Lora longer preamble by OverkillFPV · Pull Request #1954 · meshcore-dev/MeshCore

OverkillFPV · 2026-03-07T02:12:32Z

After testing i have found that for the narrow settings i am using (62.5khz, sf7) that there is a reliability issue of transmissions that did not exist on wide (250khz,sf11). I have come to the conclusion that the preamble is too short for the low spreading factor as it is 100ms shorter going from 131ms for wide, down to 32ms for narrow. the increase to a preamble length of 32 brings the preamble time up to 65ms (32ms extra). In my opinion this extra on air time is offset by the lower need to resend an entire packet. I have been running this longer preamble for a while with good results, and everyone who has tested the modification so far has either noticed a slight, or decent improvement in reliability.

This change changes the preamble length to 32 <sf9 and the standard 16 >sf8. It allows the sf to be changed through the cli and will adjust the preamble on the next tx

Copilot

Pull request overview

This PR adjusts LoRa preamble length dynamically based on the configured spreading factor (SF), aiming to improve reliability at low SF / narrow bandwidth settings by increasing preamble symbols.

Changes:

Add a getSpreadingFactor() hook to RadioLibWrapper and implement it across radio-specific wrappers.
Set preamble length to 32 symbols for SF ≤ 8, otherwise 16 symbols; update preamble automatically on the next TX when SF changes.
Extend CustomLR1110/wrapper to expose and use the current spreading factor for preamble decisions.

Reviewed changes

Copilot reviewed 9 out of 9 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
src/helpers/radiolib/RadioLibWrappers.h	Adds `_preamble_sf` state and a virtual `getSpreadingFactor()` API.
src/helpers/radiolib/RadioLibWrappers.cpp	Applies preamble length policy at init and before TX when SF changes.
src/helpers/radiolib/CustomSX1276Wrapper.h	Implements `getSpreadingFactor()` via radio’s `spreadingFactor`.
src/helpers/radiolib/CustomSX1268Wrapper.h	Implements `getSpreadingFactor()` via radio’s `spreadingFactor`.
src/helpers/radiolib/CustomSX1262Wrapper.h	Implements `getSpreadingFactor()` via radio’s `spreadingFactor`.
src/helpers/radiolib/CustomSTM32WLxWrapper.h	Implements `getSpreadingFactor()` via radio’s `spreadingFactor`.
src/helpers/radiolib/CustomLLCC68Wrapper.h	Implements `getSpreadingFactor()` via radio’s `spreadingFactor`.
src/helpers/radiolib/CustomLR1110Wrapper.h	Updates preamble reset logic and implements `getSpreadingFactor()` via the radio.
src/helpers/radiolib/CustomLR1110.h	Exposes `getSpreadingFactor()` accessor used by the wrapper.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-03-07T02:17:01Z

src/helpers/radiolib/RadioLibWrappers.cpp

 void RadioLibWrapper::begin() {
  _radio->setPacketReceivedAction(setFlag);  // this is also SentComplete interrupt
+  _preamble_sf = getSpreadingFactor();
+  _radio->setPreambleLength(_preamble_sf <= 8 ? 32 : 16); // longer preamble for lower SF improves reliability


RadioLibWrapper::begin() now unconditionally overrides the preamble length to 16/32 based on SF. This can silently override board/variant-specific preamble lengths configured during radio initialization (e.g., variants/ebyte_eora_s3/target.cpp passes a preamble length of 8 to radio.begin). If that shorter preamble is intentional for some targets, consider preserving the configured default (or making the 16/32 policy conditional/configurable) rather than forcing it here.

Suggested change

_radio->setPreambleLength(_preamble_sf <= 8 ? 32 : 16); // longer preamble for lower SF improves reliability

// Only override preamble length if none has been explicitly configured.

// This avoids clobbering board/variant-specific settings provided via radio.begin().

int desiredPreamble = (_preamble_sf <= 8) ? 32 : 16; // longer preamble for lower SF improves reliability

int currentPreamble = _radio->getPreambleLength();

if (currentPreamble <= 0 || currentPreamble == desiredPreamble) {

_radio->setPreambleLength(desiredPreamble);

}

Copilot · 2026-03-07T02:17:02Z

src/helpers/radiolib/RadioLibWrappers.cpp

+    _preamble_sf = sf;
+    _radio->setPreambleLength(sf <= 8 ? 32 : 16); // update preamble when SF has changed


startSendRaw() forces preamble length back to 16 whenever SF > 8. If some targets intentionally use a different default preamble (e.g., 8 symbols configured at radio.begin), switching from SF<=8 to SF>=9 will overwrite that default and keep it overwritten for subsequent packets. Consider tracking/restoring the configured default (or only changing preamble when entering/leaving the low-SF case) to avoid target-specific regressions.

Suggested change

_preamble_sf = sf;

_radio->setPreambleLength(sf <= 8 ? 32 : 16); // update preamble when SF has changed

bool wasLowSf = (_preamble_sf <= 8);

bool isLowSf = (sf <= 8);

// Only adjust preamble length when crossing between low-SF (<= 8) and high-SF (> 8) regimes.

if (wasLowSf != isLowSf) {

_radio->setPreambleLength(isLowSf ? 32 : 16);

}

_preamble_sf = sf;

JDat · 2026-03-07T09:38:54Z

Is this compatible with existing MeshCore versions? By changing preamble will this code work with existing repeaters on network? Just asking, not dive deep into SX12xx datasheets yet.

