PDU header format 0x15 is the generic unconfirmed data header; unconfirmed
multi-block trunking control (UMBTC) is only its trunking control SAP
(61/63) use. PDUHeaderFactory built a plain PDUHeader for every 0x15
header and PDUMessageFactory sent every such sequence to createUMBTC(), so
outbound unconfirmed IP data (SNDCP RF UNCONFIRMED DATA with SAP 0, 4 or
6) was logged raw as
PDU HEADER FORMAT:UNCONFIRMED MBTC UNCONFIRMED VENDOR:STANDARD TO LLID:n
DATA BLOCKS:4 MSG:4100450000...
while the confirmed equivalent (format 0x16) was decoded as IPPKT with
its LRRP / ARS payload.
- PDUHeaderFactory: build a PacketHeader for a 0x15 header whose SAP is
UNENCRYPTED_USER_DATA, PACKET_DATA or SNDCP_PACKET_DATA_CONTROL. Other
SAPs, including trunking control 61/63, keep the plain PDUHeader.
- PDUMessageFactory: route a 0x15 sequence whose header is a PacketHeader
through createPacketData(); everything else stays on the UMBTC path.
No payload changes are needed: PacketMessage already sizes unconfirmed
data blocks at 12 octets, and the header fields it uses (SAP, pad octet
count, data header offset) sit at the same bit positions in the confirmed
and unconfirmed headers. The ARS RegistrationAcknowledgement parser only
reads the extension octet, so the 5-octet trailer (0x04 + 32-bit time)
some systems append does not affect it.
Adds PDUMessageFactoryTest, built from two unconfirmed PDUs captured on a
Motorola P25 Phase 1 system (ARS registration acknowledgements to UDP
port 4005, one with the trailer), plus header routing checks for SAP
0/4/6, SAP 61/63 and the confirmed 0x16 header.
Co-authored-by: csmith_ai <carrottop@gmail.com>
Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
Motorola LRRP request tokens that carry an integer (trigger interval
0x31, trigger distance 0x78 and the unknown parameters 0x42 / 0x44)
encode it as a 7-bit variable length integer: each byte holds seven
value bits, most significant group first, and the high bit of a byte
flags a continuation byte. The parser read a single byte, so a real
triggered location start request
09 11 22 03 FF FF EE 52 44 64 42 9C 10 62 57 34 31 84 58
was logged as
ID:15663103 TRIGGER ON GPIO TRIGGER EVERY:132 SECS REQUESTED TOKENS
[TIME,REQUEST-64,REQUEST-62,HORIZONTAL DIRECTION,TIMESTAMP,
UNRECOGNIZED(10,44,58,9C)]
and now decodes to
ID:FFFFEE PERIODIC INTERVAL:600 SECS PARAM-42:3600 PARAM-44:100
REQUESTED TOKENS [TIME,HORIZONTAL SPEED,HORIZONTAL DIRECTION]
The encoding and the token names follow node-dmr-lib (LRRP.js
parseLRRPInt/toLRRPInt and its quoted TRBONet captures, DMRConst.js)
and Moto.Net (LRRPPacket.cs "using VLQ for most of the fields").
- Add VarInt, a shared 7-bit varint reader, and VarIntToken, a token
base class whose byte length is measured from the value and bounded
by the remaining payload. Use them for 0x31, 0x78, 0x42 and 0x44.
- Add TokenType.fromRequestValue(): 0x34 is the periodic trigger flag
in a request and the timestamp in a response. TokenFactory accepts
the resolved TokenType so the request path is not re-resolved.
- Rename request tokens: 0x31 TRIGGER_INTERVAL (TriggerInterval),
0x4A TRIGGER_CONDITION (TriggerCondition: 02 emergency, 42 pin
status change, 50 PTT, 51 PTT emergency), 0x78 TRIGGER_DISTANCE
(TriggerDistance), 0x62 HORIZONTAL_SPEED. 0x42 and 0x44 become
REQUEST_PARAMETER_42/44 (RequestParameter) with an optional varint
value; their meaning is unknown (observed 3600 and 100). Remove
TriggerGpio and TriggerOnMove.
- Identity: assemble the request id big-endian (it was accumulated
LSB-first), render it in hex, and clamp the read to the message.
- RequestedTokens: keep a list that holds only unrecognized ids; add
getters for the recognized and unrecognized lists.
- Remove leftover debug locals from LRRPPacket.parseTokens().
- Add LRRPPacketTest (both captured bodies plus two copies with a
corrupted type byte) and VarIntTest.
Built with the project's Gradle build (JDK 27 toolchain, in Docker);
compileJava, compileTestJava and the 9 new tests pass.
Co-authored-by: csmith_ai <carrottop@gmail.com>
Co-authored-by: Claude Fable 5.1 <noreply@anthropic.com>
Golay24.checkAndCorrect indexed the word from bit 0 (getInt(0, 22), xor(12, ...),
flip(index)) and took the parity over the whole message, so it was only right for a
message that is exactly one 24-bit word. P25 TDULC words 2-12 and the DMR voice
superframe words at 24 and 48 were "corrected" in the wrong bits. Decode any other
word in a 24-bit copy and write it back.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
A lagging sync detect on the first symbol after a buffer reload, followed by a
coarse timing search that steps all the way left, scores a candidate whose first
sync symbol is one sample before the buffer start. score() only checked the end
of the buffer, so it threw ArrayIndexOutOfBoundsException: Index -1.
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
* #2535 Heterodyne Channelizer enhancements and corrections.
* #2537 Complex vector oscillator SIMD lane drift.
Realigns the SIMD lanes every 262,144 samples. At frequencies where the
rotation for a set of lanes is nearly a whole or half turn, rounding affects
each lane differently and the lanes drift apart until the next frequency
change, producing spurs at multiples of the sample rate divided by the lane
count. Lane phase error at 3125 Hz with a 50 kHz sample rate was about -38 dBc
after one hour and is now about -83 dBc with 16 lanes and -73 dBc with 8 lanes.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
* #2537 test updates
* #2537 test updates
---------
Co-authored-by: denny <dennis.sheirer@gmail.com>
Co-authored-by: Claude Opus 5.5 <noreply@anthropic.com>
- Fix two-channel synthesis asymmetry across polyphase channel boundaries.
- TwoChannelSynthesizerM2: replace incorrect 2-path polyphase split (which applied
different responses to each channel) with channel sum + prototype filter
modulated to the channel boundary (+FS/4).
- ComplexPolyphaseChannelizerM2: start on the middle block map when taps per
channel is odd so adjacent channel outputs are phase coherent.
- TwoChannelOutputProcessor: translate channel 2 by FS/2 using the absolute
block index parity derived from the even-sized dispatch batches.
- SynthesizerViewer: translate channel 2 before synthesis.
- Add TwoChannelSynthesizerM2Test for flat, symmetric boundary response.
Co-authored-by: denny <dennis.sheirer@gmail.com>
* Adopts same architecture and symbol processor as P1 C4FM/LSM, DMR and NXDN.
* Soft symbol sync detection, processing and erro correction.
* Scalar and vector soft sync detection and calibration
* Improved time slot detection/prediction and ISCH processing.
* Fixes bug in parsing sequential mac structures in SACCH
* Vector calibration(s) cleanup across multiple calibrations
* Error correction bug in ISCH
* Renames differential demod shared with P1 LSM decoder.
Co-authored-by: denny <dennis.sheirer@gmail.com>
* Uses same shared architecture and symbol processor pattern as other C4FM decoders.
* Adds soft symbol processing and sync detection.
* Removes RRC filter and replaces with improved channel filter.
* Viterbi soft symbol decoding with demodulator reliability weighting
* Improved sync reacquisition in low SNR
* Improved handling for phase wrapping and interpolation
* NAC Tracker updated to track by NAC count
* Native buffers returning partial buffers
* Message processor uses only valid LDU1 words
* Reduced phantom TDU messages
* Tuner/Channe frequency error corrections don't reset the symbol trackers and fix possible divide by zero error
* LSM timing error detector now updates per symbol block size instead of sample buffer size - faster convergence
* New unit tests
* Resolves several issues, including a lagging sync detect offset issue in the recently updated P25P1 C4FM decoder.
* Consolidate filter designers for each protocol back into the FilterFactory class.
* DMR, P25 Phase 1 C4FM and LSM, and P25 Phase 2 decoders design the baseband filter once and share the taps across the I and Q filters.
Co-authored-by: denny <dennis.sheirer@gmail.com>
* Add failing test for toCSV()
* Add missing " to timeslot value
* Add failing test for quotes and commas in details
* Separate comma and quote tests
* Escape toCSV cells using utility
* Use CSVFormatter instead of escape utils
Allows us to control exactly how we want to format a CSV-valid row.
* Never have a null cell
Instead of using a stream map, opting for an explicit handling so the logger can explode when an unexpected null shows up. Prevents the situation where things are mysteriously blank, forcing developers to handle known null scenarios.
