From: baobrien <baobrien@01035d8c-6547-0410-b346-abe4f91aad63>
Date: Tue, 16 Aug 2016 04:19:24 +0000 (+0000)
Subject: Added estimator reset to fsk, got TDMA thoughts down
X-Git-Url: http://git.whiteaudio.com/gitweb/?a=commitdiff_plain;h=1ab8585ce1ba24ba12d1e66ad971a0af4740b362;p=freetel-svn-tracking.git

Added estimator reset to fsk, got TDMA thoughts down


git-svn-id: https://svn.code.sf.net/p/freetel/code@2842 01035d8c-6547-0410-b346-abe4f91aad63
---

diff --git a/codec2-dev/notes_and_todo_tdma b/codec2-dev/notes_and_todo_tdma
new file mode 100644
index 00000000..97022ea4
--- /dev/null
+++ b/codec2-dev/notes_and_todo_tdma
@@ -0,0 +1,34 @@
+Aug 15, 2016
+
+Random Thoughts and todos for FreeDV FSK TDMA:
+
+Ideas for TDMA layer structure-
+-TDMAModem Struct containing
+    -half-frame 'pilot' modem struct for pilot sync 
+        -Pilot modem used to locate a UW in half frame chunks with a large bit error tolerance
+        -Once a frame is located, TDMA can set up coarse timing for slot modems
+        -Pilot modem would not keep freq est info between half frame chunks
+        -Slot modems would not keep clock offset tracking between frame-length sample chunks, this would be taken care of by coarse timing
+    -modem and deframer struct for each slot in TDMA setup
+        -Deframer would look for most likely frame coarse offset within sample buffer while in 'sync' state
+        -Slot modem deframer will feed frame offset back into main TDMA modem to handle both coarse timing and clock offset
+    -support for N slot setup at configuration
+    -sync state machine:
+        Unsynched - No sync at all
+        Pilot sync - Pilot demod has found UW but slot demods have not locked on
+        Full sync - Any slot has pulled out a UW and is individually in 'sync' state
+    -TX -- Need some way of scheduling TX bursts
+        -Should be fairly hardware-independant
+        -Hardware timer of some sort synched with ADC?
+        -Should ADC keep sampling during TX period or throw away samples? Software option, maybe?
+        
+API Skeleton-
+
+TDMA_init/TDMA_destory - do what you think they do
+    Should TDMA init take a simple mode or a more granular config? Different inits for both?
+    - Probably just a mode to begin with
+TDMA_rx(samples) - Accept some number of samps -- probably variable n samples up to some number
+TDMA_get_frames() - Get frame structs if any frames have been decoded by rx -- only hold up to 1 frame per slot -- could probably be rolled into TDMA_rx
+TDMA_reset_slot(n) - reset the estimators/sync state for any single slot -- useful from a protocol point of view if a master is scheduling slots to multiple clients
+TDMA_sched_tx(slot_n,tx_frame,tx_samps_out,tx_time_out) - Schedule a frame transmission
+    -Difficult part is to actually ensure timing here
diff --git a/codec2-dev/src/fsk.c b/codec2-dev/src/fsk.c
index 1e9f363b..e073a994 100644
--- a/codec2-dev/src/fsk.c
+++ b/codec2-dev/src/fsk.c
@@ -377,6 +377,17 @@ void fsk_set_nsym(struct FSK *fsk,int nsyms){
     
 }
 
+
+void fsk_clear_estimators(struct FSK *fsk){
+    int i;
+    /* Clear freq estimator state */
+    for(i=0; i < (fsk->Ndft/2); i++){
+        fsk->fft_est[i] = 0;
+    }
+    /* Reset timing diff correction */
+    fsk->nin = fsk->N;
+}
+
 uint32_t fsk_nin(struct FSK *fsk){
     return (uint32_t)fsk->nin;
 }
diff --git a/codec2-dev/src/fsk.h b/codec2-dev/src/fsk.h
index 5b07c7b0..4b9d3a73 100644
--- a/codec2-dev/src/fsk.h
+++ b/codec2-dev/src/fsk.h
@@ -119,6 +119,11 @@ void fsk_set_nsym(struct FSK *fsk,int nsym);
  */
 void fsk_set_est_limits(struct FSK *fsk,int fmin, int fmax);
 
+/* 
+ * Clear the estimator states
+ */
+void fsk_clear_estimators(struct FSK *fsk);
+
 /*
  * Set a MODEM_STATS struct in which to deposit demod statistics
  */