Real-Time EEG System & Closed-Loop BCI: Low-Latency Neural Decoding, Synchronization, and Stimulation

From brain signals to stimulation in as little as 4 ms. Discover how g.tec EEG systems enable real-time brain-computer interfaces, precise synchronization, and advanced closed-loop neuroscience.

Real-time brain-computer interfaces (BCIs) are transforming neuroscience by enabling researchers to record, analyze, and respond to brain activity within milliseconds. From EEG-triggered transcranial magnetic stimulation (TMS) and functional electrical stimulation (FES) to neurofeedback, motor imagery, and adaptive neural decoding, closed-loop experiments require more than high-quality EEG signals.

They depend on low-latency acquisition, predictable processing, accurate synchronization, and reliable communication between hardware and software.

With EEG-to-TMS trigger latency as low as 4 ms and EEG-to-Stimulation with g.Estim FES or g.Estim PRO command latency as low as 5 ms, g.tec provides technologies for developing time-sensitive closed-loop research applications.

But what determines real-time EEG performance? How do latency, jitter, and sampling frequency affect experimental timing? And how can researchers build reproducible closed-loop BCI systems?

What Is a Real-Time Brain-Computer Interface?

A real-time BCI continuously acquires neural signals, processes them, extracts relevant information, and generates an output within a time frame appropriate for the application. In a closed-loop BCI, the output is used to provide feedback, control an external device, or trigger stimulation based on ongoing brain activity.

Typical applications include:

• EEG-triggered TMS: Detecting brain states and generating TMS triggers based on neural activity.
• Functional electrical stimulation: Using decoded EEG signals to control electrical stimulation through g.Estim FES.
• Adaptive neurostimulation: Generating stimulation commands based on real-time neural features.
• Motor imagery BCIs: Translating imagined movements into commands for external devices.
• Neurofeedback: Providing visual or auditory feedback based on ongoing EEG activity.
• Real-time neural decoding: Applying signal processing and classification algorithms to incoming brain signals.
• Multimodal neuroscience: Combining EEG with eye tracking, fNIRS, motion capture, and other measurement technologies.

The fundamental requirement is a reliable relationship between the recorded neural activity and the resulting system response.

Low-Latency Closed-Loop EEG: From Brain Signals to Stimulation in as Little as 4 ms

In closed-loop neuroscience, every millisecond can matter. The time required to acquire EEG signals, process neural information, and generate a stimulation command directly influences the temporal precision of an experiment.

g.tec supports low-latency closed-loop applications using its EEG amplifiers together with dedicated processing and stimulation technologies.

Closed-Loop Performance Specifications

These figures refer to specific EEG-to-trigger and EEG-to-command processing pathways. They should not be confused with the complete physical response time of an external stimulation device.

EEG-Triggered TMS with 4 ms Latency

Using g.HIamp or g.USBamp, researchers can achieve EEG-to-TMS trigger latencies as low as 4 ms under suitable processing conditions. This enables researchers to develop brain-state-dependent TMS experiments in which EEG activity is continuously monitored and used to determine when a stimulation trigger should be generated.

Potential applications include:

• Brain-state-dependent TMS
• EEG-triggered stimulation protocols
• Oscillation-dependent stimulation
• Studies of cortical excitability
• Investigations of neural dynamics and plasticity

By integrating EEG acquisition, real-time processing, and external TMS triggering, researchers can investigate relationships between ongoing neural activity and stimulation timing.

EEG-Controlled Electrical Stimulation with 5 ms Latency

When g.tec EEG amplifiers are combined with g.Estim PRO or g.Estim FES, EEG-to-stimulation command latency can be as low as 5 ms. This enables closed-loop experiments in which EEG activity is analyzed and translated into stimulation commands.

Applications include motor imagery-based stimulation, adaptive stimulation protocols, and research into neural feedback mechanisms. The total time until physical stimulation occurs also depends on the connected stimulation hardware, its configuration, and the experimental processing pipeline.

Understanding EEG Latency: Acquisition, Processing, and End-to-End Performance

Latency describes the delay between a defined input event and a corresponding output event. In real-time EEG systems, different processing stages contribute to the overall delay.

