Six modes on one frequency pair, and the four side by side tables. Working notes, v1.1, 10 September 2026.
This is the part with no mode of its own. It covers the amateur multi-mode repeater, which runs all six on one frequency pair, and it holds the four side by side comparison tables that only mean anything with all six together. The six mode pages carry their own bandwidth, routing, roaming and trunking sections; this page is where they line up against each other.
The four comparison tables on this page:
This is the most amateur-specific thing in the whole subject. One frequency pair, one MMDVM board, and six digital modes on it. No commercial repeater does this, because no commercial user would want it. It exists because amateurs have one coordinated pair and six modes they would like to run on it.
The rule that governs everything else: one mode at a time. Always. A multi-mode repeater is not running six modes at once. It is running one mode, and switching between them when it gets the chance.
It takes the mode from the last thing it heard, on RF or from the network, whichever came last. It then locks onto that mode for a set time before it will consider anything else. That lock is the mode hang timer.
Read from the G4KLX MMDVM ini rather than from a guide:
| Setting | Default | What it does |
|---|---|---|
| RFModeHang | 10 seconds | How long the repeater stays locked to a mode after RF traffic in it |
| NetModeHang | 3 seconds | The same, after network traffic. Shorter, so RF users get the longer hold |
| ModeHang | 10 seconds | A global override, and each mode can carry its own |
| Timeout | 180 seconds | Longest single transmission before the repeater drops you |
| TXHang, DMR | 4 seconds | How long the transmitter stays up after a call, per mode |
| TXHang, System Fusion | 4 seconds | |
| TXHang, P25 | 5 seconds | |
| TXHang, NXDN | 5 seconds | |
| CallHang, DMR only | 3 seconds | Reserves the slot for replies on the current talkgroup before anyone else may key |
One constraint follows from that and it is easy to get wrong: a mode's TX hang must be the same as or less than the mode hang. Otherwise the transmitter is still up for a mode the repeater has already stopped being in.
Key up in D-STAR while the repeater is locked in DMR and your transmission is simply ignored. The repeater may look busy and you will hear nothing, or depending on setup you get a "repeater not available" message. Nothing is broken. You arrived during somebody else's mode hang. Wait out the timer and try again.
Because DMR has two timeslots, a DMR-locked repeater can carry two simultaneous users. D-STAR, C4FM, NXDN, P25 and M17 are single channel, so one user each. That is worth knowing before anyone concludes the repeater is broken because two people got in on DMR and only one can on YSF.
MMDVMHost has a CW identifier, on by default, keying the repeater for a few seconds every 10 minutes. It is there because DMR carries no callsign in the data stream that an ordinary listener can decode, and Part 97 still requires the station to identify. So a digital repeater periodically sends morse in the clear. It is not a legacy artifact, it is the identification requirement being met the only way that works across every listener.
Six modes on one pair means six modes competing for one pair. Add mode hang, TX hang and a 10 minute CW ID, and a busy multi-mode machine spends a noticeable fraction of its time either locked to somebody else's mode or identifying. It is a real tradeoff, and it is the price of not needing six coordinated frequency pairs.
RF plumbing is deliberately out of scope in these notes, because a duplexer does not know or care what modulation is passing through it. Bandwidth is the exception. It differs by mode and it is worth knowing.
