Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

Re: [802.3_ISAAC] Link Delay - ACT vs TDD cable length



Hi Scott,

Thanks for the response. Several key technical questions remain unanswered, and the Task Force needs closure on them before a 90ns delay ceiling can be considered.

 

  1. TDD timing budget still not provided – These are not fully documented
    1. Propagation alone for 17-18m at typical VF = 0.66 is ~85-90ns
    2. That leaves no margin for:

                                                               i.      Turn around – channel must settle before valid data can be interpreted

                                                             ii.      Guard interval – Quite period is inserted between TX and RX bursts to prevent collisions, intersymbol overlap, and signal residuals – https://www.ieee802.org/3/dm/public/adhoc/062625/jonsson_3dm_01_06_26_25.pdf

                                                           iii.      Clock uncertainty due to noise – possible causes: CDR wander, jitter, AFE gain settling, etc.

                                                           iv.      AFE/DSP switching – possible causes: TX filters, RX gain and filters, DSP must relock EQ task, etc.

 

  • These overheads are not optional in any TDD PHY as these are 10s of nanoseconds, and these have not been shown to the Task Force.
  • This means realistic delays would be proposed 90ns + (30-50ns overhead) = Link Delays around 120-140ns
  • TDD proposals have hinted at these, but nothing numerical has been provided – These values could be sufficiently higher than what I suggested above https://www.ieee802.org/3/dm/public/0325/Chini_3dm_01a_0325.pdf

 

  1. Velocity factor inconsistency
    1. For about a year, 5ns/m has been used consistently for automotive coax. Your recent reply instead assumes 4.3ns/m which is best-case VF for a single cable and not the guaranteed minimum
    2. If 5ns/m was correct in the past work – which you co-authored it must be used now. https://www.ieee802.org/3/dm/public/050125/gorshe_3dm_01a_250501.pdf (suggesting 0.66 = 5ns/m should be used, as there are coax types that are in this range)

 

  1. ACT should not be limited by a TDD constraint
    1. ACT has no duplex turn around penalty or guard interval
    2. Its reach is determined by IL not duplex timing

 

  1. September presentation already showed the correct method
    1. In Sept 2025 presentation Houck/Cordaro/Chimento we demonstrated

                                                              i.      Insertion Loss must drive link length – not delay

                                                             ii.      IL budget of -23.08dB and cable IL ~0.8dB/m achieves 28.9m reach

                                                           iii.      At 4-5ns/m this naturally results in 100-150ns propagation delay

                                                           iv.      Therefore, delay should be derived from length and IL, not imposed first

 

  • This is consistent with established 802.3 methodology

 

  1. Why would 802.3dm adopt a link delay limit that is lower than 802.3ch?
    1. 802.3ch allow 94ns.
    2. Proposing 90ns for more complex topologies than 802.3ch is technically inconsistent

 

  • IL must justify any fixed delay ceiling and channel characteristics – not by TDD turn around limitations

 

Requests

  1. TDD timing numbers (t_turn, t_guard, t_uncertainty)
  2. Clarification of VF assumptions (5ns/m vs. 4.3ns/m)
  3. Explanation of why ACT should inherit TDD timing constraints and limit the ACT link length below the IL capabilities
  4. Alignment with the IL-based reach method shown in September
  5. Justification for adopting a link-delay limit lower than 802.3ch?

 

Thanks,

TJ Houck

 

 

From: Scott.Muma@xxxxxxxxxxxxx <Scott.Muma@xxxxxxxxxxxxx>
Sent: Monday, November 17, 2025 6:52 PM
To: Houck TJ (ATV ES M S) <TJ.Houck@xxxxxxxxxxxx>; STDS-802-3-ISAAC@xxxxxxxxxxxxxxxxx
Subject: RE: Link Delay - ACT vs TDD cable length

 

CautionThis e-mail originated outside Infineon Technologies. Please be cautious when sharing information or opening attachments especially from unknown senders. Refer to our intranet guide to help you identify Phishing email.

 

Hi TJ,

Thanks for your thoughtful comments, I appreciate your time and effort providing this additional information. 

 

The presentation did not claim that higher link segment delays are “out of scope” but made the point that the data presented and available so far support that the vast majority of automotive link segments (including within commercial vehicles) are <=15m. 

 

The PAR does not state “up to at least 15 m”, but through consensus the 802.3dm task force approved the objective “up to at least 15m reach on at least one type of automotive cabling”.  I would agree that this means greater than or equal to 15m on at least one type of automotive cabling in automotive operating conditions is the minimum.  I believe the 15m objective was established through robust discussion, consensus building, and data sharing, including some on the reflector which is consistent with some of the points you make below.  I get the sense that similar information to what you provide below was available when establishing the 15m objective, but if that was not the case or the group wishes to update the consensus objectives that is welcome.  You argue below for using a minimum of 32m of 5ns/m cable to establish the maximum propagation delay, and going well beyond (>2x) the objective is great in the case that it doesn’t lead to suboptimal results for the vast majority. 

 

As discussed in the presentation a link delay of 90ns allows >15m for cables with 5ns/m propagation delay and >20m for cables with 4.3ns/m propagation delay.  So 90ns does not put a ceiling at 15m.  Margin for TDD turnaround, EMC filtering, PoC filter delay are on top of the link segment delay, so do not need to be included in the 90ns (not that you said they are, just to be clear). 

