start downsampled sbr

encoder/ixheaace_api.c

@@ -2473,6 +2473,11 @@ static IA_ERRORCODE ia_enhaacplus_enc_init(ixheaace_api_struct *pstr_api_struct,
                  ? core_ch
                  : pstr_api_struct->config[ele_idx].i_channels_mode);
 
+  WORD32 is_single_rate = (
+    pstr_api_struct->pstr_state->aot == AOT_AAC_ELD && 
+    (!pstr_api_struct->pstr_state->mps_enable
+      || (pstr_api_struct->pstr_state->mps_tree_config == TREE_212 && pstr_aac_config->sample_rate < 27713)));
+
   if (!(pstr_api_struct->config[ele_idx].aac_classic) &&
       ixheaace_is_sbr_setting_available(
           pstr_aac_config->bit_rate,
@@ -2481,7 +2486,7 @@ static IA_ERRORCODE ia_enhaacplus_enc_init(ixheaace_api_struct *pstr_api_struct,
                : pstr_api_struct->config[ele_idx].chmode_nchannels),
           pstr_aac_config->sample_rate, &core_sample_rate,
           pstr_api_struct->spectral_band_replication_tabs.ptr_qmf_tab,
-          pstr_api_struct->pstr_state->aot)) {
+          pstr_api_struct->pstr_state->aot, is_single_rate)) {
     ixheaace_str_sbr_cfg spectral_band_replication_config;
     ixheaace_initialize_sbr_defaults(&spectral_band_replication_config);
 
@@ -2490,6 +2495,7 @@ static IA_ERRORCODE ia_enhaacplus_enc_init(ixheaace_api_struct *pstr_api_struct,
     spectral_band_replication_config.crc_sbr = 0;
     spectral_band_replication_config.parametric_coding = 1;
     spectral_band_replication_config.is_esbr = pstr_api_struct->config[0].esbr_flag;
+    spectral_band_replication_config.sample_rate_mode = is_single_rate ? IXHEAACE_SINGLE_RATE : IXHEAACE_DUAL_RATE;
     if (pstr_api_struct->pstr_state->aot == AOT_AAC_ELD) {
       spectral_band_replication_config.is_ld_sbr = 1;
       spectral_band_replication_config.sbr_codec = ELD_SBR;

encoder/ixheaace_sbr_freq_scaling.c

@@ -19,6 +19,8 @@
  */
 
 #include <math.h>
+#include <stdio.h>
+#include "iusace_type_def.h"
 #include "ixheaac_type_def.h"
 #include "ixheaac_constants.h"
 #include "ixheaace_error_codes.h"
@@ -371,6 +373,22 @@ ixheaace_get_sbr_start_freq_raw(WORD32 start_freq, WORD32 qmf_bands, WORD32 fs)
   return result;
 }
 
+WORD32
+ixheaace_get_sbr_stop_freq_raw(WORD32 stop_freq, WORD32 qmf_bands, WORD32 fs) {
+  WORD32 result;
+
+  if (stop_freq < 0 || stop_freq > 13) {
+    return -1;
+  }
+
+  result = ixheaace_get_stop_freq(fs, stop_freq);
+
+  result = (result * fs / qmf_bands + 1) >> 1;
+
+  return result;
+}
+
+
 static WORD32 ixheaace_number_of_bands(WORD32 b_p_o, WORD32 start, WORD32 stop,
                                        FLOAT32 warp_fac) {
   WORD32 result = 0;
@@ -416,6 +434,24 @@ static VOID ixheaace_cum_sum(WORD32 start_value, WORD32 *ptr_diff, WORD32 length
   }
 }
 
+WORD32
+ixheaace_find_stop_band_downsample(const WORD32 sampling_freq, const WORD32 num_channels, 
+                                   const WORD32 start_freq, const WORD32 stop_freq,
+                                   const ixheaace_sr_mode sample_rate_mode, WORD32 sbr_ratio_idx,
+                                   ixheaace_sbr_codec_type sbr_codec) {
+  WORD32 stop_freq_max = sampling_freq / 2;
+  WORD32 k0, k2;
+  IA_ERRORCODE ret;
+
+  while (stop_freq > 0 && (ixheaace_get_sbr_stop_freq_raw(stop_freq, num_channels, sampling_freq) > stop_freq_max)) stop_freq_max--;
+  if ((ixheaace_get_sbr_stop_freq_raw(stop_freq, num_channels, sampling_freq) > stop_freq_max)) return -1;
+
+  ret = ixheaace_find_start_and_stop_band(sampling_freq, num_channels, start_freq, stop_freq, sample_rate_mode, &k0, &k2, sbr_ratio_idx, sbr_codec);
+  if (ret != IA_NO_ERROR) return -1;
+
+  return k2;
+}
+
 IA_ERRORCODE
 ixheaace_find_start_and_stop_band(const WORD32 sampling_freq, const WORD32 num_channels,
                                   const WORD32 start_freq, const WORD32 stop_freq,
@@ -440,6 +476,8 @@ ixheaace_find_start_and_stop_band(const WORD32 sampling_freq, const WORD32 num_c
     return IA_EXHEAACE_INIT_FATAL_SBR_INVALID_SAMPLERATE_MODE;
   }
 
