Re: stds-80220-requirements: Section 4.1.2 Spectral Efficiency
Hi Dan
Thanks for organizing your comments. Here is my response to a few of
them while the others I agree with:
1.
"Sustained spectral efficiency is computed in a loaded
[DG: need to define 'loaded']
MY: I'm not sure why this is unclear. The term loaded is pretty
standard. Please clarify.
2. [DG: it was suggested that for the evaluation purposes, we use a 57-cell
network model. Thus, I propose to change 'multi-cellular' to '57-cell
evaluation network model']
MY: I think that this change is overly specific.
The evaluation group can make (and has made) progress to defining
what is meant by "multi-cellular" so I don't see this as a point of
confusion that the requirements group needs to address. Do others have
an opinion on this?
3.
"It is defined as the ratio of the expected aggregate throughput
(taking out all PHY/MAC overhead)"
[DG: 'expected' is still a vague term. Can you suggest a better one?]
MY: I think the aim of using the word "expected" is to capture the fact
that a good model of what happens in the real world is statistical (at least
most simulations to predict spectral efficiency are statistical).
..so how about "statistically expected"?
4.
"the system bandwidth (the aggregate spectral allocation in use across
the entire system)."
[DG: I don't understand the terms 'system bandwidth' and 'aggregate
spectral allocation'. I suggest the following replacement text:
'...the total spectrum employed in the 57-cell network, divided by
the total number of sectors'.]
I believe your calculation is incorrect. In essence it would
not penalize frequency planned systems for occupying more spectrum
than systems with universal reuse.
Take a cell with 3 sectors:
sector 1 occupies 2.5Mhz of spectrum in block A,
sector 2 occupies 2.5Mhz of spectrum in block B,
and sector 3 occupies 2.5Mhz of spectrum in block C
where blocks A, B, and C are separate.
Let 7.5Mbps be the total throughput of the cell including all three
sectors.
Your calculation would lead to:
spectral eff = 2.5Mbps/(7.5Mhz/3) = 1 bits/sec/hz/sector
Whereas the right calculation (and my intent is) is:
spectral eff = 2.5Mbps/7.5Mhz = 0.33 bits/sec/hz/sector
So how about a "network wide bandwidth" along with a definition?
Network Wide Bandwidth:
The network wide bandwidth is defined as the spectral allocation in
use by a network. For example, a 7 cell network with 3 sectors per
cell each occupying the same 2.5MHz allocation has a network wide
bandwidth of 2.5Mhz. A 7 cell network with each cell occupying a
different 2.5Mhz allocation has a network wide bandwidth of 7*2.5Mhz = 17.5Mhz.
Here is the text with your comments and some of the above remarks
incorporated. I'll await further clarification from you and the rest
of the group.
Current text
Sustained spectral efficiency is computed in a loaded multi-cellular network
setting. It is defined as the ratio of the expected aggregate throughput
(taking out all PHY/MAC overhead) to all users in an interior cell divided
by the system bandwidth. The sustained spectral efficiency calculation
shall assume that users are distributed uniformly throughout the network.
[Downlink > 2 bps/Hz/sector]
[Uplink >1 bps/Hz/sector]
Proposed Text
Proposed Text
A fair comparison of the spectral efficiencies of different air interfaces
can best be achieved for a single specified baseline configuration. To
facilitate the subsequent evaluation of different proposals for the 802.20
air interface, the spectral efficiency of the 802.20 air interface shall
be quoted as b/s/Hz/sector for a three sector baseline configuration.
Within this subsequent evaluation, proposals may submit respective
deployment models over and beyond the base configuration.
Consistent with the PAR requirement that is stated on a per cell basis and
the definitions shown in Appendix A, the sustained spectral efficiency of
the 802.20 air interface shall be greater than 1 b/s/Hz/sector [note there
is discussion on the specific value also see below].
Sustained spectral efficiency is computed in a loaded multi-cellular network
setting. It is defined as the statistically expected aggregate
throughput (taking out all PHY/MAC overhead) for all users in one sector divided
by the network wide bandwidth (see definition below). The sustained spectral
efficiency calculation shall assume that users are distributed uniformly
throughout the network and shall include a specification of the minimum expected
throughput per user.
Network Wide Bandwidth: The network wide bandwidth is defined as the spectral
allocation in use by a network. For example, a 7 cell network with 3 sectors per
cell each occupying the same 2.5MHz allocation has a network wide
bandwidth of 2.5Mhz. Furthermore, for example, a 7 cell network with each cell
occupying a different 2.5Mhz allocation has a network wide bandwidth of
7*2.5Mhz = 17.5Mhz.
Rationale
This proposal captures the need for a baseline comparison methodology
for ease of evaluating various proposals. By establishing a baseline
three sectored configuration, we also finesse the subtleties in
the per cell or per sector nature of the definition of spectral
efficiency.
[NOTE: The definition of "cell" in Appendix A states, "The term "cell"
refers to one single-sector base station or to one sector of a base
station deployed with multiple cells." As such, the PAR requirement of
1 bits/sec/Hz/cell is equivalent to 1 bits/sec/Hz/sector.