OverkillFPV · 2026-03-07T12:08:34Z

Yes, this affects the tx of the node it is on, and is completely interchangeable between nodes. Currently all my repeaters and companions are running 32 and are working perfectly with the rest of the mesh

weebl2000

Minor: the preamble expression is repeated 4 times — a small helper would reduce duplication.

weebl2000 · 2026-03-07T12:17:40Z

src/helpers/radiolib/RadioLibWrappers.h

@@ -38,6 +39,7 @@ class RadioLibWrapper : public mesh::Radio {
  }

  virtual float getCurrentRSSI() =0;


Consider extracting the repeated ternary to a helper to keep the 4 call sites DRY:

Suggested change

virtual float getCurrentRSSI() =0;

virtual uint8_t getSpreadingFactor() const { return LORA_SF; }

static uint16_t preambleLengthForSF(uint8_t sf) { return sf <= 8 ? 32 : 16; }

weebl2000 · 2026-03-07T12:33:29Z

@OverkillFPV what do you think about 24 instead of 32.

jbrazio · 2026-03-07T13:03:29Z

LGTM but why sf9 as the pivot point ?

OverkillFPV · 2026-03-07T14:56:06Z

I tried 24, 32 and 64 and 32 had an improvement over 24 with only a slight improvement going up to 64. Sf9 was just a middle ground i picked. As it would require many many iterations to determine where the swing should be.

OverkillFPV · 2026-03-07T14:56:47Z

Minor: the preamble expression is repeated 4 times — a small helper would reduce duplication.

Possibly I'll have a look at simplifying it

OverkillFPV · 2026-03-11T14:06:11Z

I'm currently playing around with a companion of 64 and repeaters at 32. As I am seeing if it helps with the random dropped messages

weebl2000 · 2026-03-11T14:16:35Z

Interested to hear the results!

OverkillFPV · 2026-03-11T20:02:03Z

Early testing is looking to suggest the tag being a bit more reliable for tx and less rx biased. Only my first impressions though

Edit: hasn't actually had much of an affect when testing from known hard to reach spots. My early thoughts was most likely due to good conditions

OverkillFPV added 3 commits March 5, 2026 21:39

Increased preamble to 32 for <sf9 and left at 16 for sf9 and above

534a0c9

Update preamble dynamically to match runtime SF changes via CLI

a61add2

Cache preamble SF to avoid redundant setPreambleLength() on every TX

6fd8c28

Copilot AI review requested due to automatic review settings March 7, 2026 02:12

Copilot started reviewing on behalf of OverkillFPV March 7, 2026 02:12 View session

Copilot AI reviewed Mar 7, 2026

View reviewed changes

weebl2000 reviewed Mar 7, 2026

View reviewed changes

simplified the calls

a434f9e

-  _radio->setPreambleLength(_preamble_sf <= 8 ? 32 : 16); // longer preamble for lower SF improves reliability
+  // Only override preamble length if none has been explicitly configured.
+  // This avoids clobbering board/variant-specific settings provided via radio.begin().
+  int desiredPreamble = (_preamble_sf <= 8) ? 32 : 16; // longer preamble for lower SF improves reliability
+  int currentPreamble = _radio->getPreambleLength();
+  if (currentPreamble <= 0 || currentPreamble == desiredPreamble) {
+    _radio->setPreambleLength(desiredPreamble);
+  }

		_preamble_sf = sf;
		_radio->setPreambleLength(sf <= 8 ? 32 : 16); // update preamble when SF has changed

-    _preamble_sf = sf;
-    _radio->setPreambleLength(sf <= 8 ? 32 : 16); // update preamble when SF has changed
+    bool wasLowSf = (_preamble_sf <= 8);
+    bool isLowSf = (sf <= 8);
+    // Only adjust preamble length when crossing between low-SF (<= 8) and high-SF (> 8) regimes.
+    if (wasLowSf != isLowSf) {
+      _radio->setPreambleLength(isLowSf ? 32 : 16);
+    }
+    _preamble_sf = sf;

		@@ -38,6 +39,7 @@ class RadioLibWrapper : public mesh::Radio {
		}

		virtual float getCurrentRSSI() =0;

	virtual float getCurrentRSSI() =0;
	virtual uint8_t getSpreadingFactor() const { return LORA_SF; }
	static uint16_t preambleLengthForSF(uint8_t sf) { return sf <= 8 ? 32 : 16; }

Conversation

OverkillFPV commented Mar 7, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Mar 7, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Mar 7, 2026

Choose a reason for hiding this comment

Uh oh!

JDat commented Mar 7, 2026

Uh oh!

OverkillFPV commented Mar 7, 2026

Uh oh!

weebl2000 left a comment

Choose a reason for hiding this comment

Uh oh!

weebl2000 Mar 7, 2026

Choose a reason for hiding this comment

Uh oh!

weebl2000 commented Mar 7, 2026

Uh oh!

jbrazio commented Mar 7, 2026

Uh oh!

OverkillFPV commented Mar 7, 2026

Uh oh!

OverkillFPV commented Mar 7, 2026

Uh oh!

OverkillFPV commented Mar 11, 2026

Uh oh!

weebl2000 commented Mar 11, 2026

Uh oh!

OverkillFPV commented Mar 11, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

5 participants

OverkillFPV commented Mar 11, 2026 •

edited

Loading