|
Joanne,
David
Shiveley and I conversed about the details on how to specify the bandwidth,
which I agreed to writeup and put on the reflector. I'll do this next week
when I return to the office.
This
detail will probably go into the eval criteria. Spectrum is allocated to
providers based on a specific block size. It is up to the provider (&
standard) to make maximum use of the spectrum within the constraints of
out-of-block/band interference. The proposal will specify an
out-of-band emission mask to be used for technology proposal evaluation.
Stay tuned.
Joseph
Cleveland
John,
In light of my email message of November 24th, which
received some support and no opposition, I propose the following addition to the text in section
4.1.2:
%%%%%%%%%%%%%%%%%%%%%%
Proposal:
Add
the following sentence to the end of section 4.1.2:
The calculation for
spectral efficiency shall be determined within a fixed size block assignment and will include the use
of sufficient
guard bands at the block edges to confine the transmitted signal energy to
a 99% level within that fixed block. Additional
details on how to calculate spectral efficiency will be addressed
within the context of the evaluation criteria.
Rationale:
This approach allows for a consistent
comparison of the spectral
efficiency in an arrangement that is consistent with
general deployment practices.
%%%%%%%%%%%%%%%%%%%%%%%
Best
regards,
Joanne
Wilson
The requirements CG seems to be blurring the lines
between evaluation criteria and the requirements for an 802.20
system. The difference between the two is that a requirement
specifies the functionality that the widget, called 802.20, must or shall
support. The evaluation criterion specifies how to determine if a proposal
has met all of the requirements.
The discussion on whether or not guard bands
should be included in the spectral efficiency calculation falls more into
the how to measure it category rather than what the target should be
category. There are some really good points in the emails below,
however I think that the issue of including or excluding guard bands is
better addressed within the context of the evaluation team.
There needs to be a foot note
added to this section stating how to treat the guard bands when
calculating SE. This can be as simple as "The calculation for spectral
efficiency [shall] [shall not] take into account the impact of guard bands
located at the edge of the block assignment". Additional details on how to
calculate SE belongs in the evaluation criteria.
If for some reason the
evaluation CG feels that this is insufficient then they are free to come
back to the requirements CG to ask for further clarification.
John J.
Humbert
6220 Sprint Parkway
Mailstop KSOPHD0504 -
5D276
Overland Park, KS
66251-6118
PCS (816)
210-9611
-----Original
Message-----
From:
owner-stds-80220-requirements@majordomo.ieee.org
[mailto:owner-stds-80220-requirements@majordomo.ieee.org] On Behalf Of Joseph
Cleveland
Sent:
Tuesday, November 25,
2003 8:54
AM
To: 'Joanne Wilson';
'Shively,
David'; 'stds-80220-requirements@ieee.org'
Subject: RE: stds-80220-requirements:
Spectral Efficiency (4.1.2)
I
concur with Bob Love's advice to 'keep it simple' and with Joann that the
bandwidth is the spectral width that contains 99% of the emitted
power. This is consistent with the FCC
definition.
-----Original
Message-----
From:
owner-stds-80220-requirements@majordomo.ieee.org
[mailto:owner-stds-80220-requirements@majordomo.ieee.org] On Behalf Of Joanne
Wilson
Sent:
Monday, November 24,
2003 11:48 AM
To: Shively, David;
stds-80220-requirements@ieee.org
Subject: RE:
stds-80220-requirements: Spectral Efficiency (4.1.2)
David,
Since we agree
that a consistent approach for all technologies is the way to
go, I think now we are just struggling a bit with the details
of how to achieve it. To some extent, this discussion has
become as much about the evaluation process (which is the
purview of another CG) as it is about how to define "network-wide
bandwidth". In keeping with Bob Love's good advice not to dive to
deeply into the details at the requirements level, let's trust (or
encourage) the Evaluation CG to create a process that provides a
consistent approach to simulating the performance of the various
proposals within a fixed block size (possibly 5 MHz, which seems to
be preferred by many) and takes into consideration the need to
confine the transmitted signal energy, say at the 99% level, to
within that fixed block. To calculate throughput in a specific
scenario, the amount of spectrum that must be allocated to guardbands to
meet that 99% level would be subtracted from the total available
spectrum prior to multiplying by the calculated spectral
efficiency. I believe the 99% number guarantees that the
variations in guardband requirements from one technology to the
next won't be all that significant. It appears to me that this
level of detail should be made a proposal to the Evaluation CG and
the definition of network-wide bandwidth in the Requirements
document could be left as is, or at least it should be less
detailed than the above.
