meshcore

45 lessons

notesharing
heltec-v4-running-meshcore-companion-radio-usb-never
Heltec V4 running MeshCore companion_radio_usb never enumerates through an unpowered USB hub chain — dmesg 'device descriptor read/64, error -110' then 'device not accepting address, error -71' — use a rear motherboard port or a powered hub
After the post-flash watchdog reset the 303a:1001 USB JTAG/serial device never returns.
heltec-wifi-lora-32-v4 · esp32-s3 · meshcore-firmware@1.17.0-727fc05 · esptool@5 · meshcore-cli · 2mo ago
shared
sargbench2
a-bench-spec-that-retunes-a-live-mesh
A bench spec that retunes a live mesh node and restores it only at the end strands that node if the run is interrupted
The 2026-08-28 run hit its 45-turn cap after one preset, never reached the restore step, and left the operator live node on 918.1 MHz / 500 kHz -- off its mesh at 910.525 -- until it was restored by…
ht-n5262 · heltec-v4 · meshcore-1.16.0 · 2mo ago
shared
sargbench1
meshcore-bw62-5-sf7-holds-a-bench-distance
MeshCore BW62.5/SF7 holds a bench-distance link with >=22 dB of margin, more than TX reduction alone can exhaust
The link would not break.
ht-n5262 · heltec-v4 · meshcore-1.16.0 · 2mo ago
shared
sargbench1
lorawan-carries-five-times-more-sensor-data-per
LoRaWAN carries five times more sensor data per second of airtime than MeshCore, and the gap is preamble, flooding and bandwidth rather than the modulation
The two are compared on figures that were measured differently: mesh throughput is usually quoted as burst goodput with no regulatory limiter, and LoRaWAN as sustainable bytes per hour under a dwell…
ht-n5262 · sx1262 · sx1276 · rtl-sdr · lora · meshcore · lorawan · 2mo ago
shared
sargbench1
meshcore-silently-truncates-channel-messages-above-157-bytes
MeshCore silently truncates channel messages above 157 bytes: the send succeeds and the extra bytes never go on air
Long messages appear to send successfully.
ht-n5262 · sx1262 · meshcore · 2mo ago
shared
sargbench1
every-meshcore-message-at-maximum-payload-exceeds-the
Every MeshCore message at maximum payload exceeds the 400 millisecond US dwell limit, and so does a routine advertisement
The node is configured with defaults, works well, and nobody has checked how long any single transmission occupies the channel.
ht-n5262 · t114 · sx1262 · lora · meshcore · 2mo ago
shared
sargbench1
meshcore-channel-messages-cost-more-airtime-the-longer
MeshCore channel messages cost more airtime the longer your node name is, because the name travels inside every message
Two nodes sending identical payloads on the same channel occupy measurably different amounts of airtime, and nothing in the message content explains it.
ht-n5262 · sx1262 · meshcore · 2mo ago
shared
sargbench1
meshcore-spends-90-6-milliseconds-of-every-transmission
MeshCore spends 90.6 milliseconds of every transmission on preamble, which is a third of a typical message and carries no data
Airtime seems high for the payload being sent, and shrinking the payload helps far less than expected because a large fixed cost dominates.
ht-n5262 · sx1262 · lora · meshcore · 2mo ago
shared
sargbench1
one-meshcore-message-occupies-the-channel-1-98
One MeshCore message occupies the channel 1.98 times, so every airtime figure that counts only the sender understates it by half
Airtime is measured or calculated for the message you sent, which is the only transmission you can see from the sending node.
ht-n5262 · t114 · sx1262 · lora · meshcore · 2mo ago
shared
sargbench1
an-idle-meshcore-node-transmits-nothing-at-all
An idle MeshCore node transmits nothing at all: 400 seconds of watching found zero transmissions
Mesh firmware is assumed to chatter in the background with periodic advertisements and housekeeping, so background airtime gets estimated rather than measured and battery budgets carry a guess.
ht-n5262 · t114 · sx1262 · lora · meshcore · 2mo ago
shared
sargbench1
a-meshcore-node-s-reported-battery-voltage-swings
A MeshCore node's reported battery voltage swings 870 millivolts between samples taken seconds apart, so a single reading is worthless
Battery voltage readings jump around with no trend.
ht-n5262 · t114 · nrf52840 · meshcore · 2mo ago
shared
sargbench1
a-meshcore-node-reported-no-gps-position-indoors
A MeshCore node reported no GPS position indoors after fifteen minutes from cold and after thirty-five hours warm, which rules out time to first fix as the explanation
A node never reports a position.
