Future mode

Design and test acquisition first, as it's tough.

+/- 100 Hz freq offset est range. Not sure how .... hence design acquistiion first

1:4 pilot symbol insertion rate:

1. This will mean a +/- 3Hz range (perhaps more), so we can tolerate a larger freq offset estimator error and still get sync quickly.
2. This will mean a little more syncronisation loss. But feel this is less important that sync on fading channels
3. It will also reduce error floor we are seeing on 1Hz fading using current 1 in 8 pilot insertion rate. This is making the LDPC code work harder.

One or two (<=40ms) codec frame (short) LDPC code, this will minimise latency, simplify packing/unpacking at the expense of some coding gain. It will allow single frame sync, and greatly speed sync by reducing need for complex state machines to evaluate sync state. Combined with better freq offset acquisition, less delay in syncing.

Aim for 600 bits/s speech codec to keep SNR down but using NN codec for > 700C speech quality.