Best
regards,
Joanne
-----Original
Message-----
From:
owner-stds-80220-requirements@majordomo.ieee.org
[mailto:owner-stds-80220-requirements@majordomo.ieee.org]On Behalf Of Shively,
David
Sent:
Wednesday,
November 19, 2003
8:35
AM
To:
stds-80220-requirements@ieee.org
Subject: RE:
stds-80220-requirements: Spectral Efficiency (4.1.2)
---------------------------------
"There seemed
last week some confusion about channelisation, freq. arrangements and
the
tabular set
of likely spectrum amounts to be entertained for viable
deployments and all coupled
with the
related issue of the evaluation
criteria."
-----------------------------------------
I completely
agree. This what I was trying to clarify. It doesn't make
sense to discuss the
adoption of
specific spectral efficiency values (e.g. 1 or 2 bps/Hz/sector)
without a consistent
way to do the
calculation. Of course, this issue also comes up
when performing comparisons
-----Original
Message-----
From:
Robert D. Love [mailto:rdlove@nc.rr.com]
Sent: Tuesday,
November 18, 2003
5:44
PM
To: djames@oak-global.com;
'Joanne Wilson'; 'Shively, David';
stds-80220-requirements@ieee.org
Subject: Re:
stds-80220-requirements: Spectral Efficiency
(4.1.2)
David, you have expressed my
concern accurately. As we get into the detailed evaluation
phase it will become more apparent how we must modify our first set
of evaluation criteria to fairly and accurately account for the
range of variability that our candidate technologies exhibit.
Therefore, we should not attempt to drill down too far, nor be too
exclusive of other options as we establish our initial
requirements.
Robert D. Love
President,
LAN Connect Consultants
7105 Leveret Circle
Raleigh, NC 27615
Phone: 919 848-6773
Mobile: 919 810-7816
email: rdlove@ieee.org
Fax: 208 978-1187
-----
Original Message -----
Sent:
Tuesday,
November 18, 2003
3:13
PM
Subject: RE:
stds-80220-requirements: Spectral Efficiency
(4.1.2)
Yes, you
just beat me to the e-mail on this one. This is really
tricky, rather like determining the length of a piece of string,
taking into account all the possible
permutations.
On the
latter point, I did not at last week's meeting also raise publicly
the related question of suitable frequency bands for MBWA (I'd
rather we called it PWDSL, for reasons I gave last week).
There are many other freq bands possible internationally,
particularly if the WG can overcome the tendency to slip back into
a mindset over always thinking in terms of 2G/3G spectrum.
Last week presentations were restricted to just PCS and similar
bands (after all we were in USA). But there is
other spectrum, and there are many other new operators to emerge
(I know, I am working with some ..).
As if
that is not presenting enough permutations, there is also
the issue we discussed some on what sort of "default"
spectrum should be used for a sensible comparison (or set of such
?) for spec. efficiency. There seemed last week some confusion
about channelisation, freq. arrangements and the tabular set of
likely spectrum amounts to be entertained for viable
deployments and all coupled with the related issue of the
evaluation criteria.
So my
provocative suggestion is - as you imply - that the regulatory
aspects be kept separate here, that the spec efficiency be defined
in terms of "number of carriers deployed" (declared) "and the
allocation block size"; that something like 10 MHz (TDD) or 2 x
5MHz (FDD) be taken as a baseline usage for all this for the
purpose of assessment. If folk want another higher (or
lower) chunk of spectrum, so be it. In any case, when it
comes to evaluation of spec. efficiency this has to
be assessed along with other related parameters which in turn
relate to how one might have to determine and invoke guard
bands or other measures (and here some bands have modest
sharing rather than so-called co-existence issues).
Proposals will clearly have to explain how the spec. efficiency
may or may not scale but may be piecewise
incremented (pilots, BCH, correlation and other
considerations etc)
So
I'd agree with your text re "...defining the spectral
efficiency ......... should be done based solely on the
technology itself". Of course that is not the end of it, one
has to supply all the other material, too. So again back to
the notion: ".......unique carriers deployed in the network,
including any required spacing between carriers.
Certainly
it seems to make things a lot simpler if one takes the aggregate
DL, UL spec. efficiency /offered traffic rather than keep carrying
the two separately, for there are so many other similar
considerations and permutations for this,
too.
I think
Bob Love made a valid point last week in this respect. There is
endless pre-occupation in some quarters (and I can see why !) over
pre-defining exactly how this calculation or that is to be done
and under a thousand permuted scenarios. I think his
valid point was that we should move ahead more firmly, not
try to settle each and every parameter prematurely; many are
inter-related, and in the longer term the requirements,
the baseline scenarios, the eval. criteria, the
proposals, the assessment and discussed possible adjustment
constitute a somewhat iterative process. (Bob, tell me if I
misconstrue, please.)
