FCC Part 15.247: The 500 kHz Minimum and LoRa Mesh

500 kHz is the number, and almost nothing in LoRa mesh is running it.

47 CFR 15.247(a)(2) sets a minimum 6 dB bandwidth of 500 kHz for digitally modulated systems in the 902 to 928 MHz band. Meshtastic’s default Long Fast preset is 250 kHz. MeshCore’s USA preset is 62.5 kHz. Neither one hops. That gap is what the mesh community has spent the better part of a year arguing about, and in the last few weeks it’s started changing what people actually run.

We’re not lawyers and this isn’t legal advice. This post does the one thing we can actually do. Read the rule text, the FCC’s own measurement guidance, and the public engineering record. Then lay out what they say, and what’s still contested. It doesn’t tell you what to run. It doesn’t tell you whether your setup is lawful. It makes no recommendation. Those are calls for a qualified professional, and in the end for the Commission.

What the Rule Actually Says

Section 15.247 opens by naming what it covers: “Operation under the provisions of this Section is limited to frequency hopping and digitally modulated intentional radiators that comply with the following provisions.” Two categories, each with its own paragraph. Hopping gets 15.247(a)(1). Digital modulation gets 15.247(a)(2), and the operative part is one sentence: “Systems using digital modulation techniques may operate in the 902-928 MHz, 2400-2483.5 MHz, and 5725-5850 MHz bands. The minimum 6 dB bandwidth shall be at least 500 kHz.”

The rest of the section fills in the other limits for a digitally modulated system here:

Read that antenna line slowly. The power limit is conducted, measured at the antenna port. The rule only claws power back once a directional antenna passes 6 dBi. It’s not a flat EIRP cap.

What “6 dB Bandwidth” Means

It’s the frequency span between the points where the signal has dropped 6 decibels below its peak. Specific measurement, specific procedure, defined for these devices in ANSI C63.10 and in the FCC’s measurement guidance, KDB 558074, which restates the limit and gives it a name: “The minimum 6 dB bandwidth of a DTS transmission shall be at least 500 kHz. Within this document, for DTS devices this bandwidth is referred to as the DTS bandwidth.”

For LoRa the answer isn’t surprising. A chirp spread spectrum signal sits close to its configured bandwidth. A 125 kHz LoRa transmission measures near 125 kHz at the 6 dB points. A 500 kHz one measures near 500 kHz. Spreading factor changes how long the symbol takes, not how wide it is. That’s why no amount of SF11 or SF12 turns a 250 kHz signal into a 500 kHz one.

One easy mix-up to avoid. The hopping rules in 15.247(a)(1) are written against the 20 dB bandwidth, not the 6 dB one. Different threshold, different number, different paragraph.

Four Doors Into the Band

Digital modulation is one way to be authorized here, not the only one. A compliance test lab’s guide to LoRa certification describes three paths under Part 15, and a fourth sits outside 15.247 entirely.

Hopping is how a 125 kHz LoRa signal becomes ordinary. LoRaWAN in the US hops 125 kHz channels across the band, meeting 15.247(a)(1) instead of (a)(2). So The Things Network running 125 kHz uplinks isn’t evidence that narrow fixed-channel operation is fine. Different paragraph, different rules.

The hybrid path is blunt about it. KDB 558074, on hybrid systems: “There is no requirement for this type of hybrid system to comply with the 500 kHz minimum bandwidth normally associated with a DTS device.” The same guidance adds that a hybrid has no minimum channel count. The hop sequence still has to be pseudorandom under 15.247(a)(1).

Door four is 15.249, which doesn’t regulate conducted power at all. It caps field strength at 50 mV/m measured at 3 meters. Convert that to radiated power, using EIRP equal to field strength times distance, squared, divided by 30, and you get about 750 microwatts. Under a milliwatt. Whatever that path is, it isn’t where a mesh node with a 22 dBm radio lives.

The Second Limit Nobody Mentions

The 500 kHz minimum gets all the attention. The power spectral density limit in 15.247(e) is the one quietly doing the arithmetic. The power “conducted from the intentional radiator to the antenna shall not be greater than 8 dBm in any 3 kHz band during any time interval of continuous transmission.”

Power spread over a narrower signal is denser. Fix the ceiling at 8 dBm per 3 kHz slice, and the total a signal can carry drops as it narrows.

Look where that lands. At 500 kHz the density limit allows just over a watt, and the separate conducted power limit is exactly one watt. The two limits meet at the bandwidth the rule sets as its minimum. The rule set is internally consistent, which is a decent hint the 500 kHz figure was picked on purpose.

That table is arithmetic, not a test report. Certification measures PSD on a spectrum analyzer under ANSI C63.10 procedures with specified resolution bandwidth and detectors, and a real LoRa chirp doesn’t lay energy down perfectly evenly. Treat the numbers as the shape of the constraint, not a compliance result.

It does explain a pattern in public grant data, though. High-power 900 MHz modules often carry authorized power well under what the hardware can do. The Skyworks-boosted RAK13302 is a nominal 30 dBm part, and its grants under FCC ID 2AF6B-13302 authorize 0.2897 W as a spread spectrum transmitter and 0.2831 W as a digital transmission system, both peak conducted. Under 290 mW either way.

Where the Presets Land

Meshtastic’s published modem presets:

Two of nine are 500 kHz. The default isn’t one of them.

MeshCore ships its suggested settings through a config service its clients read, currently 27 regional presets. The US-relevant ones:

One preset out of 27, across every region on that list, uses 500 kHz. It’s the Philadelphia entry. The list is labeled “These presets are suggested by the community.”

The Part That Isn’t Settled

The number in the rule isn’t ambiguous. Its scope is what people disagree about, and the disagreement is technical, not rhetorical. Both readings are on the public record, so here they are.

