File spandsp-raise-traintime-tolerance.diff of Package spandsp

Overview Repositories Revisions Requests Users Attributes Meta

File spandsp-raise-traintime-tolerance.diff of Package spandsp (Revision 95ecbea57c0febfc6f627c465b3c9e68)

Currently displaying revision 95ecbea57c0febfc6f627c465b3c9e68 , Show latest

Raise v17 training time tolerance, as found in iaxmodem supplied spandsp.

---
 src/v17rx.c |    8 ++++----
 1 file changed, 4 insertions(+), 4 deletions(-)

Index: spandsp-3.0.0.g15/src/v17rx.c
===================================================================
--- spandsp-3.0.0.g15.orig/src/v17rx.c
+++ spandsp-3.0.0.g15/src/v17rx.c
@@ -1064,10 +1064,10 @@ static void process_half_baud(v17_rx_sta
             s->carrier_track_i = 100.0f;
             s->carrier_track_p = 500000.0f;
 #endif
-            /* TODO: This was increased by a factor of 10 after studying real world failures.
-                     However, it is not clear why this is an improvement, If something gives
-                     a huge training error, surely it shouldn't decode too well? */
-            if (s->training_error < (V17_TRAINING_SHORT_SEG_2_LEN - 8)*FP_SCALE(4.0f)*FP_SCALE(1.0f)*constellation_spacing[s->space_map])
+            /* We need to be liberally accepting of poor fast-train training errors here.
+               Doing this affords the DTE the opportunity to make some use of whatever 
+               valid Phase C data can be decoded. */
+            if (s->training_error < (V17_TRAINING_SHORT_SEG_2_LEN - 8)*FP_SCALE(10.0f)*FP_SCALE(1.0f)*constellation_spacing[s->space_map])
             {
                 s->training_count = 0;
                 if (s->bits_per_symbol == 2)

Places

File spandsp-raise-traintime-tolerance.diff of Package spandsp (Revision 95ecbea57c0febfc6f627c465b3c9e68)

Places