Re: [802.3_EPOC] short term ad hoc committee on mulitple modulation profiles until January interim
Continues with the discussions of Hal and MareK,
First, some general comments on the statistics. 20+ million data is a very large sample space for cable modems; how to map it into a much smaller sample space such as the number of CNUs connect to one OCU which is in the order of two digits? The nice Gaussian sharp distribution will disappear. This may be the explanation of the difference of the two sets of data from tow MSOs.
Second, Many factors affect the SNR distribution; take one for example, the data from 20+ million CMs came from many different HFC plants, and even more, if we break them into cable segments – thousands of them. For a relative “bad” cable segment, all CMs may show lower SNR than a better cable segment. The correlations of the data have to be considered and analyzed.
Third, consider the above, for a much smaller sample space such as the CNUs connected to a given OCU, we may not see that kind of distribution; more true in N+1 or N+0 environment.
A side topic, if history tell us something, the Ethernet originally came out as a best effort protocol, at the time there were much ones that took more factors into consideration for better QOS, for example token ring, ATM, etc., where are them today?
Thanks,
Eugene
________________________________________
From: Marek Hajduczenia [marek.hajduczenia@xxxxxx]
Sent: Wednesday, December 05, 2012 2:11 PM
To: STDS-802-3-EPOC@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_EPOC] short term ad hoc committee on mulitple modulation profiles until January interim
Hal,
A good set of questions, indeed.
I have a few of my own, that did not get any time to be asked on the last call.
- First, I think that looking at 50+ million modems and drawing conclusions based on such data is a mistake from the get go. Statistically, we will have much smaller service groups than 50 million modems and we should be observing behavior per node, and variations in such behavior, rather than averaging everything. The law of large numbers is there for a reason – in here, we should be looking at trees, and not at the forest.
- Second, I find it hard to reconcile data presented on the call yesterday with data shown by Ed from BHN at the last meeting. To me, it seems that either both operators have completely different network designs, use different equipment, or perhaps the measurement methodology is different. While one data set supports clearly multiple profiles, the other one puts that observation into question. Do we then follow the larger data set just because it is larger?
- Third, in the SNR distribution per node shown on the yesterday’s call, it seems that two profiles would cut it for a grand majority of the connected modems. The question then becomes: how distant are these modems from the node itself? Is there any correlation between distance to node and observed SNR or it is rather a complex function of cable, passives and any other sources of noise? Furthermore, are all modems connected to the given node exactly the same? We assume all modems are alike, but I think it is only fair to assume they are not, and under the very same conditions may behave slightly different, especially in terms of measured SNR values (seems that some discussions yesterday support this conclusion).
- Fourth, and perhaps last for now. What is the premium (in relative terms, % wise if you like, taking a single profile equipment as a base value of 1.00) acceptable for equipment supporting multiple profiles versus single profile equipment? There is no free lunch as we all know and making devices more complex entails extra cost due to hardware and management complexity. I am trying to understand where the pain threshold is located and when it becomes simpler for the network operator to go and fix the coax problems rather than trying to address them through the use of super-intelligent and highly-adaptable equipment. Recall that Ethernet is first and foremost about simplicity and robustness, while adaptiveness and tons of options (and knobs to fine tune such options) does not sound very Ethernet-like to me.
Regards
Marek
From: Hal Roberts [mailto:Hal.Roberts@xxxxxxxxx]
Sent: Wednesday, December 05, 2012 17:46
To: STDS-802-3-EPOC@xxxxxxxxxxxxxxxxx
Subject: Re: [802.3_EPOC] short term ad hoc committee on mulitple modulation profiles until January interim
All,
Some observations based on Dave Urban’s presentation of downstream SNR field data. The purpose is to check if: A) I’ve understood the data properly and B) to see if there are some points we have consensus on (recognizing we only got partway through the presentation):
1. Impact of Analog Optics: It is often stated that the data we should be analyzing should subtract the impact of the analog optics, since this will not be present in some EPoC scenarios. Observation: The impact of the analog optics is negligible. The optics sets a baseline of about 42dB SNR for digital channels. However the problematic channels have noise that is 10dB worse than the optical baseline. Eliminating the optical noise will only improve the SNR of a 30dB modem by a fraction of a decibel. Therefore elimination of the analog optics will not substantially alter the problem or solution space. The higher SNR channels will, however, improve by elimination of the analog optics.
2. Source of Low SNR Modems: These are generally correlated to modems with low received signal power (RSSI) due to long in-home cable runs or high split ratios. Observation: However this leads to the following mystery when examining the Gateway Data……..
3. Gateway Data - Impact of Locating Modem at Home Entry Point: Locating the modem here would seem to solve both low signal problems as well as reduce in-home ingress. Surprisingly, locating the modem here has a relatively negligible affect with only a 2dB shift in average SNR and a ½ dB reduced Standard Deviation. Observation: How do we reconcile #2 and #3?
4. General Conclusion: The only potential way of eliminating the low SNR outliers is to reduce those modems experiencing a low signal (eliminating the analog optics will not do it). Therefore, locating the CNU at the entrance to the home where the signal level should be close to 0dBmV (and any in-home ingress at lower levels) would seem to be the way to do this. However, current data shows this is not the case. The only possibilities to explain this seem to be: A) Low signal is not the cause of low SNR or B) the ‘gateway’ location mysteriously still experiences low signal strength even though it is not behind a long cable run or high loss splitters.
Do I have the observations correct?
It would be interesting to see the SNR of the ‘gateway’ located modems paired to received signal levels to get a correlation between low SNR and low RSSI.
