TX frequency plan and calibration — a simple, repeatable recipe

Nexus-BS 2.0 transmits every mode (TETRA, DMR, P25, FM) from one SDR, as carriers offset from one TX local oscillator (LO). A direct-conversion transmitter such as the SX1255 has two unavoidable artefacts:

• LO leakage: a spike exactly on the TX LO frequency. Uncorrected on an SX1255 board it is about as strong as a carrier (≈ 0 dBc).
• I/Q images: a weak mirror of every carrier on the other side of the LO, at 2·LO − f.

PlutoSDR, LimeSDR and USRP boards usually calibrate LO leakage and I/Q balance inside the device; look at the spectrum anyway and use the fields below only if a spike or an image is visible. SX1255 boards need them.

This page is the whole procedure: where to put the LO, how to cancel both artefacts from the dashboard, what to expect, and how to save the result. It takes about 30 minutes and needs no code and no scripts.

What you need

• A way to see the TX spectrum around 430–440 MHz with ≤ 3 kHz resolution: a spectrum analyzer, or an RTL-SDR/Airspy with a waterfall (SDR#, GQRX, SDR++). Absolute accuracy does not matter; you only compare levels.
• A dummy load and attenuation. Never connect the TX port straight to an analyzer or a dongle without enough attenuation for its input limit (an RTL-SDR needs 40–60 dB). Transmit into the dummy load, sample the signal through the attenuator.
• The dashboard, Settings → SDR. The fields TX DC I, TX DC Q, TX I/Q gain (dB) and TX I/Q phase (°) are applied live when saved: no restart, the station stays on the air.

Step 1 — Put the TX LO where it cannot hurt

Rules (all offsets are multiples of 25 kHz and at least 50 kHz apart):

1. Never on a channel. Offset 0 is refused on SX1255 boards.
2. Keep the LO at least 50 kHz from the nearest carrier, not 25 kHz: a residual LO spike 25 kHz from a carrier sits in its adjacent channel.
3. Put every carrier on the same side of the LO. Each image lands at 2·LO − f, so with all carriers above the LO every image falls below it, in empty spectrum, and never on another channel.
4. Stay inside the sample rate: every |offset| + 20 kHz must be below half the sample rate, so ±275 kHz at the usual 600 kS/s.
5. RX follows TX. Each channel has one offset_hz for both directions, so set rx_lo_hz = tx_lo_hz − duplex split and the RX frequencies stay right.

Recipe: TX LO = lowest carrier − 50 kHz, then offsets +50, +100, +150, ….

Example (70 cm, 7 MHz split), every RF frequency unchanged:

Change rx_lo_hz, tx_lo_hz and every offset_hz in one save (Config file tab), then restart the radio service. Saving the offsets one by one fails validation half-way, because two carriers briefly come closer than 50 kHz.

Step 2 — Leave the gains alone

• On SX1255 boards: MIXER and DAC gain at maximum (MIXER = 30, DAC = 9). The LO leakage does not shrink with the mixer gain, so lower gain only makes the carriers weaker relative to it.
• Other SDRs: use the TX gain at which the spectrum stays clean, then treat it like the SX1255 gains: fixed, and recalibrate if you change it.
• Keep the default tx_peak values (TETRA 0.8, DMR 0.5, P25 0.7, FM 0.7). Raising the digital level drove the SX1255 baseband into distortion: at 0.99 the images rose ~10 dB while the carriers rose only 2–6 dB.
• Adjust output power with an external PA or attenuator, not by pushing the digital level.

Step 3 — Cancel the LO leakage (TX DC I / Q)

Look at the TX LO frequency: span 50 kHz centred on it, RBW 1–3 kHz, averaging on (≈ 10–20 sweeps, or a slow waterfall). Note the LO spike level.

1. Set TX DC I to +0.01, save, wait for the average to settle. Then −0.01. Keep the sign that lowered the spike.
2. Walk TX DC I in that direction while the spike keeps dropping; when it rises again, go back one step and halve the step (0.01 → 0.005 → 0.002 → 0.001 → 0.0005 → 0.0002).
3. Do the same for TX DC Q.
4. Go back to TX DC I once, with the smallest step, then TX DC Q once.

Done when the spike is ≥ 40 dB below the carriers. Typical values are in the 0.01–0.03 range. 0.0001 already moves the spike by a few dB, so stop at 0.0002 steps: below that you are chasing noise, because the residual fluctuates by a few dB on its own.

Step 4 — Cancel the images (TX I/Q phase / gain)

Look at the image of your strongest carrier at 2·LO − f, with the same settings as step 3. In the example above the TETRA image is at 438.1625 MHz.

1. TX I/Q phase (°) first: try +0.5 and −0.5, keep the better sign, refine in 0.2° then 0.1° steps.
2. Then TX I/Q gain (dB): try +0.1 and −0.1, refine in 0.05 then 0.02 dB steps.
3. One more pass of phase with the smallest step.

Done when the image is ≥ 50 dB below its carrier, or you cannot see it above the noise. Typical values are a few tenths of a degree and a few hundredths of a dB. If the image is already invisible, leave both at 0.

Step 5 — Check, then save the profile

• Let the station warm up 30 minutes and repeat steps 3–4 once with the smallest steps.
• The values belong to this board at these gains. Redo steps 3–4 after changing the board, MIXER, DAC or the TX LO by more than a few hundred kHz. A 25 kHz LO move did not need a new calibration here.
• Save a profile so the next board swap costs one minute: /opt/nexus-bs2/calibration/<board-name>.toml with the four values, the gains, the LO and a line of results. See Docs/evidence/calibration/Z32IT_sx3.0_amp.toml for the format.
• Share it: the board name and the four values are useful to everyone with the same hardware.

What to expect (measured, SX1255 HAT V3 with TQP3M9036 driver)

Analyzer RBW 3 kHz, PosPeak detector, 100 averages, carriers ≈ 0 dBm:

About −41 dBc is the practical limit for the LO: the cancellation must match the leakage to about 1 % and the leakage drifts with temperature and supply. A filter cannot remove it at 25–50 kHz from a carrier, which is why step 1 matters.

Troubleshooting

• All carriers suddenly ~14 dB weaker, driver amplifier cold: reseat the HAT. A poor contact on the header caused exactly this and cleared on reseating.

• Station does not come back on the air after changing tx_peak: update to a build that includes commit e99c3f0. Older builds accepted values up to 0.99 in Settings but refused anything above 0.8 when starting the radio.

• LO spike reappears after a restart: normal drift; if it is more than ~5 dB worse, repeat step 3 with the smallest steps.

• Spike or image does not react at all to the fields: check that you are looking at the TX LO frequency (tx_lo_hz, not a carrier) and at 2·LO − f, not 2·f − LO.

• FM looks as narrow as DMR: a carrier with no audio is only the CTCSS tone, about 1 kHz wide. It widens with speech; if it stays narrow, check that the handheld is on a 25 kHz (wide) channel.

Community profiles

Each board is different: use a profile as a starting point, then run steps 3–4.