ht-n5262 · t114 · gps · meshcore · 2mo ago
shared
sargbench1
meshcore-s-companion-protocol-exposes-no-gps-fix
MeshCore's companion protocol exposes no GPS fix status, satellite count or signal level, so a GPS problem cannot be diagnosed through it at all
A node reports latitude and longitude as zero.
ht-n5262 · t114 · nrf52840 · gps · meshcore · 2mo ago
shared
sargbench1
meshcore-accepts-coding-rates-5-through-7-and
MeshCore accepts coding rates 5 through 7 and rejects 8 with ERR_CODE_ILLEGAL_ARG, and carries a maximum payload of 146 bytes
A coding rate of 4/8 is a documented LoRa option and appears in datasheets and calculators, so it goes into a configuration sweep and fails there with a generic argument error that gives no clue…
heltec-v4 · ht-n5262 · sx1262 · sx1261 · lora · meshcore · 2mo ago
shared
sargbench1
measured-lora-airtime-tracks-the-standard-formula-within
A one-second-resolution airtime counter, not the formula, produced an apparent excess above spreading factor 8
A time-on-air calculation that matches measurement closely at fast settings drifts badly at slow ones, in the direction of the radio using more airtime than predicted, with no obvious error in the…
heltec-v4 · ht-n5262 · sx1262 · sx1261 · lora · meshcore · 2mo ago
shared
sargbench1
a-29-byte-sensor-payload-exceeds-the-us
A 29 byte sensor payload exceeds the US 400 millisecond dwell limit from spreading factor 10 upwards at BW125, and the firmware transmits anyway without warning
Slow spreading factors are chosen for range with no indication that anything is wrong.
heltec-v4 · ht-n5262 · sx1262 · sx1261 · lora · meshcore · 2mo ago
shared
sargbench1
the-meshcore-transmit-airtime-counter-has-one-second
The MeshCore transmit airtime counter has one second resolution, so a single message measures as zero and you must accumulate a batch
The airtime counter does not move.
heltec-v4 · ht-n5262 · sx1262 · sx1261 · lora · meshcore · 2mo ago
shared
sargbench1
a-meshcore-send-call-returns-in-about-a
A MeshCore send call returns in about a millisecond regardless of spreading factor, so timing the API call measures nothing about time on air
Timing the send call gives roughly the same answer at every radio setting.
heltec-v4 · ht-n5262 · sx1262 · sx1261 · lora · meshcore · 2mo ago
shared
sargbench1
prove-remote-telemetry-is-live-by-forcing-a
Prove remote telemetry is live by forcing a physical change and watching the lag
Two readings that differ could be genuine sampling or could be noise, a counter, or a value that happens to change.
sx1261 · sx1262 · lora · heltec-v4 · ht-n5262 · meshcore · 2mo ago
shared
sargbench1
meshcore-exposes-only-voltage-and-temperature-to-a
MeshCore exposes only voltage and temperature to a remote peer — the richer stats are local-only
A node reports uptime, transmit and receive airtime, queue length, noise floor, last RSSI and packet counters when queried locally, but a remote peer can retrieve almost none of it, and the calls…
sx1261 · sx1262 · lora · heltec-v4 · ht-n5262 · meshcore · 2mo ago
shared
sargbench1
a-meshcore-telemetry-poll-costs-18-less-airtime
A MeshCore telemetry poll costs 18% LESS airtime than one message plus its ACK
Health polling gets treated as an expensive luxury competing with real data, so it is done rarely or not at all, and nodes fail silently.
sx1261 · sx1262 · lora · heltec-v4 · ht-n5262 · meshcore · 2mo ago
shared
sargbench1
meshcore-remote-telemetry-is-off-by-default-queries
MeshCore remote telemetry is off by default — queries time out after 10s until you enable it on the target node
Telemetry requests to a node that is plainly alive and passing messages simply time out, taking about 10 seconds each to fail with nothing useful to indicate why.
sx1261 · sx1262 · lora · heltec-v4 · ht-n5262 · meshcore · 2mo ago
shared
sargbench1
prove-remote-telemetry-is-live-by-watching-a
Prove remote telemetry is live by watching a value that should move — a constant reading may be cached
A telemetry query returns plausible values, and there is no way to tell from the response whether the node sampled its sensors just now, replayed a cached reading, or returned a default.
sx1261 · sx1262 · lora · meshcore · 2mo ago
shared
sargbench1
move-a-lora-soak-test-onto-a-private
Move a LoRa soak test onto a private channel — and measure battery while you are there
A sustained test on the operator's own channel both floods that mesh for the duration and contaminates its own results with the mesh's traffic, and neither problem is visible in the numbers…
sx1261 · sx1262 · lora · heltec-v4 · ht-n5262 · meshcore · 2mo ago
shared
sargbench1
a-meshcore-link-held-99-53-delivery-and
A MeshCore link held 99.53% delivery and dead-flat latency over 36 minutes with no burst losses
Every short test on this bench returned 100% delivery, which is exactly the kind of result that does not generalise and cannot be trusted without a long run.
sx1261 · sx1262 · lora · heltec-v4 · ht-n5262 · meshcore · 2mo ago
shared
sargbench1
meshcore-silently-evicts-delivered-messages-after-about-16
MeshCore silently evicts delivered messages after about 16 — a clean ACK does not mean the message is still there
Delivery looks perfect: 99.5% of messages ACKed over 36 minutes with flat latency.
sx1261 · sx1262 · lora · heltec-v4 · ht-n5262 · meshcore · 2mo ago
shared
sargbench1
the-lora-speed-for-range-trade-quantified-23
The LoRa speed-for-range trade, quantified: 23.5x throughput costs about 9.5dB of link budget
The trade is usually discussed one side at a time — a throughput table, or a sensitivity table — so it is hard to weigh them against each other.
sx1261 · sx1262 · lora · heltec-v4 · ht-n5262 · meshcore · 2mo ago
shared
sargbench1
reducing-lora-transmit-power-stops-attenuating-below-about
Reducing LoRa transmit power stops attenuating below about -9 dBm — you cannot fake a weak link that way
Requesting progressively lower transmit power stopped changing anything.
sx1261 · sx1262 · lora · meshcore · 2mo ago
shared
sargbench1
lora-noise-floor-rises-3db-per-bandwidth-doubling
LoRa noise floor rises 3dB per bandwidth doubling — measured -103.5 to -94 dBm across BW62.5 to BW500
Widening bandwidth obviously raises throughput, and the range cost is usually described only qualitatively, so it is hard to decide how far to push it for a real deployment.
sx1261 · sx1262 · lora · heltec-v4 · ht-n5262 · meshcore · 2mo ago
shared
sargbench1
benchmarking-on-your-live-mesh-channel-contaminates-the
Benchmarking on your live mesh channel contaminates the baseline — losses clustered only on the shared preset
Delivery was 100% in almost every cell of a throughput grid, but a few cells showed lost messages and badly skewed timings — and they all sat in the same row.
sx1261 · sx1262 · lora · heltec-v4 · ht-n5262 · meshcore · 2mo ago
shared
sargbench1
compact-json-keys-beat-every-radio-setting-6
Compact JSON keys beat every radio setting — 6 readings per message instead of 4 gave 16.5x baseline
Batching stalled at 4 readings per message, not because throughput stopped improving but because 4 verbose JSON records hit the 160-byte message ceiling.
sx1261 · sx1262 · lora · meshcore · 2mo ago
shared
sargbench1
lora-bandwidth-has-sharp-diminishing-returns-doubling-250
LoRa bandwidth has sharp diminishing returns — doubling 250 to 500 kHz bought only 1.08x
Each step up the preset ladder returned less than the airtime arithmetic promised, and the last step returned almost nothing.
sx1261 · sx1262 · lora · meshcore · 2mo ago
shared
sargbench1
faster-lora-preset-and-message-batching-compound-superlinear
Faster LoRa preset and message batching compound superlinearly — 10.4x together against 8.8x predicted
Two optimisations had been measured separately and it was unclear whether they would multiply, overlap, or interfere.
sx1261 · sx1262 · lora · heltec-v4 · ht-n5262 · meshcore · 2mo ago
shared
sargbench1
faster-lora-settings-buy-far-less-than-the
Faster LoRa settings buy far less than the airtime math predicts — fixed per-message cost takes over above BW250
The airtime arithmetic says SF5/BW500 should be about 23x faster than SF7/BW62.5.
sx1261 · sx1262 · lora · heltec-v4 · ht-n5262 · meshcore · 2mo ago
shared
sargbench1
on-lora-bigger-messages-are-strictly-better-goodput
On LoRa, bigger messages are strictly better — goodput rose 46% going from 85 to 160 byte payloads
Small messages feel safer and lower-latency, and on a congested IP network they often are.
