Attention is currently required from: laforge, fixeria.
2 comments:
File src/coding/gsm0503_coding.c:
`sizeof(conv)` - this way it's clear that you're zero-initializing the whole buffer.
Will do.
Patch Set #1, Line 1955: osmo_con
but how would this work for EFR? As we can see, this function is used to treat EFR and FR cases, dep […]
The BFI-inducing frame of all zeros at the u(k) level is the same for FR and EFR. Strictly speaking it is neither valid FR nor valid EFR, but that's the whole point: the intended effect is that the receiving MS will flag it as a BFI, and that result will happen in both TCH/FS and TCH/EFS channel models. I have also observed experimentally that what I implemented here is exactly what sysmoBTS PHY does, and I run with EFR most of the time.
If you are wondering how I did the experimental observation of what sysmoBTS puts out, I used the TCH DL sniffing feature of FreeCalypso, tapping the DL path in the MS at the point between channel decoder output and Rx DTX handler input - that's the a_dd_0[] array in the DSP's accessible NDB page. I see a frame of 260 zero bits (including EFR) with the channel decoder bit error count reading zero, but the BFI flag is set, telling me that the CRC3 is inverted. Normal encoding of all-zeros would put 111 in CRC3, hence it is obvious to me that the proprietary PHY is generating that bad-CRC3 frame simply by emitting all zeros at the u(k) level.
To view, visit change 33063. To unsubscribe, or for help writing mail filters, visit settings.