Retronetworking » Retro-GSM

h1. Uplink Rx issues with HRv1

In the [[BFI criteria]] article we've looked at the problem of how a traditional E1 BTS sets the BFI flag in its TRAU-UL output. The questions raised in that article apply equally to all 3 classic GSM codecs: FRv1, HRv1 and EFR. However, in the case of HRv1 codec there are two additional questions:

# How does the BTS decide when to set the UFI flag?

# What thresholds are used for ternary SID classification?

h2. Frame reception status

TRAU-UL frames for FR and EFR (GSM 08.60) feature only one status bit that indicates the result of channel decoding: BFI, or Bad Frame Indicator. But the counterpart for HR (GSM 08.61) features two flag bits: BFI and UFI, the latter being Unreliable Frame Indicator. Hence the question naturally arises: what are the criteria for setting UFI?

The proposed experiments described in the [[BFI criteria]] article will need to be performed with HR as well, looking for possible effects on both BFI and UFI flags.

h2. Ternary SID classification

For all 3 codecs the TRAU-UL frame structure includes two bits that convey the ternary SID flag, classifying the received frame as non-SID speech, or as valid or invalid SID. However, the critical difference between FR/EFR and HR is that for the former two there are spec-prescribed bit counting thresholds, making this classification unambiguous and vendor-neutral, but for HR the relevant spec (GSM 06.41) leaves the details up to equipment manufacturers.

We (as in community of people who seek to bring back GSM in its full glory) will need to experiment with an E1 BTS: feed different crafted inputs over Um, constructing almost-SID frames, and see how it classifies each constructed bit pattern. Possible interactions with convolutional decoder error count will also need to be explored.

h2. Clues from GSM MS DSP

Traffic channels in GSM are quite symmetric between uplink and downlink, hence the manner in which a BTS receives TCH UL is not expected to be too terribly different from the way a GSM MS receives TCH DL. We've got a recent [[HRv1_error_flags|breakthrough in understanding of TCH/HS Rx]] thanks to experiments with TI Calypso DSP - do the same principles carry over to traditional E1 BTS receiving TCH/HS UL? It would certainly be good to experiment with such BTS to gain some empirical knowledge.