A typical closed-loop processing chain includes:

Each stage introduces its own timing characteristics.

EEG Acquisition Latency

Acquisition latency describes the time associated with making recorded EEG data available to the processing environment. It depends on factors such as sampling frequency, amplifier architecture, data buffering, and communication interface.

For example, g.HIamp technical specifications include raw data block availability in under 2 ms. This is an acquisition-related performance specification, not a universal measurement of complete closed-loop latency.

Processing Latency

Processing latency is introduced by the software algorithms operating on incoming EEG data.

Depending on the application, this may include:

• Digital filtering
• Artifact handling
• Frequency-domain analysis
• Feature extraction
• Machine-learning classification
• Decision logic
• Output generation

Processing time depends on algorithm complexity, data block size, computer performance, and software architecture.

End-to-End Latency

End-to-end latency measures the total delay across a defined processing pathway. For EEG-triggered TMS, the relevant pathway may extend from EEG acquisition to generation of the TMS trigger.

For electrical stimulation, it may extend from EEG acquisition to generation of a command for the stimulation device. The measurement boundaries must always be specified. An EEG-to-command latency does not necessarily include the physical activation delay of the external device.

Why Jitter Matters in Real-Time EEG and Closed-Loop BCI

Low latency alone does not guarantee reliable real-time performance. Jitter describes the variation in latency between repeated events or processing cycles. Two systems can have the same average latency while exhibiting different timing consistency. For experiments involving precise stimulus timing, variable delays can introduce uncertainty into the relationship between neural activity and experimental events.

Researchers should therefore evaluate:

• Average and median latency
• Latency standard deviation
• Timing jitter
• 95th and 99th percentile latency
• Maximum observed delay
• Long-term timing stability

These measurements provide a more complete understanding of real-time performance than a single average latency value.

Sampling Frequency and Timing Resolution

The sampling frequency determines the time interval between consecutive EEG samples.

• The sampling period is calculated as: Sampling Period (ms) = 1000 / Sampling Frequency (Hz)
• For multimodal synchronization, a timing-deviation criterion of less than one sampling period can be expressed as: Timing Deviation < 1 / Sampling Frequency (seconds)

For example, at a sampling frequency of 1,000 Hz, the corresponding synchronization criterion is a timing deviation below 1 ms. This criterion is useful when evaluating the temporal alignment of signals recorded by different systems. However, sampling period, measured jitter, synchronization accuracy, and end-to-end latency are distinct performance characteristics and should be reported separately.

Precise EEG Synchronization with Eye Tracking, fNIRS, TMS, and Other Devices

Modern neuroscience experiments frequently combine EEG with additional measurement or stimulation technologies.

Multimodal experiments may integrate:

• EEG and eye tracking
• EEG and functional near-infrared spectroscopy (fNIRS)
• EEG and motion capture
• EEG and TMS
• EEG and behavioral experiments
• Multiple EEG acquisition systems

The objective is to establish a reliable temporal relationship between data recorded by different devices.

How to Validate Multimodal Synchronization

A practical approach is to record the same physical event in both systems and compare the resulting timestamps. Researchers can then determine three important timing characteristics:

• Time Offset: The difference between corresponding timestamps recorded by the two systems.
• Jitter: The variability of the measured time offset across repeated events.
• Clock Drift: The change in synchronization offset over time, commonly expressed in milliseconds per hour.

For long recordings, evaluating clock drift is particularly important because independently operating devices may gradually diverge. A reproducible synchronization benchmark should document the sampling frequency, timestamping method, trigger mechanism, recording duration, and measured timing statistics.

Hardware Triggers vs. Software Synchronization

Hardware triggers provide a physical mechanism for recording or marking events and can serve as a timing reference. Software-based synchronization enables data streams from independent applications or devices to be aligned using timestamps. Lab Streaming Layer (LSL), for example, provides a framework for distributing and temporally aligning data streams in multimodal research environments.

