How to write a technical specification for EEG, fNIRS, eye tracking and biosignal systems

A technical specification for research equipment is a document that defines the functions and performance a system must deliver as measurable, verifiable requirements, without pointing to a particular brand or model. A good specification gets you the device your research actually needs, lets more than one supplier bid, and settles from the outset what will be checked at delivery. This guide explains how to write a technical specification for EEG, fNIRS, eye tracking and biosignal systems. It is practical guidance, not legal advice; finalise your specification with your institution’s procurement unit and, where needed, its legal counsel.

01Function instead of brand: measurable requirements

A specification starts from the research question, not from a product catalogue. First, write in plain sentences what the system has to do: which signals, in how many participants, in what environment, with which stimuli and alongside which other devices will they be recorded? Then turn each function into a requirement that carries a unit, a threshold and a measurement condition. Instead of “high-quality 64-channel EEG”, a line such as “at least 64 EEG channels recorded simultaneously, at least a stated sampling rate per channel, and an upper limit on noise with the bandwidth over which it is measured” makes clear both what the bidder must offer and what the committee will check.

A well-written requirement has four qualities:

Measurable

It is expressed as a number with a unit (channels, Hz, bits, µV, dB, nm, degrees) and does not rest on adjectives such as “high” or “advanced”.

Verifiable

It can be proven at delivery by a test or by the manufacturer’s technical documentation; a requirement that cannot be verified causes disputes during evaluation.

Justified

It is tied to a need of the research; asked why it is required, you can answer with a paradigm or an analysis method.

Neutral

It can be met by the products of more than one manufacturer; it does not contain a name, design or unusual value found in only one model.

02What does Article 12 of Türkiye’s Public Procurement Law say?

For purchases made with public funds in Türkiye, the framework of a specification is set by Public Procurement Law No. 4734. Under Article 12 of the law, technical criteria “shall aim to ensure efficiency and functionality, shall not contain elements that hinder competition, and shall provide equal opportunity for all bidders” (our translation). The same article states that technical specifications shall also include provisions ensuring conformity with national and international technical standards where they exist, and it sets a clear rule: “No particular brand, model, patent, origin, source or product may be specified, and no features or descriptions aimed at a particular brand or model shall be included.”

The exception is narrow: where national or international technical standards do not exist, or where it is not possible to define the technical features, a brand or model may be named on condition that the words “or equivalent” are added. In practice this means that even if no brand is named, writing into the specification a registered name, a patented design or an unusual catalogue value found only in one manufacturer’s product can count as a description aimed at that model. In addition, some purchases for research projects are made under the exceptions in Article 3 of the law, with their own procedures; your institution’s procurement unit decides which procedure applies. Whatever the procedure, a function-based, neutral specification reduces both the risk of complaints and objections and the risk of buying the wrong device.

03Typical specification lines by device class

The table below shows lines that often appear in specifications for research equipment and how each can be phrased measurably. You set the values in square brackets according to the requirements of your research; the table proposes no thresholds, it only shows how to phrase them.

Keep two points in mind when using the table. First, every numerical value is meaningful only with a measurement condition: unless the bandwidth for noise or the distance and lighting for accuracy are stated, the claims of different manufacturers cannot be compared. Second, the sampling rate must be at least twice the highest frequency of interest (the Nyquist criterion), and in practice several times that is chosen; but asking for a high value merely because it is “better” can mean needlessly large files and more expensive devices. The recording details that the EEG publication guidelines (Keil et al., 2014) ask authors to report are also a good checklist of the values a specification should define.

In fNIRS, long channels in adults are typically placed at a source-detector distance of about 3 cm, while short-separation channels sit much closer and measure only the changes in the scalp and skull, which helps separate that signal from the brain signal. The fNIRS community’s publication guidelines (Yücel et al., 2021) offer a checklist for reporting method details such as optode geometry and signal processing; asking for the same details in the specification makes it easier to produce publishable data with the system you buy. In eye tracking, accuracy is the average offset of the measured gaze point from the true point, and precision is how consistent successive measurements are while looking at the same point (Holmqvist et al., 2012). Because manufacturers’ values are usually measured under ideal conditions, requiring the measurement in your own environment and with your own participants at acceptance gives a realistic comparison.

04Scope of delivery: installation, training, support and warranty

The services that come with delivery should be written as measurably as the device itself. Otherwise, very different scopes may be offered for the same amount and the bids become impossible to compare.

