| Thread Links | Date Links | ||||
|---|---|---|---|---|---|
| Thread Prev | Thread Next | Thread Index | Date Prev | Date Next | Date Index |
|
Hi Neven, Thank you for your very insightful and valuable feedback. There is no need to apologize for only having had time for criticism in the meeting. Your constructive criticism was exactly what I was hoping for
in the meeting. Hopefully we will have more time to discuss this in the upcoming ad hoc meetings, where we should not be as pressed for time. I agree with your comments below, and I had already gone through similar thought process myself. I think that the bottom line is that we can never come up with a single RFI transfer functions (or even few
transfer functions) that perfectly describes every possible Radio Frequency Interference.
In Montreal I also discussed this with Michal Brychta, who brough in impulse noise measurements to 802.3dg: https://www.ieee802.org/3/dg/public/May_2024/Brychta_3dg_01_0724.pdf Michal told me that he saw high variability in the measured pulse shapes, as is evident from his presentation. Michal was measuring slightly different system than what we are talking about here, but the same
fundamental principles apply. The realization that it would be very difficult to create a single (or few) transfer functions that would be fully representative of RFI is actually what lead me to the very simplified, but presumably pessimistic
noise model. Remember that the purpose of this model is to compare different modulation schemes, not to qualify a specific HW design. Ragnar From: Neven Pischl <0000345b1cd921bd-dmarc-request@xxxxxxxxxxxxxxxxx>
Hi Ragnar, The problem, as I see it, is that using any simplified model is going to be way off. In my experience, the only relatively reliable way to establish correct transfer function
(EMI -> MDI) is to build physical models that can properly represent all components of a link and expose the physical model to test conditions. By that, I mean building a link segment and two "dummy" end devices with proper test hookups to monitor what gets
induced at the ends (e.g. location of where PHYs would go) during exposure to EMI. Of course, the devil is in the details, and results can - as in real applications - depend on many minute (but significant) factors. Some of them being real performance-critical
factors (component-selection, PCB layout details etc.) and some of them measurement flukes, with bad test connections (parasitic pickup e.g.) etc. It is quite a challenge designing and "sanitizing" such a test system, but it is possible. If you succeed in building such a model, and then MEASURE the transfer functions, then you can use such measured transfer function to kind of predict what is going to happen.
Such a transfer function would ONLY work for the cases that you tested. E.g. modify any component in your physical model, such as cables, connectors, PCB layout details (as minute as they may appear), could considerably change the outcome. Keep in mind, such
changes of outcome would actually represent changes of what a PHY e.g. would 'see", so it is not a bad idea to use it for gaining understanding of what really happens and how design choices affect the EMI coupling and performance. It can be quite an educational
tool. It is actually pretty large effort to do it properly. An effort in such direction was done by Ramin and the Co. in 802.3ch, but IMO overly simple and fell very short of including
realistic conditions. Anyway, that is the basic concept that can be expanded on. I am very sorry I did not have enough time in the meeting, because I did not just want to criticize. I appreciate anyone working and contributing in that direction. I wanted
to provide reasoning - hopefully for betterment of anyone's understanding - why I consider the presented model inadequate and how I believe one can obtain useful info. I feel 60sec is inadequate. Sincerely, Neven On Thursday, July 18, 2024 at 09:16:07 AM PDT, Ahmad Chini <chiniahmad.ieee@xxxxxxxxx> wrote:
Ragnar I encourage you take a look at many measurements on Coaxial noise, screening attenuation and aging effect available elsewhere that you
may have access to. At some point I hope we get similar data presented to 802.3dm as well.
Thanks Ahmad Sent from
Mail for Windows From:
Ragnar Jonsson Ahmad, Conrad, George and Neven, Thank you for your comments and suggestions related to my noise environment presentation. I would greatly appreciate if you would be
willing to work with me off-line to improve the model. I will try to reach out to you off-line, but I may not have everyone’s email address. Ragnar P.S., While I address this email to those that brought in specific comments, I would appreciate all collaborators.
To unsubscribe from the STDS-802-3-ISAAC list, click the following link:
https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-3-ISAAC&A=1 To unsubscribe from the STDS-802-3-ISAAC list, click the following link:
https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-3-ISAAC&A=1 To unsubscribe from the STDS-802-3-ISAAC list, click the following link:
https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-3-ISAAC&A=1 To unsubscribe from the STDS-802-3-ISAAC list, click the following link: https://listserv.ieee.org/cgi-bin/wa?SUBED1=STDS-802-3-ISAAC&A=1 |