defines the number of codes allocated to each of the Ranging domains.
2) According to the current specification, th
e BS can dynamically allocate the Ranging codes(N for Initial,
M for periodic and L for BW-request). I believe that the BS can
distribute the ranging code and it strongly depends on the currecnt
cell status. For, example, If the number of SSs in the traffic state
are relatively greater than those in the idle state, the BS will
re-allocate the Ranging code, M or L, for periodic and BW-request,
which is greater than initial Rangng code. This means that the Ranging
code for each group can be assigned in an imbalance status and it
can issue the more colliding from the small ranging code assigned
to SS. Would you correct what I am thinking is correct or not?
Itzik> The BS can allocate the codes according to the cell status,
implementation perspective or other reasons. The important issue
that the N,M and L values are dynamic and are learned by the SS
to react according to the strategy selected by the BS.
3) Regarding understanding(2), from the imbalance Ranging code allocation,
if the lack of Ranging code allocation occures due to cell status,
the Quasi-random selection may be one of effective control scheme
which selects Ranging code based on SS's connection ID I believe.
Would you correct me?
Itzik> I don't understand why Quasi-random selection can improve
any situation, you just change to collision probability domain from
code distribution to CID distribution.
I think that you can see from the answers of Jerry,Nico and
myself that Quasi-random does not improve the collision probability
in a major factor, this was also what I saw in my simulations.
4) From point
of implementation view, you mentioned that the Ranging code
selection by the SS are jointed with implementation and it will
be a manufacture propritery techniques. However, in traffic state,
because the BS is recognizing the connection ID of the SS and it
can estimate the Ranging code sequence corresponding to the ranging
procedures, to be transmitted to the BS, the Quasi-random selection
can give more benefit and estimated demodulation or something like
opertion such as quick acquisition, at the BS. However, it can not
have any benefits if that kind of selection shceme are not mentioned
in the specification. Would you correct me also?
Itzik> I don't
agree with you that the BS can recognize a SS using the code that
the SS selected in the Quazi-random scheme, taking a BS with 500 SS's
with uniform CID distribution, and having 16 codes for each domain,
you will have a code reuse factor of ~30. Therefore the SS is still
anonymous, from the BS's point of view.
5) In current specification, I think that there is only one backoff
valus in the UCD message for the Ranging procedures. What I have
question is that this single backoff valuse should be applied to all
Ranging procedures such as initial, periodic and BW-request.
This is also not unclear to me. Would you lead me from the cloud?
Itzik> The most dominant usage of the "Ranging procedures" or
CDMA contention access is for BW requests, so the backoff values
should be configured according to the status is this domain, the
other domains (i.e. Initial and Maintenance ranging) can use the
same values. Do you have other ideas?
Thanks
Changhoi Koo
Jerry Krinock wrote:
[This is a continuation of the "Re: How Are You" thread which has been going
on privately.]
Itzik,
You emailed something a few days ago which got my attention:
Itzik Kitroser wrote:
Some of the sub-channels can be used for Ranging sub-channels, in this case,
each Ranging sub-channels is build out of two regular sub-channels and
will contain 106 carriers. The MAC messages that contain reference to
Ranging sub-channels.
This is not my impression and not what it says in the current draft IEEE
P802.16a/D3-2002. On pg. 209 line 53 it says (in text which I wrote!):
The MAC shall define a single ranging channel. This single ranging channel
is composed of one or more subchannels.
and on page 211, line 19:
The length of each ranging code is a multiple of 53 bits. The default
is two subchannels allocated as the ranging channel; in this case the ranging
code length is 106.
I hope you see the difference between your remarks and the text in the
current draft. Your remarks say that there may be MORE THAN ONE "ranging
channel", and that each ranging channel is comprised of two subcarriers;
thus the ranging code length is always 2x53=106. In the current draft, it
says that there is ALWAYS ONE ranging channel, and that the code length
varies with the number of subchannels allocated to ranging. If two
subchannels are allocated to ranging, then the code length is 106 bits. If
four subchannels are allocated to ranging, then it is 212 bits. However,
the BS may also allocate one, three, five, six, etc. subchannels which would
result in code lengths of 53, 159, 265, 318, etc. respectively.
When I ran my simulations of the ranging process last year, I simulated it
both ways: with multiple ranging channels as you described, and with a
single ranging channel as
in t
he
current draft
. I found that, for a given
ranging subchannel allocation (of, say, 4 subchannels), performance was
slightly better using the single ranging channel as in the current draft
(say, a single ranging channel of code length 212) rather than as you
described (say, two ranging channels of 106 bits each). It seemed that the
increased coding gain was slightly more beneficial than the decreased
collisions in the frequency domain, which you get from having two channels
available.
In KiHo's contribution IEEE C802.16a-02/33r1, he describes selection of a
"Ranging Subchannel Index". According to the current draft, since there is
only a single ranging channel and therefore no need to index one channel.
Thus, it appears that KiHo he believes the same as you.
The way I wrote the current draft is the way I thought it was intended by
Runcom's original contributions. If you want it changed to the way you
underst
and, you
must
submit a comment s
omeday to change it. Although my
simulations show slightly better performance using the method in the current
draft, I personally do not care that much which way we do it, because in my
opinion the CDMA scheme performs poorly in either case. (But that is
another discussion!)
Regarding the current discussion by Changhoi, I have re-read the
contributions from Samsung and agree with Itzik that inter-cell interference
is sufficiently randomized by the our use of different subchannel
permutations in adjacent cells, assuming adjacent cells have different
values of the parameter ID_cell. I do not believe that the proposed
quasi-random selection will have any advantage.
Changhoi, thank you very much for restarting this discussion which brought
this misunderstanding to light. I hope you can contribute some simulations,
and prove us wrong if you may. May I suggest you please post a top-level
descript
ion or blo
ck diagram of your alg
orithms before you code the routines
in detail, so that others can review the basic assumptions. Otherwise,
we'll just have another set of simulation results that no one understands
and no one agrees with.
One final note: I have copied this message to the TGa email reflector. I
believe that this email reflector is underutilized, and I discourage the use
of private email lists for technical discussions. There may be new members
of the working group who can contribute, or at least need to begin
understanding the issues!
--
Changhoi Koo Ph.D
Senior Engineer
Global Stand and Strategy
Telecommunication R&D Center
Samsung Electronics CO.LTD
Suwon P.O BOX 105
21st FL., IT center.
416, Maetan-3dong, Paldal-gu
Suwon-si, Gyunggi-Do, Korea 442-742
TEL : +82-31-279-5091
FAX : +82-31-279-5130
Mobile : +82-16-9530-5091
E-mail : chkoo@samsung.com
chkoo@telecom.samsung.co.kr