-----Original
Message-----
From:
owner-stds-80220-requirements@majordomo.ieee.org
[mailto:owner-stds-80220-requirements@majordomo.ieee.org]
On Behalf Of
Joanne Wilson
Sent: 18
November 2003
21:24
To: Shively, David;
stds-80220-requirements@ieee.org
Subject: RE:
stds-80220-requirements: Spectral Efficiency
(4.1.2)
It
seems to me that the guardband requirements that
you mentioned are probably that
which
was
needed to meet the spectral mask associated with a
given frequency allocation. Those are
regulatory
matters
that are outside of the purview of 802.20 to set and
impossible for us to know without
explicit
knowledge
of the market and band of deployment. For the
purpose of defining the spectral
efficiency
of the
air interface, I believe it should be done based
solely on the technology itself. In your
example,
you
cite cdma2000 as having several different carrier
bandwidths depending on the number of
carriers
deployed
and the allocation block size. Based on that, I don't see
how one could make a valid assessment
of
cdma2000's spectral efficiency. The PCS example may
just be a deployment issue -- in a 5 MHz
block
assignment they were able to deploy three 1.25 MHz carriers with
625 kHz guardbands on the edges.
A
good test case would then be to ask how many carriers
could be deployed in a 10 MHz block
assignment.
My
guess (one of the cdma2000 suppliers or operators
could provide a more definitive answer) is
that
they
could deploy more than 6 carriers -- in fact, probably
7 carriers with 625 kHz guardbands on
the
edges.
In that case, by your method the cdma2000 carrier bandwidth
would be about 2.14 MHz and not
All
this is to say that I think the guardband issue, at least with
respect to adjacent block protection,
completely
muddies the calculation of the spectral efficiency for
the air interface.
Regarding
UMTS, it was my understanding that ETSI had a specific work item
to modify the UMTS
air
interface so that it could be deployed with the US PCS
allocation in a 5 MHz block size.
Again,
someone
from that community can correct me if that was not the
case.
-----Original
Message-----
From:
owner-stds-80220-requirements@majordomo.ieee.org
[mailto:owner-stds-80220-requirements@majordomo.ieee.org]On Behalf Of
Shively,
David
Sent:
Tuesday,
November 18, 2003
1:59
PM
To:
stds-80220-requirements@ieee.org
Subject: RE:
stds-80220-requirements: Spectral Efficiency
(4.1.2)
You
are correct that my intention is to define a way to evaluate
spectral efficiency in a consistent
manner.
In
the case of cdma2000, the individual channels can be placed
directly adjacent to one another but
there
is
still a required guard band on either end of these
channels. For 1900 MHz systems, the guard
band
on
each end is typically 625 kHz, for a total of 1.25 MHz.
Thus, for a single cdma2000 carrier the
total
spectrum
requirement is 0.625 kHz + 1.25 MHz + 0.625 kHz = 2.5
MHz. For 2 carriers, the
total
spectrum
requirement is 3.75 MHz and for 3 carriers the total is 5
MHz. As you noted, the
requirement
may
be less at 450 MHz but there is still a clear requirement for
some guard bands and this would impact
the
calculation of spectral efficiency.
For
UMTS, the chip rate is 3.84 Mcps so the basic bandwidth
requirement is commonly quoted
as
3.84
MHz. This would be the -3 dB bandwidth or, rather,
the equivalent noise bandwidth.
I do
not know what you refer to in terms of modifications for the
US PCS bands. The 3GPP standards
do
include the definitions and specifications for the 5 MHz
frequency blocks that are part of the
bandplan
for
the US PCS bands. The 3.84 Mcps carrier
completely fits into the 5 MHz block since there is
"extra"
The
bottom line is that the spectral efficiency calculations
should be done as consistently as
possible.
From
a network operator's perspective, what interests me is what
spectrum efficiency can I get in a
certain
amount of deployed spectrum (including guard
bands).
The
alternative approach would be do use only the carrier
bandwidth without any guard bands.
However,
in
this case, for example, the UMTS bandwidth would be 3.84 MHz
rather than 5 MHz.
