> For the complete documentation index, see [llms.txt](https://docs.anthriq.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.anthriq.com/xbud/experiments/surface-emg.md).

# Surface EMG

Surface electromyography records the electrical activity of muscle through the skin. Unlike EEG, it is large enough to see directly: a resting forearm sits at tens of microvolts, and a firm grip produces bursts of a millivolt or more.

That visibility makes EMG the best second recording after alpha. Students see cause and effect in the time domain without any spectral analysis, and the signal-to-noise ratio is forgiving of imperfect electrode contact.

Allow about 15 minutes.

> **Note:** Complete [Get started with xBud](/xbud/technical-guide.md) first. This guide uses the EMG armband rather than the EEG cap.

## Prepare the subject

Fit the armband over the forearm flexors: the muscle group on the palm side, about a third of the way down from the elbow. Ask the subject to make a fist and the belly of the muscle becomes easy to find.

| Signal    | Placement                                                                   |
| --------- | --------------------------------------------------------------------------- |
| EEG 0     | Over the belly of the flexor group                                          |
| EEG 1     | A few centimetres along the same muscle, in line with the fibres            |
| REF       | A bony landmark with little muscle under it, such as the wrist or the elbow |
| AGND, DRL | The other forearm, or a bony point on the same arm                          |

Disc contacts suit bare skin here; the comb contacts are for hair. Seat each contact firmly.

> **Tip:** Place the two recording electrodes along the muscle fibres rather than across them. A differential amplifier sees the potential travelling along the fibre, so alignment directly affects amplitude.

## Check the signal

Open the **Real Time Plot**. EMG needs no spectral view to confirm it is working.

1. With the arm relaxed and supported, the trace should be close to flat.
2. Ask the subject to clench a fist. The trace fills with a dense burst of activity.
3. Ask them to relax. The burst stops within a fraction of a second.

If the resting trace carries a thick 50 Hz band, improve contact before continuing. If nothing changes on clenching, the electrodes are probably off the muscle belly or the pair is aligned across the fibres rather than along them.

Filter to roughly 20–300 Hz for viewing. The high-pass removes movement artefact and the slow baseline shift that comes with arm motion; xBud's front-end already band-limits at 300 Hz.

## Run the experiment

Record a sequence of contractions with markers, so the bursts can be aligned afterwards.

1. Start the recording with the arm relaxed. Hold 10 seconds of rest as a baseline.
2. Ask for a light grip, held 5 seconds, then rest 5 seconds. Press the marker button at the start of each contraction.
3. Repeat with a medium grip, then a maximum grip, 5 seconds each with 5 seconds of rest between.
4. Stop and save as `sub01_emg_grip`.

For a fatigue run, ask for a sustained grip at about half maximum, held as long as the subject can manage, with a marker at the start and at the point they report fatigue.

> **Tip:** Build this in the [Experiment Builder](/bxi-studio/experiments/experiment-builder.md) with text epochs prompting each grip level and markers on each stimulus epoch. The prompts keep timing consistent between subjects, which matters if a class is pooling results.

## Read the result

In the Real Time Plot, each contraction is a burst whose amplitude grows with grip force. The relationship between force and amplitude is the core teaching point: more force recruits more motor units and drives them faster, so the summed surface potential grows.

| Compare                         | Expected                                                              |
| ------------------------------- | --------------------------------------------------------------------- |
| Rest against contraction        | Baseline of tens of µV against bursts of hundreds of µV to millivolts |
| Light against maximum grip      | Burst amplitude rises with force                                      |
| Onset timing against the marker | Activity precedes visible movement                                    |

### What to expect

Surface EMG amplitude depends strongly on electrode placement, skin condition and the individual, so absolute values vary widely between subjects. The pattern is what to check against.

| Measure                   | Expected                                                |
| ------------------------- | ------------------------------------------------------- |
| Resting baseline          | Tens of µV; a flat trace with no visible bursts         |
| Light grip                | A few hundred µV, clearly above baseline                |
| Maximum grip              | Around a millivolt or more, often 10× the resting level |
| Onset to visible movement | EMG leads the movement by tens of milliseconds          |
| Burst offset on release   | Activity stops within a fraction of a second            |

At xBud's fixed 100× gain a 1 mV burst arrives at the DAQ as 100 mV, which sits comfortably inside every supported input range. If bursts look clipped, the problem is the DAQ input range rather than the amplifier.

> **Note:** This module is written for the armband rather than for a specific published recording. The figures in the alpha and SSVEP experiments come from EEG sessions on this montage; the equivalent EMG reference recording has not yet been made, so the table above states what the hardware and the physiology imply rather than what one particular session measured.

For a quantitative read, export the recording and compute the RMS amplitude in a sliding window. That converts the raw burst into a smooth envelope that tracks force, which is what gesture classifiers use as an input feature. See [Exporting a dataset](/bxi-studio/streams/exporting-a-dataset.md).

## Take it further

**Fatigue.** During a sustained contraction the spectrum shifts downward as conduction velocity falls, even while amplitude holds or rises. Open the **Time Frequency Spectrogram** during a long hold and watch the band of activity drift toward lower frequencies.

**Antagonist pairs.** Add a second electrode pair over the extensors on the back of the forearm. Flexion and extension then activate opposite channels, which demonstrates reciprocal control.

**Gesture classification.** Record several distinct gestures with a marker per gesture, extract features per burst such as RMS, waveform length, and zero-crossing rate, then train a simple classifier. This is the standard route from a teaching recording to a working prototype.

> **Note:** xBud's fixed 300 Hz low-pass bounds the recorded band. Surface EMG is often quoted as 20–450 Hz, but the content above 300 Hz carries poor signal-to-noise at the skin, and amplitude, burst detection, and the usual time-domain features sit well below it.

## Next steps

* [Measure CMRR](/xbud/experiments/measure-cmrr.md): characterise the amplifier itself
* [Choose a visualization](/xbud/technical-guide.md#choose-a-visualization): the spectrogram view for fatigue