Technical Deep-Dive · FT1
The FT1 protocol
A next-generation weak-signal digital mode for amateur radio. FT1 brings techniques proven in cellular networks to the HF bands — pulling more information out of every second of air time and clawing signal back over the noise floor, all in a fast four-second cycle. It's a general-purpose mode: making a keyboard QSO move at the speed of a conversation is just one of the things it unlocks.
Cellular tech, on the ham bands
Borrowed from the phone in your pocket
The two techniques that define FT1 — incremental-redundancy Hybrid-ARQ (IR-HARQ) and turbo equalization — are workhorses of modern 4G/LTE cellular systems. They've moved billions of bits through deeply faded, interference-ridden channels. FT1 is, in essence, an experiment in pointing those same tools at HF.
IR-HARQ
When a frame fails, FT1 doesn't just repeat it — each retransmission sends new parity bits the receiver has never seen, soft-combined into a longer, lower-rate code. Energy accumulation and extra coding gain. No existing amateur text mode does this; in cellular, it's standard.
Turbo equalization
The receiver iterates: a maximum-likelihood trellis detector (BCJR) and the LDPC decoder trade soft information back and forth, each pass refining the other. It's the same iterative-decoding idea that made turbo codes the heart of 3G/4G — here it claws an estimated 1.5–2 dB out of the noise.
The trade is deliberate: FT1 gives up roughly 3 dB of raw single-shot sensitivity versus FT8 to buy its fast, conversational cycle — then leans on these cellular-grade techniques to win much of it back over several short transmissions.
The short version
Three things make FT1 different
-
It's coherent.
FT1 uses 4-ary continuous-phase modulation (4-CPM) and tracks carrier phase to demodulate it, where FT8/FT4/JS8 detect tone energy non-coherently. Coherent detection extracts more information per second of air time — that's the lever that makes a fast cycle competitive.
-
It has IR-HARQ.
A retransmission carries new parity bits, soft-combined with the original frame to decode a longer, lower-rate code. No FT8/FT4/JS8 mode does this — they're one-shot block codes per frame. This is the headline differentiator.
-
It's conversational.
A 4-second transmit/receive cycle — versus 15 s for FT8 and 6–30 s for JS8. Short cycles make a back-and-forth feel like a conversation instead of a slideshow.
The tradeoff it's fighting
Weak-signal text faces a tradeoff that is fundamental physics, not an engineering shortfall: cycle time versus weak-signal reach. You can integrate the signal over a long window to dig it out of the noise, or keep the cycle short so the conversation flows — but every second you remove from the integration window costs sensitivity. There is no single waveform optimal at both ends of that curve.
That's why the Tempo app ships two tiers rather than one compromise waveform: FT1 (fast, coherent) for stable paths, and DX1 (slow, non-coherent) for fading. Both carry the identical 77-bit message and the same LDPC(174,91) FEC — only the modem and the clock change, and the operator picks, every transmission, from a visible toggle.
The waveform
FT1 at a glance
| Modulation | Coherent 4-CPM (h = 1/2, BT = 0.3) |
|---|---|
| Channel symbols | 99 — 12 Costas sync + 87 data |
| Symbol rate | 28 Bd |
| Waveform duration | ≈ 3.536 s inside a 4.0 s UTC slot |
| Sample rate | 12 kHz (standard WSJT-X audio rate) |
| Error correction | LDPC(174,91) + turbo equalization, IR-HARQ option |
| Simulated AWGN threshold | ≈ −15 dB (simulation only) |
The signal chain
From text to air, and back
shared message + FEC layer
text ──▸ 77-bit WSJT-X payload ──▸ +14-bit CRC = 91 bits ──▸ LDPC(174,91) ──▸ 174 coded bits
FT1 physical layer
174 bits ──▸ 87 data + 12 Costas sync = 99 symbols @ 28 Bd
──▸ 4-CPM (h=1/2, BT=0.3) ──▸ 3.536 s waveform in a 4.0 s UTC slot ──▸ AIR
receive
AIR ──▸ Costas sync search (time + freq) ──▸ downconvert to ~222 Hz baseband
──▸ turbo decode: iterative BCJR ⇄ LDPC belief-propagation ──▸ OSD fallback ──▸ SIC
──▸ [IR-HARQ soft-combine across retransmissions — designed, not yet live]
──▸ 91 bits ──▸ CRC check ──▸ textThe differentiator
How redundancy accumulates
FT1's IR-HARQ design is rate-compatible — a mother code, LDPC(348,91), extended from the baseline. Each retransmission reveals more of it. The redundancy version rides on Costas-pattern variants, so it costs zero signaling overhead, and a legacy station simply decodes the standard RV0 frame.
| Transmission | Bits sent | Effective code | Cumulative |
|---|---|---|---|
| RV0 — 1st TX | 174 | LDPC(174,91) | baseline |
| RV1 — 2nd TX | 87 new parity | LDPC(261,91) | original + new parity |
| RV2 — 3rd TX | 87 new parity | LDPC(348,91) | mother code |
Simulated design analysis: 3-TX IR-HARQ reaches roughly −21 dB in 12 seconds total — on par with FT8 — while a full QSO at −20 dB completes in ~43 s, still faster than FT8's ~60 s at the same SNR. (Design/simulation figures, not on-air results.)
Two tiers, one conversation
FT1 vs DX1
| FT1 — Fast | DX1 — Robust | |
|---|---|---|
| Modulation | Coherent 4-CPM | Non-coherent 8-FSK |
| T/R cycle | 4 s (~3.5 s waveform) | 15 s |
| Occupied bandwidth | narrow (~tens of Hz) | ~50 Hz |
| Sim. AWGN threshold | ≈ −15 dB | ≈ −18.6 dB |
| Best for | Conversation on stable paths | Fading / disturbed paths |
DX1's non-coherent 8-FSK never relies on carrier phase, so it loses only ~3.7 dB under Rayleigh fading — where coherent FT8-class modes lose 10+ dB. That small fading penalty is the entire reason the robust tier exists.
Go deeper
This is the overview. The full protocol write-up covers the message layer, Costas synchronization, the LLR-combining math, and the complete FT8/FT4/JS8 comparison.