The appropriate method depends on the experimental requirements. For timing-critical measurements, researchers should distinguish hardware-recorded events from software-generated timestamps and validate the accuracy of the complete synchronization pathway.

g.HIamp: High-Density EEG for Real-Time BCI Research

g.HIamp is a high-density biosignal amplifier designed for advanced neuroscience, brain-computer interfaces, and multimodal research. Supporting up to 256 channels per amplifier and multi-amplifier configurations of up to 1,024 channels, g.HIamp enables researchers to acquire large-scale neural datasets while developing sophisticated processing workflows.

Key Capabilities

• Up to 256 channels per amplifier
• Multi-amplifier configurations up to 1,024 channels
• Raw data block availability under 2 ms
• Synchronization specification under 1 µs
• Integration with real-time signal processing environments
• Support for advanced BCI and multimodal research applications

g.HIamp is particularly relevant for experiments requiring high-density EEG acquisition, external device synchronization, and advanced neural signal processing.

Real-Time EEG-TMS Research

In suitable EEG-triggered TMS configurations, g.HIamp supports EEG-to-TMS trigger latency as low as 4 ms. This makes it relevant for experiments investigating the relationship between ongoing neural activity and precisely timed stimulation triggers.

Researchers can combine EEG acquisition with processing algorithms designed to identify relevant neural features and generate TMS triggers when predefined conditions are met. The achievable latency depends on the configured processing pathway.

g.USBamp: Low-Latency EEG Acquisition for Closed-Loop Experiments

g.USBamp is a biosignal amplifier used in neuroscience, BCI development, and real-time experimental applications. It supports research workflows involving EEG acquisition, signal processing, external triggering, and integration with experimental software.

EEG-Triggered TMS with g.USBamp

Like g.HIamp, g.USBamp can achieve EEG-to-TMS trigger latency as low as 4 ms in suitable closed-loop configurations. This enables researchers to develop EEG-triggered stimulation protocols using ongoing neural activity.

g.USBamp can also be incorporated into research environments involving multimodal recording, event synchronization, and custom BCI processing. For each experiment, researchers should document sampling frequency, software configuration, processing algorithms, and external trigger behavior.

g.Nautilus: Wireless EEG for Real-Time and Multimodal Research

g.Nautilus provides wireless EEG acquisition for neuroscience, neurofeedback, and BCI research. Wireless acquisition is especially useful for experiments that benefit from reduced cabling and greater freedom of movement.

Potential applications include:

• Motor imagery experiments
• Neurofeedback
• Human-computer interaction
• Wireless EEG recording
• EEG combined with eye tracking
• EEG-fNIRS research
• Multimodal behavioral experiments

Synchronizing g.Nautilus with fNIRS or Eye Tracking

For experiments combining g.Nautilus with external measurement systems, synchronization should be validated across the entire recording. A shared physical event can be recorded in both systems to determine timestamp offset, jitter, and clock drift.

A synchronization criterion of less than one sampling period can be expressed as: Timing Deviation < 1 / Sampling Frequency

This provides a sampling-rate-dependent reference for evaluating temporal alignment. Actual synchronization performance depends on the devices, connection methods, timestamping architecture, and experimental configuration.

Unicorn BCI: Flexible EEG Development and Experimentation

The Unicorn BCI family provides EEG acquisition technologies for developers, researchers, and educators exploring brain-computer interfaces. Systems such as Unicorn BCI Core-8 and Unicorn Hybrid Black support EEG-based experimentation and application development.

Depending on the device and software configuration, researchers can develop applications involving:

• EEG signal acquisition
• Neural signal visualization
• Custom signal analysis
• Brain-computer interface development
• Interactive applications
• Neurofeedback experiments

Unicorn BCI systems provide an entry point for exploring EEG applications while allowing developers to build custom workflows around supported acquisition interfaces. For time-sensitive closed-loop experiments, researchers should evaluate the specific device’s data streaming behavior, software interface, processing pipeline, and synchronization capabilities.

Real-Time BCI Software: g.HIsys, g.Pype, MATLAB/Simulink, and Python

Hardware performance is only one component of a real-time BCI system. Researchers also require software that supports signal acquisition, processing, classification, and integration with external devices. g.tec provides software technologies for different stages of the EEG and BCI research workflow.

g.HIsys for Real-Time Signal Processing

g.HIsys supports the development of real-time signal-processing and BCI applications using MATLAB/Simulink.

