When working with EMOTIV’s Flex 2.0 EEG system, one key technical detail often raises questions: the data is initially sampled at 2048 Hz but is later downsampled for transmission to 128 or 256 Hz. Why does this happen? And how does it affect the quality of the data — especially in real-world scenarios like motor task experiments?

Let’s explore the rationale behind this important design decision.

The Problem: Noise from Power Lines

One of the biggest challenges in mobile EEG systems is electromagnetic interference from the environment — particularly from electrical power lines and artificial lighting. These systems radiate energy primarily at the base power line frequency (either 50Hz or 60Hz, depending on the region), but not exclusively.

Due to the non-sinusoidal nature of current flow in power systems, significant energy is also radiated at harmonic frequencies — for example, 100Hz, 150Hz, 200Hz, and so on. These harmonics can overlap and distort actual brain signals during data collection, posing a serious challenge for accurate EEG analysis.

Why 2048 Hz Sampling?

To deal with this, Flex 2.0 samples all EEG channels at a high frequency of 2048 Hz. This high sampling rate gives a wide alias-free range (up to 1024 Hz, according to the Nyquist theorem) that preserves both the real brain signals and the noise sources in their true spectral positions. In other words, we get a clean, full-fidelity version of the signal before any processing occurs.

The role of Filtering

Once the signal is captured at 2048 Hz, EMOTIV applies a digital low-pass filter with a 43 Hz cutoff. This effectively removes most of the high-frequency noise, including muscle artifacts and other non-cerebral signals that are not typically useful in EEG research. Additionally, notch filters are built in at both 50 Hz and 60 Hz to directly suppress line noise from power systems, ensuring the remaining signal is clean and usable.

Why downsample?

After filtering, the EEG signal is downsampled to either 128 Hz or 256 Hz, as selected by the user. This serves three major purposes:

1. Prevention of Aliasing: Because filtering was done before downsampling, the signal no longer contains frequencies above the Nyquist limit of 64 Hz (for 128 Hz sampling) or 128 Hz (for 256 Hz). This eliminates the risk of higher-frequency noise folding back into the lower-frequency EEG band — a phenomenon known as aliasing.

2. Efficient Transmission: Transmitting raw EEG data wirelessly at 2048 Hz would be highly inefficient and require considerable bandwidth. Downsampling drastically reduces the amount of data being transmitted — by a factor of 8x or 4x — resulting in longer battery life and more stable wireless performance.

3. Preserved EEG Fidelity: EEG signals of interest (such as Alpha, Theta, Beta, and even low Gamma) typically occur well below 43 Hz. That means sampling at 128 or 256 Hz is more than sufficient to capture these brainwave patterns with high accuracy.
   
   

Real-World application in motor tasks

In experiments involving motor function, attention, or decision-making, the frequencies of interest generally remain under 40 Hz. Therefore, downsampling to 128 Hz or 256 Hz does not degrade the quality or usefulness of the EEG data for real-world behavioral research.

The initial high-frequency sampling and rigorous filtering ensure that the signals are clean and precise — and the downsampled data retains that quality for effective, real-time or post-hoc analysis.

Summary

EMOTIV’s decision to sample EEG signals at 2048 Hz and then downsample them is a deliberate and technically sound approach designed to balance data quality with power efficiency and wireless performance. By addressing the problem of environmental noise at the hardware and preprocessing level, Flex 2.0 ensures researchers receive high-fidelity data optimized for real-world EEG applications.

Whether you’re studying motor behavior, cognitive load, or emotional states, you can trust that the downsampled EEG from EMOTIV’s Flex 2.0 has been engineered for both accuracy and usability.