+  printf("start freq: %d, stop freq: %d\n", start_freq, stop_freq);
+
   if (stop_freq < 14) {
     switch (sbr_codec) {
       case USAC_SBR: {
@@ -495,14 +533,17 @@ ixheaace_find_start_and_stop_band(const WORD32 sampling_freq, const WORD32 num_c
   } else {
     if (sampling_freq <= 32000) {
       if ((*ptr_k2 - *ptr_k0) > MAXIMUM_FREQ_COEFFS_LE32KHZ) {
+        printf("error: bad coefs (32000): %d\n", *ptr_k2 - *ptr_k0);
         return IA_EXHEAACE_INIT_FATAL_SBR_INVALID_FREQ_COEFFS;
       }
     } else if (sampling_freq == 44100) {
       if ((*ptr_k2 - *ptr_k0) > MAXIMUM_FREQ_COEFFS_EQ44KHZ) {
+        printf("error: bad coefs (44100): %d\n", *ptr_k2 - *ptr_k0);
         return IA_EXHEAACE_INIT_FATAL_SBR_INVALID_FREQ_COEFFS;
       }
     } else if (sampling_freq >= 48000) {
       if ((*ptr_k2 - *ptr_k0) > MAXIMUM_FREQ_COEFFS_GE48KHZ) {
+        printf("error: bad coefs (48000): %d\n", *ptr_k2 - *ptr_k0);
         return IA_EXHEAACE_INIT_FATAL_SBR_INVALID_FREQ_COEFFS;
       }
     } else {
@@ -511,6 +552,7 @@ ixheaace_find_start_and_stop_band(const WORD32 sampling_freq, const WORD32 num_c
   }
 
   if ((*ptr_k2 - *ptr_k0) < 0) {
+    printf("error: negative coefs: %d\n", *ptr_k2 - *ptr_k0);
     return IA_EXHEAACE_INIT_FATAL_SBR_INVALID_FREQ_COEFFS;
   }
 
@@ -572,6 +614,7 @@ ixheaace_update_freq_scale(UWORD8 *ptr_k_master, WORD32 *ptr_num_bands, const WO
       ixheaace_shellsort_int(diff0, num_bands0);
 
       if (diff0[0] == 0) {
+        printf("bad number bands at diff0\n");
         return IA_EXHEAACE_INIT_FATAL_SBR_INVALID_NUM_BANDS;
       }
 
@@ -594,6 +637,7 @@ ixheaace_update_freq_scale(UWORD8 *ptr_k_master, WORD32 *ptr_num_bands, const WO
       ixheaace_shellsort_int(diff0, num_bands0);
 
       if (diff0[0] == 0) {
+        printf("bad number bands at diff0\n");
         return IA_EXHEAACE_INIT_FATAL_SBR_INVALID_NUM_BANDS;
       }
 
@@ -643,6 +687,7 @@ ixheaace_update_freq_scale(UWORD8 *ptr_k_master, WORD32 *ptr_num_bands, const WO
   }
 
   if (*ptr_num_bands < 1) {
+    printf("bad number bands at ptr_num_bands\n");
     return IA_EXHEAACE_INIT_FATAL_SBR_INVALID_NUM_BANDS;
   }
 

encoder/ixheaace_sbr_freq_scaling.h

@@ -43,3 +43,8 @@ ixheaace_find_start_and_stop_band(const WORD32 sampling_freq, const WORD32 num_c
                                   ixheaace_sbr_codec_type sbr_codec);
 
