Noise floor by RCGV1 · Pull Request #9347 · meshtastic/firmware

RCGV1 · 2026-01-18T03:27:43Z

I implemented a noise floor calculation by getting the radios current RSSI if its not transmitting nor receiving, this value is then added to a sliding window average which can hold up to 20 measurements. Noise Floor is sent via local stats.
This is extremely useful to determine if a cavity filter is needed for a particular location.

Depends on:
meshtastic/protobufs#846

iOS Interface:

🤝 Attestations

I have tested that my proposed changes behave as described.
I have tested that my proposed changes do not cause any obvious regressions on the following devices:
- Heltec (Lora32) V3
- LilyGo T-Deck
- LilyGo T-Beam
- RAK WisBlock 4631
- Seeed Studio T-1000E tracker card
- Other (please specify below)

Copilot

Pull request overview

This pull request implements noise floor calculation for Meshtastic firmware by sampling the radio's RSSI when idle, maintaining a sliding window average of up to 20 measurements, and exposing the value through local stats telemetry.

Changes:

Added noise floor tracking infrastructure to RadioLibInterface with rolling window averaging
Implemented getCurrentRSSI() for all radio chip variants (SX126x, SX128x, LR11x0, RF95, SimRadio)
Integrated noise floor measurement into telemetry module for reporting

Reviewed changes

Copilot reviewed 13 out of 13 changed files in this pull request and generated 12 comments.

Show a summary per file

File	Description
src/mesh/RadioLibInterface.h	Adds noise floor tracking member variables, methods declarations, and getCurrentRSSI() pure virtual method
src/mesh/RadioLibInterface.cpp	Implements updateNoiseFloor(), getAverageNoiseFloor(), and resetNoiseFloor() with rolling window logic
src/mesh/SX126xInterface.h/cpp	Implements getCurrentRSSI() for SX126x radios
src/mesh/SX128xInterface.h/cpp	Implements getCurrentRSSI() for SX128x radios
src/mesh/LR11x0Interface.h/cpp	Implements getCurrentRSSI() for LR11x0 radios
src/mesh/RF95Interface.h/cpp	Implements getCurrentRSSI() for RF95 radios
src/mesh/RadioInterface.h	Adds default getCurrentRSSI() implementation returning 0
src/platform/portduino/SimRadio.cpp	Implements getCurrentRSSI() for simulated radio returning -120 dBm
src/modules/Telemetry/DeviceTelemetry.cpp	Integrates noise floor value into local stats telemetry and logging

src/mesh/RadioLibInterface.cpp

src/modules/Telemetry/DeviceTelemetry.cpp

Copilot · 2026-01-19T18:32:54Z

src/mesh/RadioLibInterface.cpp

+        // Update noise floor after transmitting (radio is now in a good state to sample)
+        updateNoiseFloor();


Logic issue: After transmitting, sendingPacket is set to NULL at line 498, and then updateNoiseFloor() is called at line 514. However, updateNoiseFloor() checks if sendingPacket != NULL at line 265 and returns early. This means the noise floor will never be updated after transmitting, contradicting the comment that says "Update noise floor after transmitting (radio is now in a good state to sample)".

Well, if sendingPacket == NULL then it doesn't return early.

But indeed I don't think we should do it here, we should first return to receive mode as fast as possible after transmitting.

src/mesh/RadioLibInterface.h

src/mesh/RadioLibInterface.cpp

Copilot · 2026-01-19T18:32:56Z

src/mesh/RadioLibInterface.cpp

+
+    updateNoiseFloor();


Performance concern: updateNoiseFloor() is called at the beginning of every onNotify() callback (line 351), which is invoked for every ISR event (ISR_TX, ISR_RX, TRANSMIT_DELAY_COMPLETED). This could be inefficient. The function already has rate limiting (5 second interval), but it still performs multiple checks on every notification. Consider moving the call to only the ISR_RX case or after startReceive() calls where it's most relevant.

Copilot has a valid comment here. We shouldn't call this here, because every delay added here can cause the LoRa radio queue to get full (we have to read it out immediately because it can hold only 256 bytes). Doing it after entering receive mode is fine, and a period call e.g. at the end of Router::runOnce() or something.

Copilot · 2026-01-19T18:32:56Z

src/mesh/RadioLibInterface.cpp

+    // Get current RSSI from the radio
+    int16_t rssi = getCurrentRSSI();
+
+    if (rssi == 0 || rssi < NOISE_FLOOR_MIN) {


Validation logic issue: The condition checks rssi == 0 || rssi < NOISE_FLOOR_MIN to skip invalid readings. However, RSSI values are typically negative (e.g., -120 to -30 dBm), so checking for rssi == 0 might be too strict. A value of 0 dBm is technically valid though extremely unlikely. Consider whether this check is necessary, or if it should instead check for positive values which would truly indicate an error.