-----Original
Message-----
From: Joanne Wilson
[mailto:joanne@arraycomm.com]
Sent:
Tuesday,
November 18, 2003
12:38
AM
To: Shively, David;
stds-80220-requirements@ieee.org
Subject: RE:
stds-80220-requirements: Spectral Efficiency
(4.1.2)
In
principle, I agree with your underlying premise which I
believe is that all proposals should state their
overall
bandwidth
requirements in a consistent way. However, I don't
think your example is quite correct. For
example,
I
know that three cdma2000 carriers can be deployed in less
than 2x5 MHz of spectrum (e.g. cdma450
deployments
which
I believe are in about 2x4.5 MHz blocks) but W-CDMA
cannot be deployed in less than 2x5 MHz and in
fact,
I
understand that it had to be modified to be deployable
in the US PCS bands that are multiples of 2x5
MHz block
sizes. So,
I
don't believe that the 5 MHz W-CDMA channel bandwidth
includes guardbands.
I
believe it would be impossible to determine what would
be the needed guardbands for protection of adjacent block
licensees.
I
think it would be make sense to include whatever spacing
would be needed between multiple carriers in a
single
If
you agree, how about the following
approach?:
"The
network wide bandwidth is the total spectrum in use
by the unique carriers deployed in the network,
including any
required spacing between
carriers."
-----Original
Message-----
From:
owner-stds-80220-requirements@majordomo.ieee.org
[mailto:owner-stds-80220-requirements@majordomo.ieee.org]On Behalf Of
Shively,
David
Sent:
Monday,
November 17, 2003
4:06
PM
To:
stds-80220-requirements@ieee.org
Subject: RE:
stds-80220-requirements: Spectral Efficiency
(4.1.2)
Regarding these
definitions, it should be clearly understood
whether
or not
guard bands are accounted for in the calculation of
spectral
efficiency. For
UMTS (W-CDMA), the channel is usually quoted as
being
5 MHz
wide. In this case the guard bands have been
included. However,
for
cdma2000 1X (and IS-95) the channel is usually quoted as
being
1.25 MHz
wide which does not include the necessary guard
bands.
I propose the
following:
Network Wide
Bandwidth: The network wide bandwidth is the total
spectrum in
use by
the unique carriers deployed in the network, including any
required
guard bands.
Best
regards,
David
Shively
-----------------------------------------------------------
Dr. David
Shively
Cingular
Wireless
5565
Glenridge Connector, Mail Stop 950
Atlanta, GA
30342
Phone: 404 236
5909
Mobile:
404 285 5731
FAX:
404 236 5949
email:
david.shively@cingular.com
pager:
dshively@imcingular.com
-----Original
Message-----
From:
Humbert, John J [NTWK SVCS] [mailto:JHumbe01@sprintspectrum.com]
Sent:
Monday, November 17,
2003 12:08
PM
To:
stds-80220-requirements@ieee.org
Cc:
mike@arraycomm.com
Subject:
stds-80220-requirements: Spectral Efficiency
(4.1.2)
Below is the latest
version of the text that was developed at the Plenary in
Albuquerque along with a list of
the open issues for this section.
*
4.1.2 System Spectral Efficiency
(b/s/Hz/sector)
*
The system spectral efficiency of the 802.20 air interface
shall be quoted for the case of a three sector baseline
configuration [Footnote 1]. It shall be computed in a
loaded multi-cellular network setting, which shall be
simulated based on the methodology established by the
802.20 evaluation criteria group. It shall consider among
other factors a minimum expected data rate/user and/or
other fairness criteria, and percentage of throughput due
to duplicated information flow. The values shall be quoted
on a b/s/Hz/sector basis. The system spectral efficiency
of the 802.20 air interface shall be greater than X
b/s/Hz/sector.
*
Footnote 1: Since the base configuration is only required
for the purpose of comparing system spectral efficiency,
proposals may submit deployment models over and beyond the
base configuration.
*
Definition:
*
System spectral efficiency - System spectral efficiency is
defined as the ratio of the aggregate throughput
(bits/sec) to all users in the system divided by the
network wide bandwidth (Hz) and divided by the number of
sectors in the system.
*
Aggregate Throughput: Aggregate throughput is defined as
the total throughput to all users in the system (user
payload only).
*
Network Wide Bandwidth:The network wide bandwidth is the
total spectrum in use by the unique carriers deployed in
the network.
*
Open items
-
Single value vs. multiple for uplink and downlink
-
X bits/sec/Hz [note 1 b/s/Hz -or- downlink > 2
b/s/Hz/(cell or sector?) @ 3km/hr ;uplink > 1
b/s/Hz/(cell or sector?) @ 3 km/hr].
-
Actual values of spectral efficiency at higher speeds
-
TDD/FDD
John J.
Humbert
6220
Sprint Parkway
Mailstop
KSOPHD0504 - 5D276
Overland
Park, KS 66251-6118
PCS
(816) 210-9611
|