The first reading treats 15.247(a)(2) as a condition of operation. The section says operation under its provisions is limited to hopping and digitally modulated radiators complying with the following provisions, and one of those is the 500 kHz minimum. On that reading, a fixed-channel system narrower than 500 kHz isn’t operating under 15.247 at all. What’s left in this band is 15.249, with its sub-milliwatt field strength limit. Semtech’s own compliance guidance for the band, and the test lab guide above, both describe the 500 kHz floor for direct sequence and equivalent digitally modulated operation in those terms.

The second reading splits certification from operation. On this view the 500 kHz requirement governs the mode used to demonstrate compliance during equipment authorization. Certified modules ship with plenty of narrower modes nobody tested, the argument goes. Once a grant exists, those narrow modes become a power spectral density question rather than a bandwidth prohibition.

The document that speaks most directly to that second reading is the FCC’s own measurement guidance, which takes on multi-mode devices head-on. It notes that many devices under test “provide for operation in multiple transmission modes, where the data rate, bandwidth, modulation, coding rate, and number of data streams are often variable,” then states the requirement: “When such multiple modes of operation are possible, then compliance to the applicable technical requirements shall be confirmed for any and all realizable operational modes.”

KDB 558074 does allow a smaller test matrix, but conditionally. You can identify the modes producing “worst-case” results against every technical limit and test only those. It’s specific about the price: “Whenever this type of test reduction is utilized, a complete and detailed technical justification shall be provided in the test report, to include measurement data where applicable.”

A third argument shows up now and then, that the whole section only governs frequency hopping. The text quoted at the top of this post is what people cite against it. The scope sentence names digitally modulated radiators right alongside hopping ones. And (a)(2) is addressed to systems using digital modulation techniques.

Two structural points come up in these threads, and both are accurate descriptions of how Part 15 is built. Equipment authorization and operating authority are separate things. A grant attaches to equipment. The duty to operate within the rules sits with the operator. And the marketing rules in 47 CFR 2.803 cover importing and selling. That’s a different exposure from operating, and it lands on vendors rather than individuals.

We’re not going to tell you which reading wins. That’s a question for a lawyer or a qualified compliance engineer, and finally for the Commission.

What May Be Underway

MeshCore. The live thread is issue #945 in the MeshCore repo, “Suggestion: new USA Regulatory preset,” opened in October 2025 and still open with 377 comments, last active on September 11, 2026. The proposal swaps the 62.5 kHz / SF7 US preset for 500 kHz / SF11. The proposer figures that as a 1 dB gain in link budget for a 25% cut in data rate. It also moves the center frequency onto an evenly divided channel center. The US default in the config service hasn’t changed. What has changed is that a 500 kHz entry now exists there. That’s the Philadelphia preset, 902.250 MHz at SF11.

The technical objection on record deserves stating plainly, because it isn’t hand-waving. A Seattle-area operator reports utility meter channels every 300 kHz across the band there. At one repeater site the noise floor runs roughly 30 dB higher at 250 kHz than at 62.5 kHz. He posted a waterfall capture to back it up. A wider channel scoops up more of whatever else is in the band. Where the band is busy, the compliant bandwidth and the usable bandwidth pull against each other.

Meshtastic. The default is still Long Fast at 250 kHz per current published docs. Long Turbo and Short Turbo, both 500 kHz, already exist as selectable presets, so the firmware side of a change is a defaults-and-coordination problem rather than an engineering one. We searched the Meshtastic organization’s repos for compliance warnings referencing 15.247 or FCC bandwidth and found none in the firmware or the official apps. Screenshots of an in-app bandwidth warning have been circulating. The ones we traced came from socialmesh, a fork, not an official Meshtastic client.

Regional meshes. PhillyMesh published a migration in early September 2026. Infrastructure nodes moved to the 500 kHz settings now carried in MeshCore’s preset list. They wrote openly about the fragmentation that follows when half a mesh changes modem settings. Presets aren’t interoperable. A node at 500 kHz and a node at 62.5 kHz don’t hear each other, so this isn’t a change any one operator makes alone.

What Nobody Can Tell You Yet

A few things stayed genuinely open after this research, and writing around them would be dishonest.

We found no FCC enforcement action against a hobbyist mesh operator over bandwidth in 902 to 928 MHz. Absence in a public database isn’t proof none exists, and it says nothing about what the rules require.

We found no pending petition or rulemaking that would change the 500 kHz minimum. Petitions to lower it have been filed in the more distant past, which is how we know the figure has been contested before, but nothing current turned up.

Whether running a certified module in an untested narrow mode is an equipment matter, an operational matter, or neither is exactly the contested question above. The measurement guidance quoted here speaks to what a test report has to cover, not to what an end user may do afterward.

And spectrum occupancy is local. Seattle’s band isn’t Boston’s and neither is ours. Any bandwidth discussion that ignores what’s already transmitting where the node will live is incomplete on its own terms.

Reading the Primary Sources

If this matters to your deployment, the documents are short and public, and reading them beats reading anybody’s summary, this one included. The rule is 47 CFR 15.247. The measurement guidance is FCC KDB publication 558074, which test labs mirror widely. The MeshCore discussion is issue #945. Meshtastic’s preset table is in their radio settings documentation.

Want background on why bandwidth, spreading factor, and range trade against each other? Our explainer on how LoRa mesh networks work covers the modulation side. The protocols comparison covers how the two projects differ in what they put on the air. If you want to see what your own band looks like before forming an opinion about channel width, that starts with antenna and RF measurement.

The number in the rule hasn’t moved since long before either protocol existed. What’s moving is how many operators have read it. The network you end up building sits downstream of choices being made in public right now. In issue threads, one regional mesh at a time.