Installation

Whether installation is done on site, whether integration with the experiment computer and stimulus software is in scope, and which test marks installation as complete.

Training

For how many people, how many hours or days, on which topics (preparation, recording, data export, basic analysis) and in which language training is given; whether a certificate is issued at the end.

Technical support

Support channels, response time, remote access and how support continues after the warranty period ends.

Warranty

Warranty period and scope (amplifier, caps, electrodes and software separately), where repairs are done and whether a replacement device is provided during repair.

Software licences

Whether the licence is perpetual or time-limited, how many users and computers it covers, and for how long updates are included.

Documentation

User manual, technical datasheet, electrical safety and conformity documents for devices connected to participants (CE marking and a declaration of conformity with standards such as IEC 60601-1), and a description of the file formats.

Consumables and spare parts

The cap sizes, quantity of electrodes and gel supplied with delivery; for how long spare parts will be available.

To make evaluation easier, ask bidders to answer every line of the specification item by item, citing the relevant page of the supporting document. This practice, common in research equipment specifications in Türkiye, lets the committee match each claim to a document.

05Acceptance tests: what to verify at delivery

Acceptance tests are the concrete steps showing that the requirements of the specification are actually met at delivery, and the inspection and acceptance procedure relies on them. Writing the tests into the specification from the start prevents disputes at delivery. For a typical research system, the following steps can be considered:

1. Inventory check: comparing every delivered item (amplifier, caps, electrodes, cables, licences) with the list in the specification.
2. Noise recording: comparing the noise level in a short recording, with the inputs shorted or using the manufacturer’s test connection, against the stated value.
3. Known-signal test: showing that a signal of given amplitude and frequency from a signal generator is measured correctly.
4. Trigger and latency test: measuring the latency and jitter between trigger markers and the actual stimulus onset with a photodiode on the stimulus screen or with an audio signal.
5. Impedance and preparation: demonstrating electrode preparation and impedance measurement on a participant.
6. Data export: seeing that a recording can be exported in an open format and opened in independent software (such as EEGLAB or MNE-Python).
7. Wireless connection and battery: verifying the stated range and battery life during a recording.
8. fNIRS and eye tracking: showing that a signal is received from all fNIRS channels and the short-separation channels; for eye tracking, measuring accuracy after calibration in your own environment.

Open standards are a long-term safeguard for data formats. EDF is a simple and flexible format developed for exchanging and storing multichannel biological and physical signals; BDF is a 24-bit version of it. Requiring that recordings can be converted to community standards such as EEG-BIDS (Pernet et al., 2019) makes it easier for other researchers and joint projects to read the data years later.

06Common mistakes

Copying one manufacturer’s datasheet

This is the most common mistake. Taking over a product’s catalogue values, registered names or a design unique to that product steers the specification towards a single model, may conflict with Article 12 and increases the risk of complaints. Derive the values from your research needs and check that more than one manufacturer can meet them.

Unverifiable superlatives

Phrases such as “highest quality”, “state of the art” or “best on the market” cannot be measured, cannot be tested at delivery and cause disputes during evaluation. Write every quality with a unit, a threshold and a measurement condition.

Over-specifying

Asking for a channel count, sampling rate or feature the research does not need narrows competition and raises the cost. For every line, ask: “Without this value, could I not answer my research question?”

Under-specifying

Forgetting trigger inputs, raw data access, software licences or training means additional purchases after delivery. In particular, state synchronisation with the stimulus presentation software explicitly.

Giving no measurement condition

Lines that state no condition for noise, accuracy or battery life lead to values measured in different ways being set side by side.

Leaving services for later

If installation, training and support are not defined in the specification, bids include them in different ways and comparison becomes hard.

07Steps to follow when preparing a specification

1. Write the research question and the experimental paradigms to be used in one paragraph.
2. List the measurements each paradigm needs: channels and regions, sampling, triggers, accompanying devices, recording environment.
3. Turn these needs into requirements with a unit, a threshold and a measurement condition.
4. Check that the requirements are open to competition by comparing them with the technical documents of more than one manufacturer.
5. Add the scope of delivery, the acceptance tests and the requirement for an item-by-item compliance response.
6. Review the draft with your procurement unit and, where needed, with legal counsel.

This guide offers a practical framework and is no substitute for legal advice; the procedure to follow and the documents to require depend on your institution and on how the purchase is funded. For information and a quote on the EEG, fNIRS, eye tracking and biosignal systems in the catalogue, you can use the quote form.