| Mode | Channel bandwidth | Modulation | Bit rate |
|---|---|---|---|
| D-STAR (DV) | 6 kHz occupied, 6.25 kHz spacing | GMSK | 4800 bps total: 2400 voice, 1200 FEC, 1200 data |
| NXDN | 6.25 kHz | 4FSK | 4800 bps |
| NXDN | 12.5 kHz | 4FSK | 9600 bps |
| M17 | 9 kHz occupied, 12.5 kHz spacing | 4FSK | 9600 bps, 4800 symbols per second |
| DMR | 12.5 kHz | 4FSK, two slot TDMA | 9600 bps, 4800 symbols per second, two slots sharing it |
| C4FM | 12.5 kHz, every mode | 4FSK | 9600 bps, 4800 symbols per second |
| P25 Phase 1 | 12.5 kHz | C4FM, a form of 4 level FM. CQPSK also permitted | 9600 bps, 4800 symbols per second |
| P25 Phase 2 | 12.5 kHz | H-CPM inbound, H-DQPSK outbound, two slot TDMA | 12 kbit/s, 6000 symbols per second, two voice paths |
There are two completely different ways to be spectrum efficient, and the modes divide over them. Not every mode is in a camp. C4FM and the P25 Phase 1 that amateurs actually run occupy a full channel and carry one conversation in it, and M17 sits between the two, described below.
Narrow the RF. D-STAR and NXDN genuinely occupy 6.25 kHz. Two NXDN channels fit inside one 12.5 kHz allocation.
Divide the time. DMR and P25 Phase 2 occupy the full 12.5 kHz but carry two conversations in it. They are called 6.25 kHz equivalent, which is true per user and not true of the signal on the air. Wikipedia is blunt about P25 Phase 2: voice "transmits with the full 12.5 kHz per frequency allocation, however it does so at a faster data rate of 12 kbit/s allowing two simultaneous voice transmissions."
That distinction explains something the notes keep running into. DMR has two timeslots because it chose time division. D-STAR has none because it chose narrow RF instead. It is also why only DMR can carry two users on a multi-mode repeater while the other five carry one.
M17 sits between the two camps and publishes its occupied bandwidth separately from its channel spacing: 9 kHz of signal in a 12.5 kHz slot, with a deviation index of 0.33 giving symbol deviations of plus and minus 0.8 and 2.4 kHz.
And the practical consequence for a multi-mode repeater: none of the six exceeds 12.5 kHz, which is exactly why one coordinated pair can carry all of them.
Two things up front, because DMR is most people's baseline and it has a feature none of the others share.
Only DMR has timeslots. DMR is two slot TDMA, so every talkgroup has to be assigned to TS1 or TS2. D-STAR, C4FM, NXDN and M17 are all single channel, and P25 Phase 1 is too. For the other five, "which timeslot" is not a question that exists.
Every mode has a routing field, and all of them except D-STAR also have an access field. The two get confused constantly. The access field works like CTCSS and decides whether the repeater listens to you at all. The routing field decides where your audio goes. DMR's color code is access, its talkgroup is routing. Keeping those apart is most of the answer.
| Mode | Access field | Routing field | Range |
|---|---|---|---|
| DMR | Color code | Talkgroup, plus TS1 or TS2 | 24 bit talkgroup, 0 to 16777215 |
| D-STAR | none | Module letter, and URCALL | A to Z |
| C4FM | RX DG-ID | TX DG-ID | 0 to 99 |
| NXDN | RAN | Destination ID | RAN 6 bit, dest ID 16 bit |
| P25 | NAC | Talkgroup | NAC 12 bit, TG 16 bit |
| M17 | CAN | Destination callsign | module letter in the last position |
Each mode page works through its own two fields in detail. The short version of each:
The pattern. Four of the six put routing in a number the user dials: DMR talkgroup, C4FM DG-ID, NXDN destination ID, P25 talkgroup. Two do not: D-STAR and M17 both use a letter attached to a reflector name. And every mode except D-STAR has an access field sitting right beside the routing field that people mistake for it, which is exactly the trap NXDN's RAN sets.
The useful way to hold this is that roaming needs two halves, and each mode has a different combination.
The network half: several sites linked together so that being on any of them puts you in the same conversation.
The radio half: something the radio can measure, and logic in the radio to move by itself.
| Mode | Network half | Radio half | Roaming in amateur use |
|---|---|---|---|
| NXDN | Yes, IP linked sites | Yes, beacon plus vote scan | POSSIBLE needs commercial gear at both ends |
| DMR | Yes, but not required | Yes, entirely radio side | YES amateurs use it today |
| C4FM | Yes, MSRL and IMRS | No | NO |
| D-STAR | No | List only, operator picks | NO |
| P25 | No | No | NO |
| M17 | No | No | NO |
Roaming needs the radio to measure something. There are only two ways to guarantee there is something to measure.
