10.07.2015 Views

here - Sites personnels de TELECOM ParisTech - Télécom ParisTech

here - Sites personnels de TELECOM ParisTech - Télécom ParisTech

here - Sites personnels de TELECOM ParisTech - Télécom ParisTech

SHOW MORE
SHOW LESS

You also want an ePaper? Increase the reach of your titles

YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.

2.4. A block splitting pattern 37The γ i ′′ ’s and δ′′ i ’s are no longer pairwise in<strong>de</strong>pen<strong>de</strong>nt, even between different blocks, but r(a, t) =∑d α i − γ i ′′ + δ′′ i and the following proposition is true.Proposition 2.4.3. Let k ≥ 2 and t ∈ ( Z/(2 k − 1)Z ) ∗. Then a ∈ St,k if and only if ∑ d γ′′∑d δ′′ i .Remember that t is consi<strong>de</strong>red to be fixed so that the α i ’s and the β i ’s are consi<strong>de</strong>red to beconstants, w<strong>here</strong>as the other quantities <strong>de</strong>fined in this section <strong>de</strong>pend on a which ranges over allbinary strings on k bits and will be consi<strong>de</strong>red as random variables, whence the vocabulary weuse.2.4.2 Combining variablesIn the previous section we <strong>de</strong>fined two variables for each block. However, we are only reallyinterested in the number of carries, so one should suffice, whence the following <strong>de</strong>finition.Definition 2.4.4. We <strong>de</strong>fine ɛ i = γ i + β i − δ i , as <strong>de</strong>picted below:α 1{β 1{α i{t = 1---10---0...1---10---0...1---10---0 ,β i{α d{β d{a = ?10-0?01-1...?10-0?01-1...?10-0?01-1 .ɛ 1 ɛ i ɛ d{{Then ɛ i is “approximately” the number of carries that do not occur in the i-th block. As inthe previous subsection, we <strong>de</strong>fine ɛ ′ i = γ′ i + β i − δ i ′ and ɛ′′ i = γ′′ i + β i − δ i ′′ and Proposition 2.4.3is reformulated as follows.Proposition 2.4.5. Let k ≥ 2 and t ∈ ( Z/(2 k − 1)Z ) ∗. Then a ∈ St,k if and only if ∑ ∑d ɛ′′d β i = B = k − w H (t).2.4.3 One block: d = 1If t is ma<strong>de</strong> of only one block, we compute closed-form expressions for #C t,k,i = 2 k P (ɛ ′′ = k − i)for all i w<strong>here</strong> P (ɛ ′′ = k − i) is nothing but the probability that i carries occurs while adding tand a, or equivalently the probability that k − i carries are lost.Such a t ≠ 0 (or an equivalent one) is written t = 2 k − 2 k−α (i.e. t = 1...1} {{ }0...0} {{ }) and itsα{β=k−αweight is w H (t) = α with α ≥ 1.In the following proposition, the computations are ma<strong>de</strong> without including the binary string0...0 in contrast with what was done in Subsection 2.4.1 because it does not complicate themtoo much.Proposition 2.4.6. The distribution of ɛ ′′ is as follows:P (ɛ ′′ = 0) = 2 −β ;i

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!