Hal
From: Salinger, Jorge [mailto:Jorge_Salinger@xxxxxxxxxxxxxxxxx]
Sent: Monday, December 03, 2012 7:40 AM
To: STDS-802-3-EPOC@xxxxxxxxxxxxxxxxx<mailto:STDS-802-3-EPOC@xxxxxxxxxxxxxxxxx>
Subject: Re: [802.3_EPOC] short term ad hoc committee on mulitple modulation profiles until January interim
Dear IEEE P802.3bn members,
Expanding on Mark's Email, attached below, during the last IEEE EPoC F2F meeting we had a lot of discussion regarding multiple modulation and coding schemes (M-MCS) for EPoC. However, the time available was not sufficient to express all views an opinions, both pros and cons. Given the interest in the topic, and the desire to make a decision on whether to use M-MCS or not for EPoC rather soon, there was consensus on creating a short-lived ad-hoc for the purpose of discussing the potential benefits and draw-backs of M-MCS in EPoC. To that end, this ad-hoc will be a forum to discuss the merits and draw-backs of M-MCS for EPoC, and to try to arrive to a recommendation on whether M-MCS should be used or not in EPoC for the next EPoC F2F meeting. While we may discuss approaches for implementing M-MCS for EPoC to facilitate the discussion on its merits or draw-backs, it is not the purpose of this ad-hoc to arrive to a recommendation on how M-MCS would be implemented even if it is d!
eemed appropriate to use it.
To that end, and after considering multiple options for timeslots for this meeting, we will hold an initial discussion tomorrow, Tuesday, at 9:30 AM ET. I tried to pick this timeslot with 2 criteria in mind: a. that the meeting not overlap with existing EPoC or DOCSIS 3.1 activities, and b. that it be possible for the widest range of participants as possible. I think that the a. criteria is met with this timeslot, but it is very difficult to pick a timeslot that meets the b. criteria for everyone (in particular, I know that this timeslot is pretty early in the West coast).
During this initial meeting we will review objectives and then start with a presentation on the benefits of M-MCS based on data collected by Comcast. Subsequent meetings will include additional presentations on the benefits of M-MCS, and presentations on the draw-backs of M-MCS and/or complications from its implementation.
Please see a meeting invitation following this Email. Please let me know should you not receive that meeting invitation.
Thanks!
Jorge
From: Mark Laubach <laubach@xxxxxxxxxxxx<mailto:laubach@xxxxxxxxxxxx>>
Reply-To: Mark Laubach <laubach@xxxxxxxxxxxx<mailto:laubach@xxxxxxxxxxxx>>
Date: Monday, November 19, 2012 6:26 PM
To: EPoC Task Force <STDS-802-3-EPOC@xxxxxxxxxxxxxxxxx<mailto:STDS-802-3-EPOC@xxxxxxxxxxxxxxxxx>>
Subject: [802.3_EPOC] short term ad hoc committee on mulitple modulation profiles until January interim
Dear IEEE P802.3bn members,
At the meeting last week, it was mutually desired by everyone to have more discussion time on the topic of multiple modulation profiles as there was not enough time at the meeting for everyone to fully share their views. The following ad hoc is charted until the January P802.3bn interim meeting in Phoenix, AZ:
Name: Multiple Modulation Profiles
Chair: Jorge Salinger
Until: January 2013 P802.3bn Interim meeting
Focus: facilitate discussion and information sharing on multiple modulation profiles
For everyone having a holiday this week, please have a happy one!
Cheers,
Mark Laubach, Chair
IEEE P802.3bn EPoC PHY Task Force
Broadband Communications Group
Broadcom Corporation
1351 Redwood Way
Petaluma, CA, 94954
[cid:image001.png@01CDD318.F27E1F00]
Tel: +1.707.792.9093
Cell: +1.650.996.2219
________________________________
From: Mark Laubach <laubach@xxxxxxxxxxxx<mailto:laubach@xxxxxxxxxxxx>>
Reply-To: Mark Laubach <laubach@xxxxxxxxxxxx<mailto:laubach@xxxxxxxxxxxx>>
Date: Monday, November 19, 2012 6:26 PM
To: EPoC Task Force <STDS-802-3-EPOC@xxxxxxxxxxxxxxxxx<mailto:STDS-802-3-EPOC@xxxxxxxxxxxxxxxxx>>
Subject: [802.3_EPOC] short term ad hoc committee on mulitple modulation profiles until January interim
Dear IEEE P802.3bn members,
At the meeting last week, it was mutually desired by everyone to have more discussion time on the topic of multiple modulation profiles as there was not enough time at the meeting for everyone to fully share their views. The following ad hoc is charted until the January P802.3bn interim meeting in Phoenix, AZ:
Name: Multiple Modulation Profiles
Chair: Jorge Salinger
Until: January 2013 P802.3bn Interim meeting
Focus: facilitate discussion and information sharing on multiple modulation profiles
For everyone having a holiday this week, please have a happy one!
Cheers,
Mark Laubach, Chair
IEEE P802.3bn EPoC PHY Task Force
Broadband Communications Group
Broadcom Corporation
1351 Redwood Way
Petaluma, CA, 94954
[cid:image002.jpg@01CDD318.F27E1F00]
Tel: +1.707.792.9093
Cell: +1.650.996.2219
________________________________
<="" p="">
________________________________
<="" p="">
________________________________
________________________________
________________________________________________________________________
To unsubscribe from the STDS-802-3-EPOC list, click the following link:
https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-3-EPOC&A=1