The value established in the PAR and by the definition in the
appendix through the definition: "Spectral efficiency: Spectral efficiency is measured in terms of bits/s/Hz/cell. (In the case
of a sectorized configuration, spectral efficiency is given as
bits/s/Hz/ sector.)"
The current text in the Requirements document -- Version 8c
has:
downlink > 2 b/s/Hz/sector, and
uplink > 1 b/s/Hz/sector
While these values are proposed in the current document. I could
not find any rationale for them on the list. The Requirements Group
still needs to agree on target values for spectral efficiency.
]
Mike
On Thu, Nov 06, 2003 at 03:48:47PM -0500, Gal, Dan (Dan) wrote:
> Mike,
>
> I think that the your latest definition is still unclear in some areas. For easier reading,
> I copied your proposal below and inserted comments in the appropriate places.
>
> -------------------------------------------------------------------------------------
> "Sustained spectral efficiency is computed in a loaded
> >>>[DG: need to define 'loaded']
>
> multi-cellular network setting.
> >>>[DG: it was suggested that for the evaluation purposes, we use a 57-cell network model. Thus,
> I propose to change 'multi-cellular' to '57-cell evaluation network model']
>
> It is defined as the ratio of the expected
> >>>[DG: 'expected' is still a vague term. Can you suggest a better one?]
>
> aggregate throughput taking out all PHY/MAC overhead) to all users in a sector
> >>>[DG: change 'to all users in a sector' to: 'for all users in one sector']
>
> divided by
> >>>[DG: change 'divided by' to: 'to']
>
> the system bandwidth (the aggregate spectral allocation in use across
> the entire system).
>
> >>>[DG: I don't understand the terms 'system bandwidth' and 'aggregate spectral allocation'.
> I suggest the following replacement text: '...the total spectrum employed in the 57-cell network, divided by the total number of sectors'.]
>
> The sustained spectral efficiency calculation shall assume that users are distributed uniformly throughout the network
> and shall include a specification of the minimum expected
> data rate/user.
>
> >>>[DG: replace 'data rate/user' with 'throughput per user']
>
> A fair comparison of the spectral efficiencies of different air interfaces
> can best be achieved for a single specified baseline configuration. To
> facilitate the subsequent evaluation of different proposals for the 802.20
> air interface, the spectral efficiency of the 802.20 air interface shall
> be quoted as b/s/Hz/sector for a three sector baseline configuration.
> Within this subsequent evaluation, proposals may submit respective
> deployment models over and beyond the base configuration.
>
> Consistent with the PAR requirement that is stated on a per cell basis and
> the definitions shown in Appendix A, the sustained spectral efficiency of
> the 802.20 air interface shall be greater than 1 b/s/Hz/sector."
>
> >>>[DG: it would be more useful to move these two paragraphs to the top]
>
>
> Dan
>
> ------------------------------------------------------------------------------------
>
> -----Original Message-----
> From: Klerer Mark [mailto:M.Klerer@flarion.com]
> Sent: Thursday, November 06, 2003 2:30 PM
> To: 'Humbert, John J [NTWK SVCS]'; Marianna Goldhammer; Michael
> Youssefmir; stds-80220-requirements@ieee.org
> Subject: RE: stds-80220-requirements: Section 4.1.2 Spectral Efficiency
>
>
>
> John
>
> Actually there is a discrepancy between the text and the slide presentation, the slide presentation states "2 bits/hertz/cell at 3km/hr". That is also what I recall from the conference call. Based on the discussion of 3 sectors per cell (as per the eval reference configuration) being used to derive that number I was willing to go along with that. I definitely would have a problem if the text changes back to per sector basis.
>
> Mark
>
> -----Original Message-----
> From: Humbert, John J [NTWK SVCS] [mailto:JHumbe01@sprintspectrum.com]
> Sent: Thursday, November 06, 2003 12:56 PM
> To: Marianna Goldhammer; Michael Youssefmir; stds-80220-requirements@ieee.org
> Subject: RE: stds-80220-requirements: Section 4.1.2 Spectral Efficiency
>
>
> Michael,
>
> For the most part the new proposal does a good job of addressing the
> issues raised on the last conference call. The problem with the new
> proposal is in the last section dealing with target values.
>
> In the last conference call the group agreed on a target value of 2
> bits/hertz/sector at 3km/hr. The question on the table was whether or
> not there needed to be another target value for a higher speed.
>
> The target values in this section are consistent with the PAR because
> the PAR specifies 1 bit/hertz/sector as a minimum value.