 WORD32 ixheaace_get_sbr_start_freq_raw(WORD32 start_freq, WORD32 qmf_bands, WORD32 fs);
+WORD32 ixheaace_get_sbr_stop_freq_raw(WORD32 stop_freq, WORD32 qmf_bands, WORD32 fs);
+WORD32 ixheaace_find_stop_band_downsample(const WORD32 sampling_freq, const WORD32 num_channels, 
+                                   const WORD32 start_freq, const WORD32 stop_freq,
+                                   const ixheaace_sr_mode sample_rate_mode, WORD32 sbr_ratio_idx,
+                                   ixheaace_sbr_codec_type sbr_codec);

encoder/ixheaace_sbr_main.c

@@ -18,6 +18,7 @@
  * Originally developed and contributed by Ittiam Systems Pvt. Ltd, Bangalore
  */
 
+#include <stdio.h>
 #include <string.h>
 
 #include "ixheaac_type_def.h"
@@ -380,7 +381,7 @@ static IA_ERRORCODE ixheaace_create_env_channel(
 UWORD32
 ixheaace_is_sbr_setting_available(UWORD32 bitrate, UWORD32 num_output_channels,
                                   UWORD32 sample_rate_input, UWORD32 *ptr_core_sr,
-                                  ixheaace_str_qmf_tabs *pstr_qmf_tab, WORD32 aot) {
+                                  ixheaace_str_qmf_tabs *pstr_qmf_tab, WORD32 aot, WORD32 single_rate) {
   FLAG table_found = IXHEAACE_TABLE_IDX_NOT_FOUND;
   WORD32 idx_sr;
   WORD32 idx_ch;
@@ -399,7 +400,7 @@ ixheaace_is_sbr_setting_available(UWORD32 bitrate, UWORD32 num_output_channels,
   }
 #endif
 
-  *ptr_core_sr = sample_rate_input / 2;
+  *ptr_core_sr = single_rate ? sample_rate_input : (sample_rate_input / 2);
 
   table_found = ia_enhaacplus_enc_get_sbr_tuning_table_idx(
       bitrate, num_output_channels, *ptr_core_sr, pstr_qmf_tab, NULL, &idx_sr, &idx_ch,
@@ -407,6 +408,8 @@ ixheaace_is_sbr_setting_available(UWORD32 bitrate, UWORD32 num_output_channels,
       ((AOT_AAC_ELD == aot) ? pstr_qmf_tab->sbr_tuning_table_ld
                             : pstr_qmf_tab->sbr_tuning_table_lc));
 
+  printf("core_sr: %d, sr_input: %d\n", *ptr_core_sr, sample_rate_input);
+
   return (table_found == IXHEAACE_TABLE_IDX_NOT_FOUND) ? 0 : 1;
 }
 
@@ -490,6 +493,10 @@ VOID ixheaace_adjust_sbr_settings(const ixheaace_pstr_sbr_cfg pstr_config, UWORD
             pstr_qmf_tab->sbr_tuning_table_ld[idx_sr][idx_ch][idx_entry].stereo_mode;
         pstr_config->freq_scale =
             pstr_qmf_tab->sbr_tuning_table_ld[idx_sr][idx_ch][idx_entry].freq_scale;
+
+        if (pstr_config->sample_rate_mode == IXHEAACE_SINGLE_RATE) {
+          WORD32 stopFreq;
+        }
         break;
       }
       default: {
@@ -597,6 +604,7 @@ VOID ixheaace_initialize_sbr_defaults(ixheaace_pstr_sbr_cfg pstr_config) {
   pstr_config->sbr_harmonic = 0;
   pstr_config->sbr_ratio_idx = 0;  // NO_SBR
   pstr_config->use_low_freq_res = 0;
+  pstr_config->sample_rate_mode = IXHEAACE_DUAL_RATE;
 }
 
 static IA_ERRORCODE ia_enhaacplus_enc_update_freq_band_tab(
@@ -967,7 +975,7 @@ ixheaace_env_open(ixheaace_pstr_sbr_enc *pstr_env_encoder, ixheaace_pstr_sbr_cfg
   pstr_env_enc->str_sbr_cfg.stereo_mode =
       (params->codec_settings.num_channels == 2) ? params->stereo_mode : IXHEAACE_SBR_MODE_MONO;
 