Researchers can develop workflows involving:

• EEG acquisition
• Digital signal processing
• Feature extraction
• Classification
• Real-time decision logic
• External control and feedback

MATLAB/Simulink integration enables researchers to develop and evaluate signal-processing architectures within a familiar engineering environment. The achievable processing latency depends on the algorithm, block size, computational hardware, and complete software configuration.

g.Pype SDK for Python

g.Pype provides a Python-based environment for supported EEG acquisition and analysis workflows. Python is widely used in neuroscience for data analysis, algorithm development, machine learning, and reproducible research.

g.Pype enables researchers to work with supported g.tec acquisition technologies in Python-based workflows. It is important to distinguish these capabilities from dedicated low-latency real-time streaming and control interfaces.

g.Pype should not be treated as the dedicated real-time streaming solution for timing-critical closed-loop applications. Such applications should use the appropriate supported acquisition and real-time processing interfaces.

Integrating EEG with Open Research Environments

Many neuroscience laboratories use software ecosystems involving Python, MATLAB, Lab Streaming Layer, and specialized analysis frameworks.

When developing an EEG workflow, researchers should consider:

• Supported device APIs and SDKs
• Access to raw EEG data
• Streaming and buffering behavior
• Event markers and timestamps
• Integration with external analysis tools
• Synchronization requirements
• Processing latency
• Reproducibility of the complete experiment

The best software architecture depends on whether the objective is offline analysis, online visualization, real-time neural decoding, or closed-loop stimulation.

How to Measure and Validate End-to-End BCI Latency

A reproducible latency benchmark requires a clearly defined input event, an observable output event, and a reliable timing reference. Researchers should avoid comparing latency specifications without understanding how the measurements were obtained.

Step 1:  Define the Measurement Boundaries

Determine exactly which events define the start and end of the measurement.

For example:

• EEG input to data availability
• EEG input to classifier output
• EEG input to TMS trigger
• EEG input to stimulation command
• EEG input to physical stimulation onset

These measurements represent different performance characteristics.

Step 2: Document the Acquisition Configuration

Record the amplifier model, sampling frequency, channel count, data block size, and communication interface.

Step 3: Document the Processing Pipeline

Specify the filtering algorithms, feature extraction methods, classifiers, decision logic, and output interfaces.

Step 4: Use a Hardware-Based Reference

Where appropriate, use an independent timing reference, such as an oscilloscope or logic analyzer, to measure the relationship between input and output events.

Step 5: Measure Repeated Events

Repeat the measurement under representative experimental conditions.

Analyze average latency, timing variability, latency percentiles, and maximum observed delays.

Step 6: Evaluate Long-Term Stability

For extended recording sessions, measure synchronization offset and clock drift over time.

Step 7: Publish the Measurement Conditions

A useful latency benchmark should include the hardware configuration, software versions, measurement method, sample rate, processing pipeline, and resulting timing statistics. This allows other researchers to understand and reproduce the measurement.

Choosing the Right EEG System for Closed-Loop Research

Different applications require different combinations of acquisition hardware, processing software, and external interfaces.

The optimal solution depends on the experiment’s required channel count, acquisition environment, timing requirements, software architecture, and external device integration.

From Real-Time EEG to Advanced Closed-Loop Neuroscience

Real-time brain-computer interfaces require more than fast data acquisition. They depend on the interaction of signal quality, processing performance, synchronization, and reliable output control.

With EEG-to-TMS trigger latency as low as 4 ms and EEG-to-g.Estim command latency as low as 5 ms, g.tec provides technologies for developing low-latency closed-loop experiments.

High-density acquisition with g.HIamp, flexible EEG recording with g.USBamp, wireless research with g.Nautilus, and BCI development with Unicorn offer different options for building experimental systems.

Combined with appropriate signal-processing software and validated synchronization methods, these technologies support research into neural decoding, neurofeedback, brain-state-dependent stimulation, and multimodal neuroscience.

Develop your next real-time BCI experiment with g.tec’s EEG acquisition and closed-loop research technologies.

EXPLORE OUR REAL-TIME BCI SOLUTIONS