Merge dr-qam16-cport into main by rafael2k · Pull Request #1 · Rhizomatica/codec2

rafael2k · 2026-02-18T02:31:48Z

This PR proposes merging dr-qam16-cport into main.

Note: the branch currently has merge conflicts with main that need resolution before merge.

…kets mpp at 15dB-ish SNR

Resolve src/ofdm_mode.c merge conflict while preserving both qam16 and qam16c2 mode configuration.

Copilot

Pull request overview

This PR merges the dr-qam16-cport branch into main, extending the existing OFDM/LDPC stack to support a new high-SNR 16-QAM OFDM data mode (qam16c2) and generalizing several QPSK-specific paths to handle bps=4.

Changes:

Add qam16c2 mode plumbing across C modem, FreeDV API/raw tools, and Octave reference scripts; increase UW capacity to support larger UW bitfields.
Generalize packet assemble/disassemble, demod, and LDPC soft-decision (LLR) generation from QPSK-only to PSK/QAM (2 or 4 bits/symbol).
Add/adjust CMake tests and documentation for the new mode.

Reviewed changes

Copilot reviewed 25 out of 25 changed files in this pull request and generated 4 comments.

Show a summary per file

File	Description
unittest/tqam16.c	Updates qam16 demod call signature in unit test.
unittest/tofdm.c	Updates Demod2D call signature and modem packet assemble helper.
stm32/unittest/src/tst_ofdm_mod.c	Refactors formatting/includes; switches to PSK packet assembler.
stm32/unittest/src/tst_ofdm_demod.c	Refactors formatting/includes; switches to PSK packet disassembler and updated LLR flow.
src/reliable_text.c	Updates `symbols_to_llrs()` call to include `bps`.
src/ofdm_mode.c	Adds `EsNodB` to configs and introduces `qam16c2` mode parameters/UW.
src/ofdm_mod.c	Switches to PSK packet assembler.
src/ofdm_internal.h	Raises `MAX_UW_BITS`, adds `EsNodB` fields, updates function prototypes for generalized PSK/QAM paths.
src/ofdm_demod.c	Uses `EsNodB`->linear EsNo conversion; switches to PSK disassembly; updates demod/error-count paths for bps=4.
src/ofdm.c	Adds amplitude-normalized qam16 demod, generalized UW symbol creation/mod/demod to bps=2/4, and new PSK packet helpers.
src/mpdecode_core.h / src/mpdecode_core.c	Generalizes Demod2D/symbol-to-LLR pipeline to variable constellation sizes (QPSK/QAM16).
src/interldpc.h / src/interldpc.c	Generalizes LDPC interleave TX modulation and uncoded error counting for bps=2/4.
src/freedv_*	Adds new FreeDV mode constant and routes OFDM data open/tx/rx through qam16c2.
octave/*	Updates Octave mode/config scripts and QAM16 test to match C port changes.
README_data.md	Documents the new QAM16C2 mode in the data mode table.
CMakeLists.txt	Adds qam16c2 test coverage (Octave Tx/C Rx and C Tx/C Rx).

Comments suppressed due to low confidence (1)

src/ofdm_demod.c:530

In the non-LDPC (hard decision) path, the code always calls qpsk_demod() and writes only two bits per symbol (bits[1]/bits[0]). For ofdm->bps == 4 (e.g. qam16c2) this will generate incorrect output and leave parts of rx_bits_char uninitialized. Either gate this branch with an assert(ofdm->bps == 2) or add the corresponding 16-QAM demod + bit packing logic for bps == 4.

          for (i = 0; i < Npayloadsymsperpacket; i++) {
            int bits[ofdm->bps];
            complex float s = payload_syms[i].real + I * payload_syms[i].imag;
            qpsk_demod(s, bits);
            rx_bits_char[ofdm_config->bps * i] = bits[1];
            rx_bits_char[ofdm_config->bps * i + 1] = bits[0];
          }

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-02-18T03:07:21Z

                       unsigned char tx_bits_char[]);
 void qpsk_modulate_frame(COMP tx_symbols[], int codeword[], int n);
 void ldpc_decode_frame(struct LDPC *ldpc, int *parityCheckCount, int *iter,
                       uint8_t out_char[], float llr[]);
 int count_uncoded_errors(struct LDPC *ldpc, struct OFDM_CONFIG *config,


interldpc.c renamed qpsk_modulate_frame() to psk_modulate_frame(int bps, ...), but this header still declares qpsk_modulate_frame(...). Please update the header to match the implementation (rename the prototype or provide a compatibility wrapper) to avoid stale/incorrect declarations for users of interldpc.h.

Copilot · 2026-02-18T03:07:21Z

  int i;
-
-  float symbol_likelihood[nsyms * QPSK_CONSTELLATION_SIZE];
-  float bit_likelihood[nsyms * QPSK_BITS_PER_SYMBOL];
-
-  Demod2D(symbol_likelihood, rx_qpsk_symbols, S_matrix, EsNo, rx_amps, mean_amp,
-          nsyms);
-  Somap(bit_likelihood, symbol_likelihood, QPSK_CONSTELLATION_SIZE,
-        QPSK_BITS_PER_SYMBOL, nsyms);
-  for (i = 0; i < nsyms * QPSK_BITS_PER_SYMBOL; i++) {
+  int constellation_points = 1 << bps;
+  float symbol_likelihood[nsyms * constellation_points];
+  float bit_likelihood[nsyms * bps];
+


symbol_likelihood and bit_likelihood are VLAs sized by nsyms * (1<<bps) and nsyms * bps. With bps=4 and large frames (e.g. 16200-bit LDPC → nsyms=4050), this is >300KB of stack per call and can overflow on smaller stacks (or in embedded/threaded contexts). Consider allocating these buffers on the heap, reusing a preallocated workspace, or otherwise bounding stack usage.

