GSM EFR payload format

Thread view:

Enable EMail Alerts

Start New Thread

Thread rating:

Kim Hyldgaard - 02 Sep 2003 21:37 GMT

Hi,

I have a quite specific question. Perhabs somebody with technical insight
can help me:

The EFR algorithm converts 20 ms. of voice (which would be 160 bytes in
plain G711 encoding) into 244 bits = 30.5 bytes.
In the GSM specification however, two additional bytes are added. One byte
is used to protect the 65 most important bits with a CRC, the other byte is
used to to add redundancy for 4 selected bits.
These 4 bits are transmitted in 12 bits repeating each bit two times.

OK - this was the information. Now the question:
How should these redundant bits be treated? In my world redundancy means
identical. However, I have seen the T68i from SonyEricsson to produce a
strange pattern - for instance 101 or 110. I think the only valid pattern
should be 111 or 000.

Does anybody have a comment? - Or perhabs point me to the right group if
there's a better forum for my question.

Kind regards and thanks in advance
Kim Hyldgaard

Reply to this Message

jer - 03 Sep 2003 01:16 GMT

> Hi,
>
[quoted text clipped - 19 lines]
> Kind regards and thanks in advance
> Kim Hyldgaard

Huh? (blink)

Signature

jer email reply - I am not a 'ten' ICQ = 35253273
"All that we do is touched with ocean, yet we remain on the shore of
what we know." -- Richard Wilbur

Reply to this Message

John Henderson - 03 Sep 2003 09:22 GMT

> I have a quite specific question. Perhabs somebody with
> technical insight can help me:
[quoted text clipped - 15 lines]
> Does anybody have a comment? - Or perhabs point me to the right
> group if there's a better forum for my question.

I'm not familiar with EFR encoding or with G711. But it seems to
me there're many ways that 4 bits could be redundantly encoded
into 12 according to some algorithm. If the original bits are
labelled a, b, c & d, the 12 bits could be sent as aaabbbcccddd,
abcdabcdabcd, abcbcdcdadac, and so on.

Some could be negated too. Examples are a~aab~bbc~ccd~dd,
abcd~a~b~c~dabcd, etc.

Is there another piece to the puzzle? Redundancy which is not
just contiguous copying could have clear integrity advantages in
certain environmental conditions.

John

Reply to this Message

Kim Hyldgaard - 03 Sep 2003 20:38 GMT

> I'm not familiar with EFR encoding or with G711. But it seems to
> me there're many ways that 4 bits could be redundantly encoded
[quoted text clipped - 4 lines]
> Some could be negated too. Examples are a~aab~bbc~ccd~dd,
> abcd~a~b~c~dabcd, etc.

Good points.
However, usually negating is done to produce a bias, and I don't think this
is the case.
- I will check your ideas. Thanks for your reply.

Kind regards
Kim

Reply to this Message