Criteria for setting BFI flag¶

Aside from the case of BFI with no data, when does the BTS receiver for TCH UL set BFI=1 in its TRAU-UL output after a block decoding attempt? We can take as a given that BFI is set when CRC-3 fails, or when CRC-8 fails in the case of EFR - the most canonical BFI criterion - but are there other conditions under which BFI is set despite passing CRC? Does BFI get set when the number of convolutional decoding bit errors exceeds a certain threshold? Or when fewer than all 8 bursts were received?

Findings answers to this question is yet another avenue for potential investigation with an E1 BTS. Given such BTS, the first simple experiment would be to take a FreeCalypso GSM MS and hack it a little to selectively suppress individual bursts. Take the same logic that is already there for DTXu, but modify it to suppress just one burst, and see what happens at the BTS for the two diagonally interleaved blocks that involve that burst. Then suppress two bursts, and so on - perhaps some BFI logic in the BTS will be revealed this way.

Another line of experimentation would be a bit more difficult: it would require an SDR-based GSM MS instead of Calypso, with an actual SDR rather than virtual environment. The experiment would be to transmit blocks with some bits intentionally inverted after convolutional encoding, thereby inducing errors in the BTS channel decoder. The idea is to see if a certain number of such bit errors causes BFI to be set despite CRC still passing.

All experiments in this series would be best to perform with FRv1 codec rather than EFR, so there is only CRC-3 and not CRC-8. This way we'll have a higher chance of revealing other BFI criteria besides CRC.