-  if (params->codec_settings.sample_freq <= 24000 || !params->is_ld_sbr) {
+  if (params->sample_rate_mode == IXHEAACE_DUAL_RATE) {
     pstr_env_enc->str_sbr_hdr.sample_rate_mode = IXHEAACE_DUAL_RATE;
     if (params->sbr_codec == USAC_SBR) {
       pstr_env_enc->str_sbr_cfg.sample_freq = 2 * params->codec_settings.sample_freq;
@@ -982,6 +990,7 @@ ixheaace_env_open(ixheaace_pstr_sbr_enc *pstr_env_encoder, ixheaace_pstr_sbr_cfg
     pstr_env_enc->str_sbr_hdr.sample_rate_mode = IXHEAACE_SINGLE_RATE;
     pstr_env_enc->str_sbr_cfg.sample_freq = params->codec_settings.sample_freq;
   }
+  printf("sample rate mode: %d, sample freq: %d\n", pstr_env_enc->str_sbr_hdr.sample_rate_mode, pstr_env_enc->str_sbr_cfg.sample_freq);
   if (params->is_ld_sbr) {
     pstr_env_enc->str_sbr_bs.count_send_header_data = -1;
   } else {
@@ -1043,7 +1052,7 @@ ixheaace_env_open(ixheaace_pstr_sbr_enc *pstr_env_encoder, ixheaace_pstr_sbr_cfg
   pstr_env_enc->str_sbr_cfg.use_parametric_coding = params->parametric_coding;
 
   err_code = ia_enhaacplus_enc_update_freq_band_tab(
-      &pstr_env_enc->str_sbr_cfg, &pstr_env_enc->str_sbr_hdr, IXHEAACE_QMF_CHANNELS);
+      &pstr_env_enc->str_sbr_cfg, &pstr_env_enc->str_sbr_hdr, IXHEAACE_QMF_CHANNELS / (pstr_env_enc->str_sbr_hdr.sample_rate_mode == IXHEAACE_SINGLE_RATE ? 2 : 1));
   if (err_code) {
     return err_code;
   }

encoder/ixheaace_sbr_main.h

@@ -85,6 +85,7 @@ typedef struct ixheaace_str_sbr_cfg {
   WORD32 hq_esbr;
   ixheaace_sbr_codec_type sbr_codec;
   WORD32 use_low_freq_res;
+  ixheaace_sr_mode sample_rate_mode;
 } ixheaace_str_sbr_cfg, *ixheaace_pstr_sbr_cfg;
 
 typedef struct ixheaace_str_sbr_enc *ixheaace_pstr_sbr_enc;
@@ -92,7 +93,7 @@ typedef struct ixheaace_str_sbr_enc *ixheaace_pstr_sbr_enc;
 UWORD32
 ixheaace_is_sbr_setting_available(UWORD32 bitrate, UWORD32 num_output_channels,
                                   UWORD32 sample_rate_input, UWORD32 *sample_rate_core,
-                                  ixheaace_str_qmf_tabs *ptr_qmf_tab, WORD32 aot);
+                                  ixheaace_str_qmf_tabs *ptr_qmf_tab, WORD32 aot, WORD32 single_rate);
 
 UWORD32 ixheaace_sbr_limit_bitrate(UWORD32 bit_rate, UWORD32 num_channels,
                                    UWORD32 core_sample_rate, ixheaace_str_qmf_tabs *ptr_qmf_tab,

encoder/ixheaace_sbr_rom.c

@@ -48,7 +48,7 @@ const WORD32 vector_offset_def[] = {0, 1, 2, 3, 4, 5, 6, 7, 9, 11, 13, 16, 20, 2
 const WORD32 vector_stop_freq_32[14] = {32, 34, 36, 38, 40, 42, 44, 46, 49, 52, 55, 58, 61, 64};
 const WORD32 vector_stop_freq_44[14] = {23, 25, 27, 29, 32, 34, 37, 40, 43, 47, 51, 55, 59, 64};
 const WORD32 vector_stop_freq_48[14] = {21, 23, 25, 27, 30, 32, 35, 38, 42, 45, 49, 54, 59, 64};
-const WORD32 vector_stop_freq_64[14] = {20, 22, 24, 26, 28, 31, 34, 37, 41, 45, 49, 54, 59, 64};
+const WORD32 vector_stop_freq_64[14] = {20, 22, 24, 26, 29, 31, 34, 37, 41, 45, 49, 54, 59, 64};
 const WORD32 vector_stop_freq_88[14] = {15, 17, 19, 21, 23, 26, 29, 33, 37, 41, 46, 51, 57, 64};
 const WORD32 vector_stop_freq_96[14] = {13, 15, 17, 19, 21, 24, 27, 31, 35, 39, 44, 50, 57, 64};