Copilot · 2026-02-18T03:07:22Z

+    config->clip_en = false;
+    config->tx_bpf_en = false;


In the qam16c2 mode block, clip_en (and tx_bpf_en just below) are assigned multiple times to the same value. Please remove the redundant assignments to keep the mode configuration concise and reduce the chance of future inconsistencies when tuning parameters.

Suggested change

config->clip_en = false;

config->tx_bpf_en = false;

Copilot · 2026-02-18T03:07:22Z

    complex float symbol = qam16_mod(tx_bits);
-    qam16_demod(symbol, rx_bits);
+    qam16_demod(symbol, rx_bits, 1.0);
    if (memcmp(tx_bits, rx_bits, 4)) {


memcmp(tx_bits, rx_bits, 4) compares only 4 bytes, but tx_bits/rx_bits are int[4] so this only checks the first element on typical platforms. Use sizeof(tx_bits) (or 4 * sizeof(int)) or compare the 4 elements explicitly so the test actually validates all bits.

Suggested change

if (memcmp(tx_bits, rx_bits, 4)) {

if (memcmp(tx_bits, rx_bits, sizeof(tx_bits))) {

Update interldpc prototype, hard-decision demod path for bps=4, fix qam16 unit-test comparison width, reduce qam16c2 config duplication, and move symbols_to_llrs scratch buffers off stack.

Set qam16c2 TX BPF prototype to filtP1100S1300 so optional tx_bpf/clip paths follow OFDM mode conventions and don't assert on enable.

Restore STM32 OFDM unittest files close to main and keep only API-alignment changes required by the qam16/psk updates.

When verbose is enabled, print CRC status and a short hexdump of decoded OFDM data payloads to help diagnose application-level framing/bitstream decode issues.

drowe67 and others added 18 commits April 4, 2024 06:36

EsNo supplied to LDPC dec now a config param

e02ef2c

removed unused EsNo_est_all_symbols

6a5687c

WIP qam16c2, ofdm_mod runs but doesn't decode in Octave

4726a97

uncoded C tx works with Octave rx

10d344c

WIP qam demod

31d8044

Merge branch 'dr-freedata-001' into dr-qam16-cport

2607516

ofdm_demod uncoded demods qam16c2

1079f80

uncoded BER 0 with ofdm_mod/ofdm_demod, LDPC decode not working yet

73482ea

refactoring to support QAM16 LDPC decode

c0821c8

first pass LDPC decoder working with QAM16

1b2a4a7

adjusted UW thresh and EsNo for LDPC decoder - PER < 0.1 with 100 pac…

2e5f845

…kets mpp at 15dB-ish SNR

qam16c2 FreeDV API support

82546c4

qam16 ctests and README_data.md

43388dd

attempt to fix stm32 build issues

1fd68ac

freedv 2020 build issue

ffd020c

fix stm32 build ...

15f565e

Merge branch 'dr-freedata-001' into dr-qam16-cport

6aeb1f5

Merge branch 'main' into dr-qam16-cport

eaa3417

Resolve src/ofdm_mode.c merge conflict while preserving both qam16 and qam16c2 mode configuration.

annyanka52 requested a review from Copilot February 18, 2026 03:02

Copilot started reviewing on behalf of annyanka52 February 18, 2026 03:02 View session

Copilot AI reviewed Feb 18, 2026

View reviewed changes

rafael2k added 3 commits February 18, 2026 03:15

Address Copilot review findings

583738c

Update interldpc prototype, hard-decision demod path for bps=4, fix qam16 unit-test comparison width, reduce qam16c2 config duplication, and move symbols_to_llrs scratch buffers off stack.

Align qam16c2 with BPF mode pattern

43c093c

Set qam16c2 TX BPF prototype to filtP1100S1300 so optional tx_bpf/clip paths follow OFDM mode conventions and don't assert on enable.

Trim STM32 cosmetic churn

e2aee0a

Restore STM32 OFDM unittest files close to main and keep only API-alignment changes required by the qam16/psk updates.

rafael2k force-pushed the dr-qam16-cport branch from cb39cfa to e2aee0a Compare February 18, 2026 12:50

Add OFDM data payload debug

33eade9

When verbose is enabled, print CRC status and a short hexdump of decoded OFDM data payloads to help diagnose application-level framing/bitstream decode issues.

rafael2k merged commit 3f8d05e into main Mar 2, 2026

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Merge dr-qam16-cport into main#1

Merge dr-qam16-cport into main#1
rafael2k merged 22 commits into
mainfrom
dr-qam16-cport

rafael2k commented Feb 18, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Feb 18, 2026

Uh oh!

Copilot AI Feb 18, 2026

Uh oh!

Copilot AI Feb 18, 2026

Uh oh!

Copilot AI Feb 18, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

	if (memcmp(tx_bits, rx_bits, 4)) {
	if (memcmp(tx_bits, rx_bits, sizeof(tx_bits))) {

Uh oh!

Conversation

rafael2k commented Feb 18, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Feb 18, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 18, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 18, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Feb 18, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants