Encoders: Add a little dithering with the fractional bit for mono mixdowns so faster shifts can be used again instead of division without introducing their own DC offset into the mixed channels.

git-svn-id: svn://svn.rockbox.org/rockbox/trunk@12246 a1c6a512-1295-4272-9138-f99709370657

1 files changed, 11 insertions, 5 deletions

diff --git a/apps/codecs/aiff_enc.c b/apps/codecs/aiff_enc.c
index f1569d2..5f99308 100644
--- a/apps/codecs/aiff_enc.c
+++ b/apps/codecs/aiff_enc.c
@@ -69,9 +69,10 @@ struct aiff_header aiff_header =
 
 /* (*) updated when finalizing file */
 
-static int    num_channels;
-uint32_t      sample_rate;
-uint32_t      enc_size;
+static int      num_channels IBSS_ATTR;
+static uint32_t sample_rate;
+static uint32_t enc_size;
+static int32_t  err          IBSS_ATTR;
 
 /* convert unsigned 32 bit value to 80-bit floating point number */
 static void uint32_h_to_ieee754_extended_be(uint8_t f[10], uint32_t l) ICODE_ATTR;
@@ -240,10 +241,14 @@ static void chunk_to_aiff_format(uint32_t *src, uint32_t *dst)
             int32_t lr1, lr2;
 
             lr1 = *(*src)++;
-            lr1 = ((int16_t)lr1 + (lr1 >> 16)) / 2;
+            lr1 = (int16_t)lr1 + (lr1 >> 16) + err;
+            err = lr1 & 1;
+            lr1 >>= 1;
 
             lr2 = *(*src)++;
-            lr2 = ((int16_t)lr2 + (lr2 >> 16)) / 2;
+            lr2 = (int16_t)lr2 + (lr2 >> 16) + err;
+            err = lr2 & 1;
+            lr2 >>= 1;
             *(*dst)++ = swap_odd_even_le32((lr1 << 16) | (uint16_t)lr2);
         } /* to_mono */
 
@@ -311,6 +316,7 @@ static bool init_encoder(void)
 
     sample_rate  = inputs.sample_rate;
     num_channels = inputs.num_channels;
+    err          = 0;
 
     /* configure the buffer system */
     params.afmt            = AFMT_AIFF;