Repetition in Rate Matching

[PhyLyrDude]

Hi All,

I am looking into Rate Matching in UMTS. In de-Rate Matching (if this i
the correct term), punctured bits I believe are replaced with 0.

For repetition bits, if the inputs to the de-Rate Matching are soft bit
(LLR), how do we combine the repeat bits or just discard one of th
repeats? If we combine them, how do we do it, is it just a simpl
addition.

Thanks.

[Eric Jacobsen]

On 11/24/2009 9:47 AM, PhyLyrDude wrote:
> Hi All,
>
> I am looking into Rate Matching in UMTS. In de-Rate Matching (if this is
> the correct term), punctured bits I believe are replaced with 0.

I don't know which spec you're looking at, but puncturing is typically
only done with some sort of FEC, so what to do to recover the punctured
symbols depends partly on the FEC being used. e.g., for a Reed-Solomon
codeword it doesn't much matter, but zeros are typically used.
Typically for soft-decision decoders one substitutes a
minimum-confidence value, sometimes alternating between ones and zeros.
Some decoder implementations use internal processing that optimizes
performance for depuncturing when a "depuncture" signal is indicated for
depunctured symbols.

So it depends. ;)

> For repetition bits, if the inputs to the de-Rate Matching are soft bits
> (LLR), how do we combine the repeat bits or just discard one of the
> repeats? If we combine them, how do we do it, is it just a simple
> addition.

Simple addition works well and is easy to implement. There are other
methods, but usually with tradeoffs that are harder to justify.

--
Eric Jacobsen
Minister of Algorithms
Abineau Communications
http://www.abineau.com

[PhyLyrDude]

Hi Eric,

Thanks very much for your response. Yes, the output of the de-Rat
Matching will be fed into Turbo Decoder.


>On 11/24/2009 9:47 AM, PhyLyrDude wrote:
>> Hi All,
>>
>> I am looking into Rate Matching in UMTS. In de-Rate Matching (if thi
is
>> the correct term), punctured bits I believe are replaced with 0.
>
>I don't know which spec you're looking at, but puncturing is typically
>only done with some sort of FEC, so what to do to recover the punctured
>symbols depends partly on the FEC being used. e.g., for a Reed-Solomon
>codeword it doesn't much matter, but zeros are typically used.
>Typically for soft-decision decoders one substitutes a
>minimum-confidence value, sometimes alternating between ones and zeros.
> Some decoder implementations use internal processing that optimizes
>performance for depuncturing when a "depuncture" signal is indicated fo

>depunctured symbols.
>
>So it depends. ;)
>
>> For repetition bits, if the inputs to the de-Rate Matching are sof
bits
>> (LLR), how do we combine the repeat bits or just discard one of the
>> repeats? If we combine them, how do we do it, is it just a simple
>> addition.
>
>Simple addition works well and is easy to implement. There are other
>methods, but usually with tradeoffs that are harder to justify.
>
>--
>Eric Jacobsen
>Minister of Algorithms
>Abineau Communications
>http://www.abineau.com
>