Suggested change

if (rssi == 0 || rssi < NOISE_FLOOR_MIN) {

if (rssi > 0 || rssi < NOISE_FLOOR_MIN) {

src/mesh/RadioLibInterface.h

…nter - Use robust busyTx/busyRx checks instead of simple isReceiving check - Initialize noiseFloorSamples to NOISE_FLOOR_MIN instead of 0 - Move noise_floor assignment inside null check to prevent potential crash - Change getNoiseFloor() and getAverageNoiseFloor() to return int32_t - Fix RSSI validation to check for positive values (rssi > 0) - Fix format specifier from %.1f to %d for int32_t - Update comments to accurately reflect the sampling logic

compumike · 2026-01-20T05:38:44Z

Very interesting!

We're not measuring Noise Floor (NF) in the sense of just random (stationary random process) background: thermal, amplifier, etc.

We're really interested in interference (potentially non-LoRa as well as LoRa, though excluding samples where a receive preamble is already detected by the radio).

It might be interesting to extract other statistics of these NF samples, not just the average.

For example: if we also extract the 10th-percentile and 90th-percentile values, this could tell us whether the interference is present all the time (10th and 90th close together), versus occasional (10th and 90th far apart). If they're far apart, it might actually inform us to delay a packet transmit if we measure a getRSSI(false) above a certain percentile, so we can attempt to avoid even non-LoRa interference! (Or maybe delay if it's more than X dB above the 10th percentile NF.) That would be pretty cool. 😄

compumike · 2026-01-20T05:46:49Z

Another thought: about how long does the getRSSI(false) call take to return?

Assuming it's quick, have you looked at taking a burst of consecutive samples (say, 10) and averaging them before storing in the array? (Maybe not just averaging the 10 samples as dBm values, but instead converting them to linear power mW values, averaging in mW, and then converting back to dBm.) This might give a cleaner short-term NF measurement which then goes into the longer-term array.

kamcio2603 · 2026-02-08T14:20:21Z

any update?

jock78 · 2026-02-08T16:21:55Z

Hi can provide Seeed Xiao Esp32, LILYGO T3v1.6.1, Heltec v4 and Heltec Pocket for testing.

src/mesh/LR11x0Interface.cpp

michalpan · 2026-02-24T22:57:36Z

Hi,

this is a super cool feature. It would help a lot with installations in locations with high noise. But it would be possible to know this ahead of time, instead of hitting it during debugging why a node is deaf…

it would be super cool to see in the app.

cheers,
Michal

Per PR review feedback, calling updateNoiseFloor() in onNotify() for every ISR event (ISR_TX, ISR_RX, TRANSMIT_DELAY_COMPLETED) can cause the LoRa radio queue to get full. The noise floor sampling still happens in startReceive() and after transmitting.

RCGV1 · 2026-02-25T21:54:32Z

@GUVWAF I had to enable "GodMode" to access that fucntion for lr11

GUVWAF · 2026-02-26T19:08:16Z

@RCGV1 Ah, I see. We shouldn't do that, I'll see if I can make that a public method in RadioLib.

GUVWAF · 2026-02-26T19:10:46Z

src/mesh/RadioLibInterface.h

+  protected:
+    // Noise floor tracking - rolling window of samples
+    static const uint8_t NOISE_FLOOR_SAMPLES = 20;
+    static const int32_t NOISE_FLOOR_MIN = -120; // Minimum noise floor clamp in dBm


Why clamping to -120dBm? If there is only thermal noise, at 0 degrees Celsius and with a small bandwidth (62.5kHz), you can go down to -126dBm.

When testing with the LR1110, I noticed that a value of -128 is returned when it is invalid (i.e. not in receive mode while calling the function). Hence I would say let's check for this specific value, and not clamp it to -120.

GUVWAF · 2026-03-01T09:51:25Z

@RCGV1 For LR11x0, when we update RadioLib to include jgromes/RadioLib#1714, we can use lora.getRSSI(false, true).

For the rest, please see my comment on clamping.

Furthermore, we should still remove updating the noise floor from here, as we need the radio to go the receive mode as fast as possible in order not to lose packets. Moreover, at least for the LR11x0, it is necessary to have it in receive mode before doing the RSSI measurement.

Lastly, in startReceive() it is also done only before enabling the Rx interrupt, which is also not ideal.
I would say maybe we could call updateNoiseFloor() from within DeviceTelemetry, just before we send our telemetry to the phone, which happens every minute. Then every 15 min. when the LocalStats are sent, we have a fairly good average.

robekl · 2026-03-02T03:20:57Z

src/modules/Telemetry/DeviceTelemetry.cpp


-    LOG_INFO("Sending local stats: uptime=%i, channel_utilization=%f, air_util_tx=%f, num_online_nodes=%i, num_total_nodes=%i",
+    LOG_INFO("Sending local stats: uptime=%i, channel_utilization=%f, air_util_tx=%f, num_online_nodes=%i, num_total_nodes=%i, "
+             "noise_floor=%f",


telemetry.variant.local_stats.noise_floor is int32_t, so %d should be used

robekl · 2026-03-02T03:21:42Z

src/mesh/RadioLibInterface.cpp

+    uint8_t sampleCount = getNoiseFloorSampleCount();
+
+    if (sampleCount == 0) {
+        return 0; // Return 0 if no samples


