizes smell research.

• Before: Only rich labs with big budgets could study smell precisely.
• Now: A student with a few thousand dollars and a little bit of patience can build this in their garage or a standard university workshop.

In a nutshell:
This paper introduces a cheap, open-source, and highly precise "smell remote control." It allows scientists to study the sense of smell in humans and animals with the same ease and precision that we have long been able to study sight and sound, finally giving this "secondary sense" the attention it deserves.

Technical Summary

1. Problem Statement

Olfaction is a critical sensory modality influencing emotion, memory, and behavior, yet it remains understudied compared to vision and audition. This neglect is largely attributed to the technical challenges of olfactory research:

• Complexity of Delivery: Unlike light or sound, odorants must be vaporized, transported, and actively cleared to prevent lingering, creating temporal delays and variability.
• Cost and Accessibility: Commercial olfactometers are expensive (tens of thousands of dollars) and often rely on proprietary software, limiting customization.
• Technical Barriers: Existing custom designs often require specialized engineering skills, access to machine shops, or complex fabrication, preventing widespread adoption in neuroscience and psychology labs.
• Specific Needs: There is a lack of affordable, scalable systems that can deliver both single odors and mixtures while maintaining constant air pressure (crucial for avoiding mechanical confounds) and are suitable for diverse setups (rodent to human, fMRI compatible).

2. Methodology

The authors developed a modular, open-source olfactometer designed for assembly without specialized facilities. The system comprises three main components:

• Air Supply System:

  • Uses a medical air compressor (or cylinder) split into a carrier line (continuous clean air, 1 L/min) and a regulator/odor line.
  • A 3-way solenoid valve switches the regulator line between a "clean air" path (regulator line) and an "odor" path.
  • The system maintains a constant total output flow (1.1 L/min) by balancing the carrier line with either the regulator line (baseline) or the odor line (stimulus).

• Modular Odor System:

  • Features four independent modules (expandable), each containing a paired solvent bottle and odorant bottle.
  • Solvent Balancing: A key design feature where, if an odor bottle is not active, its paired solvent bottle is opened. This ensures the number of active airflow channels remains constant regardless of whether a single odor, a mixture, or a control is presented, thereby stabilizing output pressure.
  • Uses standard 40ml glass vials with screw-top lids and 19G needles for inlet/outlet.

• Electronic Control System:

  • Controlled by an Arduino Mega 2560 R3 microcontroller running open-source code (Arduino IDE).
  • Uses solenoid valves (2-way for individual bottles, 3-way for line switching) driven by custom "spike-and-hold" circuits on breadboards.
  • Power is supplied via 12V converters regulated to specific voltages (7V and 2.8V) for valve actuation.
  • The entire system is mounted on a wheeled trolley for portability.

• Validation Methods:

  • Technical Validation: Used a mini Photoionization Detector (miniPID) to measure odor concentration, latency, amplitude, and decay. Tested variables included tubing length (10cm to 4m), stimulus duration (100ms to 5min), air dilution ratios, liquid dilution, and mixture complexity.
  • Biological Validation:
    • Rodents: Habituation-dishabituation task in mice (investigation time of odor port).
    • Humans: Psychophysics tasks measuring detection latency and target detection accuracy in odor mixtures with increasing distractors.

3. Key Contributions

• Open-Source Design: Full schematics, parts lists (BOM), assembly instructions, and control software are provided, allowing any lab to replicate the device for approximately £1,500 / $2,000 (excluding air source/computer).
• Pressure Stabilization: The "solvent balancing" mechanism ensures that mixing odors does not alter the total airflow pressure, a critical factor often overlooked in mixture studies that can confound behavioral results.
• Modularity and Scalability: The system supports up to four odors simultaneously (expandable) and can be easily adapted for different experimental needs (e.g., fMRI compatibility due to non-metallic components in the delivery path, allowing the unit to be placed outside the scanner room).
• Versatility: Validated for use with both freely moving/head-fixed rodents and human psychophysics, supporting stimuli from milliseconds to minutes.

4. Key Results

• Temporal Precision: The system delivers odors with high trial-to-trial stability. Odor onset latency is consistent (~250ms) and scales linearly with tubing length, allowing for predictable delays in setups like fMRI.
• Duration Control: Reliable delivery was achieved for durations ranging from 250ms to 5 minutes. Short durations (<250ms) required a "pre-pulse" (opening the 3-way valve before the odor valve) to establish airflow, but the system handled standard behavioral durations (seconds to minutes) robustly.
• Concentration Control:
  • Air Dilution: Systematic changes in the carrier-to-odor flow ratio produced predictable, graded changes in signal amplitude.
  • Liquid Dilution: The system successfully detected dilutions down to 0.0001% (with signal averaging).
• Mixture Delivery:
  • Without solvent balancing, adding more odor channels increased output pressure.
  • With solvent balancing, pressure remained constant (~400 Pa) across all mixture complexities (1 to 4 components), while the odor signal amplitude increased proportionally with the number of active odorants.
• Biological Validation:
  • Mice: Successfully demonstrated habituation to solvent and dishabituation (increased investigation) upon presentation of a novel odor (Methyl Tiglate).
  • Humans: Participants showed consistent detection latencies (~2.4s) for single odors. In mixture tasks, detection accuracy significantly decreased as the number of distractor odors increased, replicating established psychophysical findings.

5. Significance

This work significantly lowers the barrier to entry for olfactory research. By providing a low-cost, fully open-source, and technically validated solution, the authors enable more laboratories to conduct high-precision olfactory experiments without relying on expensive commercial equipment or specialized engineering support.

The system's ability to deliver stable pressure during complex odor mixtures addresses a major confound in olfactory neuroscience, allowing for more rigorous studies of odor de-mixing and perception. Furthermore, its portability and compatibility with various setups (from simple behavioral arenas to fMRI environments) make it a versatile tool for exploring the neural and behavioral mechanisms of olfaction in both animal models and humans. The authors encourage the community to adapt, improve, and share modifications, fostering an open-science approach to sensory research.