Every honest CSI project starts by being clear about the tool in your hand. A single ESP32 reading Channel State Information is a real sensor, not a toy — but the gap between what it does and what the headlines imply is wide. If you have read our complete guide, this is the companion that draws the line and explains exactly why it sits where it does.
01What one antenna is really measuring
Start with the raw fact. A single-antenna link gives you one CSI reading per packet: amplitude and phase for each subcarrier, measured at the one receive antenna. That reading is the sum of every path the signal took from transmitter to receiver — the direct line plus every bounce off walls, floor, furniture and people, all collapsed into one measurement.
The crucial word is sum. The antenna cannot pull those paths apart. It cannot say this much came from the left, that much from a reflection two meters out. It only reports the combined result. Almost everything a single link can and cannot do follows from that one constraint, so hold onto it as you read the two lists below.
02What a single link genuinely can do
Within that constraint, a lot is still possible, because change over time survives even when the paths are summed together. When something moves, the combined reading changes, and that change is information.
- Presence. A person moving in the link area shifts the CSI well above the empty-room baseline. Reliable yes-or-no occupancy, covered in presence detection.
- Motion amount. How much the reading is changing scales with how much movement is happening — the basis for alarms and coarse activity sensing.
- Breathing. Someone sitting still produces a tiny, slow, periodic change as their chest moves, typically in the 0.1 to 0.5 Hz band. With clean data you can pull that rhythm out, as we show in can Wi-Fi CSI measure your breathing.
- Rough distance. Closer motion disturbs the channel more than distant motion, so you get a loose sense of near versus far — not a measured range.
These are not consolation prizes. A camera-free presence sensor, a motion alarm or a sleep-breathing monitor for your own space are real, genuinely useful builds, and a single link does them well. The mistake is expecting the next list for free.
03What it cannot do
Here is the other column, and it is just as important to say out loud.
| Claim | Reality with one antenna |
|---|---|
| Exact position on a floor plan | No — a distance is a circle; position needs several anchors |
| Direction a person moved | No — a single summed reading carries no bearing |
| Body pose or a skeleton | No — that needs many spatial measurements, not one |
| A through-wall image of a person | No — imaging needs antenna arrays and heavy compute |
| Identifying who the person is | No — not from one link; coarse motion is not identity |
When a clip appears to show any of these from a single board, one of three things is true: there is more hardware than shown, the result is heavily staged for one fixed scenario, or the claim is simply wrong. The physics in the next two sections is what makes that confident statement safe to make.
04Why: no spatial diversity
The first reason is geometric. To know where something is, or which way it moved, you need to compare the signal across points separated in space. Two antennas a known distance apart see a reflection arrive at slightly different phases, and that tiny difference encodes direction — this is how angle of arrival works. Several anchors at known spots each give a distance, and the distances cross at a position — this is trilateration. Both depend on having more than one viewpoint.
A single antenna has exactly one viewpoint. There is nothing to compare it against, so direction and position are not merely hard — they are absent from the measurement. No amount of clever software recovers a dimension the sensor never recorded. This is the same reason indoor localization always calls for multiple anchors.
05Why: noisy phase
The second reason is practical. CSI has two parts, amplitude and phase, and in theory phase is the richer one — it carries timing and, with several antennas, angle. In practice the raw phase from a commodity Wi-Fi chip is badly corrupted. The transmitter and receiver run on separate clocks that are never perfectly aligned, and each packet picks up offsets from carrier and sampling timing that can swamp the small phase change you actually care about.
Cleaning phase is a real discipline, not an afterthought — removing offsets, unwrapping, referencing against a stable subcarrier. We cover it in cleaning Wi-Fi CSI data and in amplitude versus phase in CSI. The takeaway for limits is simple: the most information-rich part of the signal is also the hardest to trust on cheap hardware, which caps what a single antenna can squeeze out.
06When to add antennas or nodes
Limits are not a reason to give up; they are a map of when to scale up. The rule is clean: a single link is enough whenever your question is about whether and how much movement. Add hardware the moment your question becomes where, which way or how many.
| Your goal | What you need |
|---|---|
| Presence, motion alarm, breathing | One link is enough |
| Direction of movement | Multiple antennas on one node |
| Position on a floor plan | Several nodes as anchors, at known spots |
| People counting beyond a rough guess | Several links crossing the room |
| Anything like imaging | Antenna arrays and research-grade compute |
So the honest workflow is to start with one board, build something that genuinely works inside its limits — a CSI motion alarm is the classic first project — and only then add nodes when a real need for direction or position appears. That order saves you from chasing capabilities the hardware never had, and from the disappointment of a demo that promised an image and delivered a wiggle.
Everything here is framed for your own devices and your own space. Understanding the limits is not only good engineering — it is also what keeps expectations honest and claims about watching other people firmly in the realm of the impossible, where a single antenna leaves them.
07FAQ
What can a single-antenna Wi-Fi CSI setup actually do?
A single link reliably detects presence and motion, measures how much movement is happening, can pick up the slow rhythm of breathing from someone sitting still, and gives a rough sense of how close that motion is. These are genuinely useful, and they are the honest ceiling of what one antenna delivers.
Why can one antenna not tell where a person is standing?
Because a single antenna has no spatial diversity. It measures the channel summed over every path the signal took, with no way to separate the direction or the distance of any one reflection. A distance alone is a circle of possible spots, and position needs several viewpoints to pin down, which one antenna cannot provide.
Can a single ESP32 see a person through a wall like a camera?
No. Wi-Fi does pass through light interior walls, so motion next door can change the CSI, but one antenna senses only that something moved, not a shape or a pose. Imaging a person needs arrays of many antennas, wide bandwidth and heavy processing. A pocket board gives a motion signal, never a picture.
When should I add more antennas or nodes?
Add them when your project needs direction, position or a count rather than plain motion. A single link is enough for presence, a motion alarm or breathing. The moment you want to know where a person is, which way they moved, or how many there are, you need multiple antennas or multiple nodes placed around the space.