544 m
1,785 ft — deepest delivery
A complete 5 KB payload, through timber, at peak summer leaf-out.
Every figure on this page was measured in the field, on the same radio stack that ships in the product. What we have not measured yet is listed at the bottom of the page, not filled in from a datasheet.
A complete 5 KB payload, through timber, at peak summer leaf-out.
On the ~410 m link across open water, antenna at 6 ft, in the rain. Six of six probes delivered and acknowledged.
Median for a 5 KB crop on a quiet channel — and it is the same 20 s at 22 m as at 426 m.
Measured canopy path loss in dense summer woods. Open ground is closer to 20.
A handheld probe walks the property and pushes an alert-sized payload onto the mesh at each stop, GPS-stamped, over the production Reticulum-on-LoRa stack. It is the same survey tool that ships with a research pilot, and its probes land in the same app your photos do.
Every probe is recorded at both ends. The gateway logs what it heard from the handheld; the handheld logs what it heard back. The two disagree: the direction going out to the camera is weaker, by a measured 13 dB. So every margin on this page is that downlink margin. Quoting the uplink instead would roughly double every headroom figure here.
Range is read from the gateway's database, not the handheld's success light. At the edge, the payload often lands while the receipt for it does not, so the device in your hand reports a failure for a photo that arrived intact. That happened on two of five edge "failures" on the second survey.
The dataset so far: 59 field probes across two sites. 27 pushed through dense summer timber at a wooded lake site over two surveys in August, and 32 from an earlier survey over mostly open ground, which anchors the near-line-of-sight comparison. Every figure below traces to a specific logged probe, and we will keep publishing as the dataset grows, including the results that cut against us.
The fit is unusually clean: five probes spanning 30 to 360 m and a 56 dB range, every one within 3.2 dB of the line. Dense summer canopy costs 50 dB per decade of distance, against 23.6 dB/decade over mostly open, near-line-of-sight ground at the comparison site. Doubling your range costs 15 dB in these woods and about 7 dB in the open.
Taken at face value, the curve says the radio reaches 890 m (0.55 mi) before it runs out of sensitivity. Do not plan a deployment on that number. The same surveys found obstruction shadows 20 dB deep, and one sat at 162 m, well inside the working radius. Walked in order, the link looked dead at 162 m and came back 12 dB stronger 135 m further out.
| What you assume about shadowing | Standard mode | Robust mode |
|---|---|---|
| Nothing — the mean curve alone | 887 m (2,910 ft) | 1,169 m (0.73 mi) |
| A 10 dB shadow can happen anywhere | 560 m (1,837 ft) | 738 m (2,421 ft) |
| A 20 dB shadow can happen anywhere (the case we measured) | 354 m (1,161 ft) | 466 m (1,529 ft) |
So we plan camera placements at this site around ~350 m through timber, not 890 m. The gap between those two numbers is what the robust radio mode is for: at 360 m the mean signal is comfortable, and a 22 dB hole drops it below what the standard mode can hear but still inside what the robust one can.
The surprise of the second survey: group every probe in the 395–545 m band by compass bearing, and bearing explains more of the signal than distance does.
| Bearing from the gateway | Distance | Signal at the gateway |
|---|---|---|
| Across open field and water | 395–426 m | −98.8 dBm |
| Along the lake's east side | 406–506 m | −112.5 dBm |
| Due north through timber | 374–544 m | −116.0 dBm |
17 dB between the wooded bearing and the open-water bearing, at the same range, on the same afternoon. At this site's canopy slope that is worth more than doubling your range: turning a link 15° off a good bearing costs more than moving it twice as far away.
Small moves matter too. Two probes at the same bearing, same distance, same height, 20 m apart, came in 11 dB apart, a bigger spread than the entire distance effect across those six probes. This is why a survey probe ships with every pilot instead of a coverage-radius promise: walk the spot, press the button, put the post where the meter says.
Rain did not matter. The entire second survey ran in light rain and no penalty shows anywhere in the data. At 915 MHz, over these distances, falling rain costs thousandths of a dB.
A MeshCam camera always captures and keeps full resolution. What crosses the radio first is a tight crop of the animal, a few kilobytes, and the bigger versions follow on request. So "how fast" has three answers, and only the first one is a number you wait for.
| What arrives | Size | Measured time | Conditions |
|---|---|---|---|
| Alert — the detector's crop of the animal, plus what it is and when | 5 KB | 20.0 s | median of 15 field probes, quiet channel 20 s was also the mode: identical at 22 m and at 426 m |
| The same alert, on a channel shared with two chattering nodes | 5 KB | 59.1 s | median of 18 probes at the second site |
| Detail — a bigger crop or the full frame, pulled on request | 16 KB | 97–131 s | two chunks of one 32 KB transfer, on the busy
channel we have not yet run this size on a quiet one |
| Archive — the untouched full-resolution original | — | instant | never crosses the radio unless you ask; it is on the camera's own SD card from the moment of capture |
Distance is not a row in that table because it never showed up in the data. On a quiet channel, 20 s was the ordinary case at every range we tested. What slows a photo down is other radios talking on the channel.
The two sites ran the same protocol against a busy channel and a quiet one, which makes them a clean A/B on the one variable that is usually impossible to isolate.
| Channel | Probes | Median | Mean |
|---|---|---|---|
| Two other nodes checking in every ~30 s | 18 | 59.1 s | 58.0 s |
| Quiet | 15 | 20.0 s | 26.9 s |
A 2.95× airtime tax, from two neighbours. That is a bigger effect than transmit power, antennas, or radio-mode tuning can buy back, which is why MeshCam nodes stay silent between events by design. It also makes latency a per-site number: your neighbours' behaviour sets it more than your terrain does.
This is the part a spec sheet leaves out, so it gets its own heading.
Not measured, so no number yet. The design budget is under 50 microamps asleep, and an indirect bench check on a pilot board is consistent with that, but a resistance reading is not a current measurement. An ordinary bench multimeter cannot make this one: its own shunt browns the board out the instant the radio wakes to transmit.
The right instrument is a power profiler that can follow the draw from microamps to an amp in one capture. It is next on the bench list. When it produces a real deep-sleep figure, a real transmit burst, and a real milliamp-hours-per-photo number, they go here.
What we can already say: the energy that matters is the radio's transmit burst, the burst lasts about as long as the delivery times in the table above, and a camera that sleeps between events is a different power problem than a relay that listens all day. That is why the mesh separates the two roles.
None of this is a coverage promise for your property. The measurements above found 17 dB between two bearings and 11 dB across 20 m of ground at a single site; no page can promise coverage over terrain it has never seen.
The design answer is to never depend on one heroic link: short, reliable hops, relayed, with the full-resolution original on every camera's own card so an outage delays a photo instead of losing it. And before anything gets mounted, the same handheld probe that produced this page walks your property and puts real measurements where the guesses would have been. It is included with every research pilot.
Signal figures are given in dBm as recorded by the receiving radio; margins are stated against the radio's rated sensitivity floor for the mode in use, in the downlink direction. Distances are GPS-measured and carried inside the probe payload, not inferred from signal strength. Timber-site measurements were taken 5 and 7 August 2026 in northeast Ohio, at peak leaf-out, with the gateway antenna on a house-mounted tower and the probe handheld at about 6 ft; the open-ground comparison site was surveyed in mid-July 2026. Antenna gain on a shipping gateway is capped below what our test tower carried, which costs roughly 3.5 dB on the outbound half; the margins above are stated before that adjustment. Questions about method, or want the raw probe logs? hello@getmeshcam.com.