Satellite phones without GPS: per-beam reference points agreed, two-root PRACH narrowed

A phone talking to a satellite has to time and tune its first transmission for a long signal path and a large Doppler shift. It knows where the satellite is, because the network broadcasts the orbit data (ephemeris), but without GNSS it does not know where it is itself, so it cannot work out the distance or the Doppler. The Release 20 work gives the network two tools: a reference location the phone can use in place of its own position, and a random-access signal built so the base station can measure the error that remains.

For that signal, 3GPP narrowed the leading design — two PRACH sequences in the existing format — to Alt 1-4 as the baseline: the two sequences use different, non-conjugate roots, which lets the base station separate the timing error from the frequency error. Other variants, such as two sequences in different PRACH formats or several sequences tied to multiple reference locations, remain under study. Thales, the feature lead running the discussion, put it forward with Ericsson, Nokia, Huawei, Qualcomm and about ten others behind it; OPPO objected. 3GPP decides by consensus, so one objection can stall a proposal, but here it did not. Whether the two sequences share one random-access occasion or use separate ones will be decided at the next meeting.

In quasi-Earth-fixed cells, where the satellite steers each beam so that its footprint stays in place for a while before switching, the network will broadcast a reference location: one per cell when the cell has a single beam, one per beam when it has several. About two dozen companies argued that beam-level granularity is needed to keep residual errors within what the phone can correct. ZTE wanted a per-cell baseline, and Nokia, while backing the approach, warned that a reference point per beam adds broadcast overhead; whether a cell-level location can still be configured in multi-beam cells is left for further study. Earth-moving cells get the equivalent: a common timing-advance and Doppler value, per beam where a cell has several.

The corrections the network sends back in the random-access response are getting bigger too. The timing-advance command will be extended to cover at least the largest residual errors in the study report TR 38.742, with today's granularity kept; a 3 ms range was proposed but the value is still open. A frequency-adjustment command will be defined with its own range, granularity and positive and negative values, with the message format left to the protocol group; the granularity, either a fraction of the subcarrier spacing or a ppm step, is due to be chosen at the next meeting. How to signal the common timing advance in Earth-moving cells is not settled: reusing the 26-bit format of the existing ta-Common parameter, which reaches 270.73 ms, was only proposed as a working assumption.

On the device side, the revised work item approved at the plenary gives high-power satellite phones (Power Class 1.5) both two-antenna transmit diversity and single-layer uplink MIMO — both send one data stream from two antennas — and rules out dual-layer MIMO. MediaTek, LG, Huawei, OPPO and Samsung had leaned toward transmit diversity only; SpaceX, Apple, Amazon, Qualcomm, ViaSat and AST kept single-layer MIMO in. The same item adds half-duplex requirements for VSAT terminals in the Ka and Ku bands, which are dish- or panel-sized satellite terminals far larger than a phone, and targets the new downlink carrier aggregation in the L band at all existing power classes, PC3, PC2, PC1.5 and PC1.