API/docs state uncalibrated should be -120 dBm (src/mesh/RadioLibInterface.h:129). This leaks impossible telemetry (0 dBm) via src/modules/Telemetry/DeviceTelemetry.cpp:122 and hurts data accuracy for fresh boot / low-traffic nodes.

robekl · 2026-03-02T03:22:41Z

src/mesh/RF95Interface.cpp

    RadioLibInterface::configHardwareForSend();
 }

 void RF95Interface::startReceive()


RF95Interface::startReceive() does not call RadioLibInterface::startReceive(), so the new noise-floor sampling hook in base (src/mesh/RadioLibInterface.cpp:616) is bypassed on RF95. Result: RF95 devices may never refresh noise floor except after TX-complete path, which biases or stalls measurements on RX-heavy nodes.

robekl · 2026-03-02T03:23:18Z

src/mesh/RadioLibInterface.cpp

+    LOG_DEBUG("Noise floor: %d dBm (samples: %d, latest: %d dBm)", currentNoiseFloor, getNoiseFloorSampleCount(), rssi);
+}
+
+int32_t RadioLibInterface::getAverageNoiseFloor()


Noise-floor buffer/index state is read and written without synchronization (getAverageNoiseFloor() vs updateNoiseFloor()), and telemetry calls into this from a different module thread (src/modules/Telemetry/DeviceTelemetry.cpp:122).

robekl · 2026-03-02T03:24:09Z

src/mesh/LR11x0Interface.cpp

+template <typename T> int16_t LR11x0Interface<T>::getCurrentRSSI()
+{
+    float rssi = -110;
+    lora.getRssiInst(&rssi);


return value should be checked instead of always assuming success

RCGV1 added 4 commits January 17, 2026 09:30

add noise floor

d4b8828

Sliding window noise floor

5fd527a

Add getCurrentRSSI() to SimRadio for noise floor support

e1b995f

Remove sendLocalStatsToPhone call from runOnce

c12de75

RCGV1 mentioned this pull request Jan 18, 2026

Noise floor meshtastic/Meshtastic-Apple#1566

Open

5 tasks

RCGV1 and others added 2 commits January 19, 2026 10:08

Change noise floor to int32_t type

02c980b

Merge branch 'develop' into noise-floor

1f7cd76

thebentern requested a review from Copilot January 19, 2026 18:20

RCGV1 added 2 commits January 19, 2026 10:22

Use int32_t for RSSI sample storage in noise floor

3ce3112

Remove float cast from noise floor assignment

3e52ca3

Copilot started reviewing on behalf of thebentern January 19, 2026 18:28 View session

Copilot AI reviewed Jan 19, 2026

View reviewed changes

RCGV1 added 2 commits January 19, 2026 21:52

Fix RSSI condition to include zero value

514b5de

Change noise floor initialization to zero

866abf0

jp-bennett added the enhancement New feature or request label Jan 31, 2026

Merge branch 'develop' into noise-floor

dbb0a96

Disable noise floor for LR11x0 chips: getRSSI(bool) unsupported

b830c8d

GUVWAF reviewed Feb 12, 2026

View reviewed changes

src/mesh/LR11x0Interface.cpp Outdated Show resolved Hide resolved

RCGV1 requested a review from GUVWAF February 25, 2026 01:29

GUVWAF requested changes Feb 25, 2026

View reviewed changes

fix lr11x0 current rssi

6dd2e2f

GUVWAF reviewed Feb 26, 2026

View reviewed changes

robekl reviewed Mar 2, 2026

View reviewed changes

		// Update noise floor after transmitting (radio is now in a good state to sample)
		updateNoiseFloor();

	if (rssi == 0 \|\| rssi < NOISE_FLOOR_MIN) {
	if (rssi > 0 \|\| rssi < NOISE_FLOOR_MIN) {

Uh oh!

Conversation

RCGV1 commented Jan 18, 2026

🤝 Attestations

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Copilot AI Jan 19, 2026

Choose a reason for hiding this comment

Uh oh!

GUVWAF Feb 25, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Copilot AI Jan 19, 2026

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 19, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

compumike commented Jan 20, 2026

Uh oh!

compumike commented Jan 20, 2026

Uh oh!

kamcio2603 commented Feb 8, 2026

Uh oh!

jock78 commented Feb 8, 2026

Uh oh!

Uh oh!

michalpan commented Feb 24, 2026

Uh oh!

RCGV1 commented Feb 25, 2026

Uh oh!

GUVWAF commented Feb 26, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

GUVWAF commented Mar 1, 2026

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

10 participants

GUVWAF Feb 25, 2026 •

edited

Loading