>
> I propose that we stick with the text in the original proposal for this
> paragraph:
>
> [Downlink > 2 bps/Hz/sector] @ 3km/hr
> [Uplink >1 bps/Hz/sector] @ 3km/hr
>
>
>
>
> John J. Humbert
> 6220 Sprint Parkway
> Mailstop KSOPHD0504 - 5D276
> Overland Park, KS 66251-6118
> PCS (816) 210-9611
>
> -----Original Message-----
> From: Marianna Goldhammer [mailto:marianna.goldhammer@alvarion.com]
> Sent: Thursday, November 06, 2003 6:19 AM
> To: Michael Youssefmir; Humbert, John J [NTWK SVCS];
> stds-80220-requirements@ieee.org
> Subject: RE: stds-80220-requirements: Section 4.1.2 Spectral Efficiency
>
> Mike,
>
> Both definitions are a mix of requirement / attempt to define
> conditions.
>
> If you look at evaluation methodology used in 3GPP/2, there are tenths
> of
> conditions influencing the spectral efficiency, still not addressing
> sector/beam number and frequency planning issue.
>
> I would propose to state in this document only the requirement. Should
> be done a study to see which are the most appropriate deployment
> architectures, for broadband data traffic, NOT limited by using the
> existing
> cellular infrastructure (see Dan's e-mails).
>
> And besides, the evaluation criteria should permit creative/advanced
> modes
> of improving spectral efficiency.
>
> Marianna
>
> -----Original Message-----
> From: Michael Youssefmir [mailto:mike@arraycomm.com]
> Sent: Thursday, November 06, 2003 9:45 AM
> To: Humbert, John J [NTWK SVCS]; stds-80220-requirements@ieee.org
> Cc: Michael Youssefmir
> Subject: stds-80220-requirements: Section 4.1.2 Spectral Efficiency
>
>
>
>
> Folks,
>
> I was on tap for updating the text in this section
> to reflect the last requirements call. The general consensus on
> the call was to move forward with specifying a definition that
> would also lead to a baseline configuration for the evaluation
> group. (Please note previous recent text from Dan, Faroukh,
> and Samir on this section)
>
> Current text
>
> Sustained spectral efficiency is computed in a loaded multi-cellular
> network
> setting. It is defined as the ratio of the expected aggregate throughput
> (taking out all PHY/MAC overhead) to all users in an interior cell
> divided
> by the system bandwidth. The sustained spectral efficiency calculation
> shall assume that users are distributed uniformly throughout the
> network.
>
>
> [Downlink > 2 bps/Hz/sector]
> [Uplink >1 bps/Hz/sector]
>
> Proposed Text
>
> Sustained spectral efficiency is computed in a loaded multi-cellular
> network
> setting. It is defined as the ratio of the expected aggregate throughput
> (taking out all PHY/MAC overhead) to all users in a sector divided
> by the system bandwidth (the aggregate spectral allocation in use across
> the entire system). The sustained spectral efficiency calculation shall
> assume that users are distributed uniformly throughout the network
> and shall include a specification of the minimum expected data
> rate/user.
>
> A fair comparison of the spectral efficiencies of different air
> interfaces
> can best be achieved for a single specified baseline configuration. To
> facilitate the subsequent evaluation of different proposals for the
> 802.20
> air interface, the spectral efficiency of the 802.20 air interface shall
> be quoted as b/s/Hz/sector for a three sector baseline configuration.
> Within this subsequent evaluation, proposals may submit respective
> deployment models over and beyond the base configuration.
>
> Consistent with the PAR requirement that is stated on a per cell basis
> and
> the definitions shown in Appendix A, the sustained spectral efficiency
> of
> the 802.20 air interface shall be greater than 1 b/s/Hz/sector.
>
> Rationale
>
> This proposal captures the need for a baseline comparison methodology
> for ease of evaluating various proposals. By establishing a baseline
> three sectored configuration, we also finesse the subtleties in
> the per cell or per sector nature of the definition of spectral
> efficiency.
>
> [NOTE: The definition of "cell" in Appendix A states, "The term "cell"
> refers to one single-sector base station or to one sector of a base
> station deployed with multiple cells." As such, the PAR requirement of
> 1 bits/sec/Hz/cell is equivalent to 1 bits/sec/Hz/sector.
> The value established in the PAR and by the definition in the
> appendix through the definition: "Spectral efficiency: Spectral
> efficiency is measured in terms of bits/s/Hz/cell. (In the case
> of a sectorized configuration, spectral efficiency is given as
> bits/s/Hz/ sector.)"
>
> The current text in the Requirements document -- Version 8c
> has:
> downlink > 2 b/s/Hz/sector, and
> uplink > 1 b/s/Hz/sector
>
> While these values are proposed in the current document. I could
> not find any rationale for them on the list. The Requirements Group
> still needs to agree on target values for spectral efficiency.
> ]
>
> Mike
>
>
>
>
>
>
>
> This mail passed through mail.alvarion.com
>
> ************************************************************************
> ****
> ********
> This footnote confirms that this email message has been scanned by
> PineApp Mail-SeCure for the presence of malicious code, vandals &
> computer
> viruses.
> ************************************************************************
> ****
> ********
> This mail was sent via mail.alvarion.com
>
> ************************************************************************
> ************
> This footnote confirms that this email message has been scanned by
> PineApp Mail-SeCure for the presence of malicious code, vandals &
> computer viruses.
> ************************************************************************
> ************