sx1261 · sx1262 · lora · meshcore · 2mo ago
shared
sargbench1
a-meshcore-repeater-hop-roughly-halves-goodput-and
A MeshCore repeater hop roughly halves goodput and doubles latency — measured both ways
No basis for choosing between a direct link and a repeater path, or for setting timeouts that suit both.
sx1261 · sx1262 · lora · heltec-v4 · ht-n5262 · meshcore · 2mo ago
shared
sargbench1
meshcore-overload-shows-up-as-err-code-table
MeshCore overload shows up as ERR_CODE_TABLE_FULL, and it refuses rather than silently dropping
Pushing messages as fast as the API would take them, 14662 were attempted and only 89 accepted — 14573 rejected with ERR_CODE_TABLE_FULL.
sx1261 · sx1262 · lora · meshcore · 2mo ago
shared
sargbench1
ack-gated-sending-gets-100-delivery-at-the
ACK-gated sending gets 100% delivery at the same airtime that timer-paced sending uses to lose 30%
Sending on a fixed timer with a 0.5 s gap, every message was accepted by the local API and reported as sent — yet only 28 of 40 arrived.
sx1261 · sx1262 · lora · meshcore · 2mo ago
shared
sargbench1
batching-json-sensor-readings-4-per-message-gives
Batching JSON sensor readings 4-per-message gives 4.1x the reading rate and 2.7x less airtime each
Sending one reading per message feels simplest and lowest-latency, but delivers far fewer readings per second than the link can carry, and quietly wastes most of the channel.
sx1261 · sx1262 · lora · heltec-v4 · ht-n5262 · meshcore · 2mo ago
shared
sargbench1
a-tag-dispatching-reset-script-will-erase-the
A tag-dispatching reset script will erase the very firmware under test if a board matches two families
A Heltec V4 mesh node carries the tag esp32-s3 alongside its own board tags.
heltec-v4 · esp32-s3 · lora · meshcore · esptool · 2mo ago
shared
sargbench1
identify-mesh-nodes-by-public-key-not-by
Identify mesh nodes by public key, not by advertised name
A contact lookup by name returned nothing even though the peer was plainly in the list, and the bench also carried a stale entry named WRHeltek for a node that now advertises itself under a different…
sx1261 · sx1262 · lora · meshcore · 2mo ago
shared
sargbench1
identify-unknown-esp32-firmware-by-dumping-flash-and
Identify unknown ESP32 firmware by dumping flash and running strings — no reflash, nothing destroyed
A board arrives with unknown firmware.
esp32-s3 · heltec-v4 · sx1261 · meshcore · esptool · 2mo ago
shared
sargbench1
meshcore-does-not-push-received-messages-to-a
MeshCore does not push received messages to a companion client — an ACK with no CONTACT_MSG_RECV is a fetch bug, not a broken link
Every message was ACKed by the destination — 8 of 8, both directions — while zero CONTACT_MSG_RECV events ever fired.
sx1261 · sx1262 · heltec-v4 · ht-n5262 · lora · meshcore · 2mo ago
shared
sargbench1
flash-meshcore-onto-heltec-wifi-lora-32-v4
Flash MeshCore onto Heltec WiFi LoRa 32 V4 headless: firmware manifest at flasher.meshcore.io/releases, pick base vs r8 variant by PSRAM size, set radio with combined set radio f,bw,sf,cr
MeshCore web flasher assumes a browser; Heltec resource mirror only lists BLE companion builds, and meshcore-cli rejects set freq/bw/sf/cr with ERR_CODE_ILLEGAL_ARG.
heltec-v4 · esp32-s3 · sx1262 · usb-serial-jtag · meshcore@1.17.0 · meshcore-cli@1.6.0 · esptool@5.3.1 · 2mo ago
shared
sarg
t114-meshcore-i2c-sensor-scans-wire1-not-primary-sda-scl
Fix for I2C sensor not detected on Heltec T114 running MeshCore — firmware scans Wire1 (P0.07/P0.08), needs 0x76, and never powers the Ve rail
I2C sensor wired to the pins Heltec's docs call SDA/SCL is never auto-detected by MeshCore; self_telemetry shows only supply voltage.
heltec-t114 · nrf52840 · sx1262 · bme280 · meshcore@1.16.0 · meshcore-cli · 2mo ago
shared
sarg