Connect responses select the setup speed

Stock VARA HF 4.9.0 caller replays on 2026-09-20 correct the earlier reading of lower lattice positions as damaged-request rejections. They are connect offers selecting the speed of the caller-ID setup transmission.

Positions use the bandwidth's existing connect-response descriptor, advancing 30 PRNG draws per position. Each payload remains keyed to the called station. This table concerns connection setup, not DATA ACK/NAK interpretation.

Independent evidence

KD7UHR's responses after requests 4, 5 and 6 in 20260920T050126Z-W9SSJ-KD7UHR.wav each match BW2750 position 0, 15/15 tones. Each recording was replayed at original gain to a separately launched stock caller. Every trial reported UNENCRYPTED LINK, BITRATE (1) 18 bps TX, then transmitted a CRC-valid caller-ID frame for W9SSJ:

5e44d32800000080048bd63e

Fresh stock-caller trials also cover the remaining lower positions at all three bandwidths and the BW2750 level-4 control. Decoded frames, source hashes, speed announcements and capture status are retained in tests/kestrel/fixtures/kd7uhr-response-zero-0920/stock-setup-frames.json. Level-1 trials with KD9ZZZ and W9SSJ-10 independently check the identity format. Full local audio and host/PTT ledgers remain in the private station archive.

An earlier multi-case experiment reused one caller after ABORT. Some later CONNECT commands returned WRONG, so those trials are excluded. Every reported caller result above uses a fresh Wine instance, virtual cables, recorded input and output, and verified owned-process cleanup.

The reverse interoperability test also reached stock's CONNECTED W9SSJ KD7UHR status at every bandwidth/level combination in the table. A fresh stock responder heard a clean, corrupted-tone or noisy request; its actual offer was measured before our transmitter rendered setup at that speed. All 12 cases had clean capture status and verified cleanup. Exact run IDs and input/output/source hashes are in stock-setup-accepted.json beside the frame fixture. Labels such as accept-500-2 are experiment names: offered_level and the measured tones, not the name, determine which speed was tested.

These are setup-interoperability tests on virtual cables; they do not establish live KD7UHR mail delivery or later DATA speed negotiation.

The September 9 damage experiments established that request damage changes the response position. They did not establish rejection semantics: our own caller ignored those positions, and no stock caller was fed them. That circular inference caused Kestrel to discard valid lower-speed offers.

Setup formats

The first eight body bytes are the existing packed caller/SSID/setup header. Levels 2..4 append 14 <caller CRC high>, zero padding, and 04 82. Level 1 instead ends the ten-byte body directly with 04 8b. All frames append the big-endian CRC-16/GENIBUS of their body.

The setup uses the corresponding DATA waveform. The chosen host level is kept separately from the operator's DATA-level setting. Repeated offers may select a different setup speed, within the existing transmitted-setup budget. Receiving an offer alone never marks the session CONNECTED; confirmation is still owed.

All four positions retain the called-station, bandwidth, tone-count and comparability checks. The stream and segmented-audio paths accept short-preamble responses using their payloads. No new unkeyed or cross-bandwidth acceptance is introduced.