Passive. The repeater transmits a beacon on a schedule
and the radio listens and compares.
Active. The radio transmits a probe on the repeater input and listens for the
repeater to answer.
Every serious implementation offers both. Kenwood pairs its beacon with Active Site Hunt, where "the transceiver sends a Wakeup message to the repeater to prompt synchronization", governed by a Sync Wakeup Wait Time and a retry count. Motorola pairs Passive Site Search with Active Site Search. Anytone offers only the active probe, because a ham repeater historically does not beacon and its design cannot assume one.
An earlier draft of these notes said DMR and NXDN were exact inverses, one receive side and one transmit side. That was wrong on both counts and is corrected here. Kenwood's radio does transmit, and MMDVM's DMR repeater can beacon. The difference between the two modes is narrower and more specific than that: MMDVM implements the beacon for DMR and not for NXDN. It is an implementation gap, not a design philosophy.
The detail behind each verdict is on the mode's own page: the NXDN beacon and what it would take to use it, the several distinct DMR roaming schemes including Kenwood's conventional beacon and MMDVM's own, C4FM having the network half and nothing else, and D-STAR's GPS-ranked list with the operator choosing.
Three of the six have trunking.
D-STAR, C4FM and M17 have no trunking at all. Not unused. Not present.
The frequency count is the first wall. A conventional repeater is one pair, two frequencies. A trunked system needs a pool of pairs, one of which is given over entirely to the control channel, which carries no voice. A modest five channel trunked site is five pairs, ten frequencies, four of them usable for talking. Five times the spectrum for four times the capacity.
The control channel transmits continuously, for as long as the system exists, carrying no conversation. That alone would be a hard conversation with a frequency coordinator.
But the real reason is simpler: trunking solves a problem amateur radio does not have. Trunking exists so many users can share few channels with automatic assignment, because in public safety the channels are congested and the traffic is constant. Amateur repeaters are mostly idle. Automatic channel assignment on an empty system buys nothing.
On top of that, MMDVM implements trunking for no mode, so the cheap route is closed, and every user would need a trunking capable subscriber radio programmed for that specific system.
It is not the FCC. The claim that Part 97 prohibits trunking gets repeated in forums without a citation and no such provision was found. In the United States the barrier is coordination councils rejecting the applications outright. That is a different thing and these notes should not blur them.
Amateur trunked systems exist, and every one this search found is DMR. Two leads were never read, listed as an open item below, so that is what was found and not a survey. They are described on the DMR page: two Motorola Capacity Plus systems in Canada, VA3UEC in southern Ontario and a five site system in central Alberta, plus dmrtc, an open source DMR Tier III trunked controller written for amateur and educational use by Adrian Musceac, YO8RZZ.
No amateur P25, NXDN, LTR, SmartNet or EDACS trunked system was found anywhere. Conventional P25 on ham bands is common. Trunked P25 is not.
A deep search closed three of the five items that stood in earlier drafts. What follows is what survived it, plus what the search itself turned up. Each item is repeated on the page it belongs to.
Transmit Interval Time in KPG-D7, but its range is dealer-gated and not in any
public document. Icom publishes nothing at all. Two forum figures do not cleanly agree.
Getting this needs a dealer with the programming software open, or an off air capture.
(NXDN)Closed by the deep search, and worth recording as closed: whether the beacon is part of NXDN (no, it is vendor design, five origins), the Yaesu repeater program (alive, the contrary article is an April Fools piece), and whether anyone has run an amateur trunked system (yes, at least two in Canada).
Honest counts, because these notes will feed a public page. Each mode page repeats the entries that belong to it.
MMDVM-Host.cpp and Conf.cpp.