 

I expect the ACT proposed protocol can support links >15m with a propagation delay of 90ns, so a propagation delay of 90ns doesn’t impose any changes on ACT.  The current TDD protocol proposal would not have overlap between the HS and LS bursts with a 90ns delay because of the chosen parameters, while a much larger delay would lead to overlap.  So the proposed 90ns link segment delay works for both ACT and TDD and allows both to meet the approved objectives.  If the ACT proposal specifies a much higher link segment delay it would still meet the objectives, but I wouldn’t be able to say what that limit could be.

 

If consensus can be reached that higher link segment propagation delays or longer cables is an objective of the task force, or a necessary specification in the draft text, then new TDD parameters could be proposed and studied. 

 

Thanks for clarifying your goal.  My goal is to propose specifications that will allow both ACT and TDD to be optimized and meet the approved objectives.  If the approved objectives are inadequate vs. the physical reality or scope definition, or no longer the consensus, I’m happy to see that amended so we can once again have a common goal.

 

Best Regards,

Scott

 

 

From: TJ Houck <TJ.Houck@xxxxxxxxxxxx>
Sent: Monday, November 10, 2025 2:23 AM
To: STDS-802-3-ISAAC@xxxxxxxxxxxxxxxxx
Subject: [802.3_ISAAC] Link Delay - ACT vs TDD cable length

 

EXTERNAL EMAIL: Do not click links or open attachments unless you know the content is safe

Hi Scott,

Thanks for the presentation today. I want to address the recurring claim that longer-reach or higher-delay channels are “out of scope” or represent a market too small to justify inclusion. I think this characterization is inaccurate.

 

The OICA commercial-vehicle segment—which explicitly includes light-commercial vans, heavy trucks, and buses—represents tens of millions of vehicles globally each year. These platforms use camera and sensor links with harness lengths that can exceed 15meters and one-way delays of >90 ns, even in baseline configurations. These are in-scope automotive applications, not niche prototypes.

 

The IEEE 802.3dm PAR states “up to at least 15 m,” establishing a minimum, not a ceiling. Nothing in the PAR prevents support for longer channels when the insertion-loss and EMC budgets remain valid. Designing for only 15m assumes away real, measurable vehicle topologies already in production.

 

1. The technical issue remains unresolved

As Ragnar noted Relationship between TDD IBG , TDD proponents have not addressed the fundamental problem. Instead of reconciling this with actual vehicle geometries, the conversation keeps shifting to “market adequacy” (“few vehicles need longer runs”). That framing doesn’t eliminate the electrical risk. It only changes the topic.

 

2. The relationship between insertion loss and delay is being reversed. This is not a “delay-limit” problem but a channel-design problem.

The standard should not constrain insertion loss because of link delay — it should define insertion-loss first, and let the resulting delay follow from the physical channel.

When delay is capped at 90 ns, we are effectively forcing an arbitrary length ceiling, reducing the permissible IL margin for longer-body vehicles. The spec should instead allow full IL cable lengths.

 

3. Why 160 ns is the technically defensible number?

Automotive coax/STP VF ≈ 0.66 – 0.78 c (≈ 5 – 4.3 ns/m). Some cabling can be <0.66 velocity

160 ns ÷ 5 ns/m ≈ 32 m of cable reach or shorter cable with a lower Velocity factor cable.

 

This aligns with realistic routing in the upper end of OICA commercial classes (e.g., articulated buses ≈ 18 m, tractor + trailer ≈ 16 m + coupling ≈ 24–28 m effective path).

A 160 ns budget therefore covers the complete OICA “commercial vehicle” range — LCV, heavy truck, and bus — while preserving adequate guard margin for TDD turnaround, EMC filtering, and PoC filter delay.

 

4. Public data contradicts “15 m is enough”

Passenger cars already reach 12 – 13 m: Krieger (VW Group) presented Typical Automotive Harness Topologies to 802.3ch with 0.5 – 12.5 m total channel lengths for sedans/compacts. As stated earlier 15meters is the minimum, not the ceiling.

 

Long-body LCVs: adding roof-rail, high-roof A-pillar, cross-dash, service loops, and inlines yields ≈ 16 – 19 m which could lead to delays >90ns

 

Breakdown

  • Roof-rail and pillar drop extension = +2-3meters
  • Cross dash traverse = +1-2meters
  • Service and door swing = 1-1.5meters
  • Connectors – detouring = 0.5-1meter

Total additional over Sedan = 4-6meters = 16.5 to 18.5meters

 

Trailers/attachments:

  • Van + car-hauler (16 – 24 ft) → 15 – 20 m total.
  • Tractor + 53′ semi → 24 – 28 m continuous run.

 

Public references (all open):

 

  1. Krieger – Typical Automotive Harness Topologies, IEEE 802.3ch (public, 0.5–12.5 m channels).

https://grouper.ieee.org/groups/802/3/ch/public/mar18/krieger_3ch_01a_0318.pdf 

  1. IndexBox – Trailer and Semi-Trailer World Market Overview 2024 (~11 M units).

https://www.indexbox.io/blog/trailer-and-semi-trailer-world-market-overview-2024-1/

  1. Mercedes Upfitter Guidelines – Sprinter 170 Ext High Roof Video Harness Lengths (3 × 15 m).

https://www.mercedes-benz-vans.ca/content/dam/mb-vans/us/upfitter/220523_ARL_Sprinter_907_2022_ENU_fin.pdf

https://www.mbvans.com/en/upfitter/tech-info/bulletins

 

My goal in raising this is to ensure that the Task Force bases its delay assumptions on physical reality and scope definition, not on an unverified estimate of “market adequacy.”

 

Best Regards,

TJ Houck

 

Infineon Technologies Americas Corp. – Detroit

System Architect

tj.houck@xxxxxxxxxxxx

 


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