• State TI Thinks Resolved
  • Locked Locked
  • Replies 8 replies
  • Answers 3 answers
  • Subscribers 34 subscribers
  • Views 281 views
  • Users 0 members are here
Support feedback
Options
Options
Related

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

DP83869HM: Non-transparent mode

Tommy Pedersen
Prodigy 40 points
Part Number: DP83869HM
Other Parts Discussed in Thread: DP83869

We are making three designs based on the DP83869HM. We will use it in media converter applications both for 100 and 1000 Mbps and we are using only straps to configure. To differentiate between 100 and 1000M we have a switch to pull pin 22 (JTAG_TDO/GPIO_1) either up or down. We strap down pin 45 & 46 (LED_1 & _2) to set copper AN to 10/100 and 100/1000 depending on if 100 or 1000 mode is set. We have the following customer requirements:

1. Link loss pass through, solved by pulling pin 32 (RX_CLK) down

2. Fiber force mode, solved by strap pin 47 (LED_0) up to disable fiber auto-negotiation (802.3 clause 37)

3. Non-transparent mode, i.e. copper interface auto-negotiation depends on the fiber side.

I am unsure how to solve the 3rd requirement. I have noted the text "Copper interface will support auto-negotiation, but the user will have to ensure that the speed negotiated on the copper side matches the speed fixed on the fiber side. In cases of speed mismatch between copper and fiber, interface data transmission will not be successful." in section 9.4.7 in the data sheet but is unsure what that means for this application where we have FULL control that device attached to the copper side supports the selected fiber speed and that the TP cable that connects the device to the media converter is a sufficiently good cable (typically 25 cm Cat5e not changeable by the user). I.e. if the device on the copper side supports 100M and we have fixed the fiber to 100M, will we "always" get a 100M link? And the same question for 1000M?

I also need an answer to the formal question if the DP83869HM supports non-transparent mode. Does it support that? If the answer is NO, which IC would you propose for us to use instead?

  • 0
    TI__Genius 9920 points

    Hi Tommy,

    Thanks for the question. I will have to reach out to my team to get you more information about this as like you have mentioned it is not explicitly mentioned whether or not the DP83869 supports non-transparent mode. I will have your response in about 1-2 days. Thanks for your patience.

    Best,

    Cecilia

  • 0 in reply to Cecilia Reyes
    TI__Genius 9920 points

    Hi Tommy,

    It seems that the DP83869 does not support non-transparent mode and would require a small mcu to monitor and configure the speed on the cable side for this to work. I do not believe we have a device in our portfolio that supports this mode. 

    Thanks,

    Cecilia

  • 0 in reply to Cecilia Reyes
    Prodigy 40 points

    Hi Cecilia,

    After studying the DP83869 data sheet in even higher detail, it seems that we would be able to use it for non-transparent mode anyway. I have two follow-up questions:

    1. Can you please verify that setting the ANEGSEL_1 and ANEGSEL_0 to 1 in table 17 (100M) and table 18 (1000M) would mean non-transparent mode (i.e that the copper interface only advertises the speed that the fiber interface is configured for)?

    2. Can you please also comment on how the answer in https://e2e.ti.com/support/interface/f/138/t/847431 relates to our application which have the requirements:

    • a) full duplex
    • b) fiber force mode
    • c) non-transparent mode
    • d) configurable 100/1000M

    We are concerned about the following things in the mentioned support case:

    • A) need to disable Auto MDIX, which cannot be done with straps? (we would like to keep Auto MDIX and want to only use straps)
    • B) need to have auto-negotiation for fiber (which does not fullfil the customer requirement)

    Best regards,

    Tommy

  • 0 in reply to Tommy Pedersen
    TI__Genius 9920 points

    Hi Tommy,

    Can you explain more about the use case for non-transparent mode? I am trying to understand the applications or reasons for its use.

    The DP83869 does not support this feature currently as there is no auto-neg available for it and it is a fixed speed that must be configured. 

    We are currently accepting feedback to see if this feature can be implemented in our newer devices, though. It would be very helpful to understand the need of it.

    Thanks,

    Cecilia

  • 0 in reply to Cecilia Reyes
    Prodigy 40 points

    Hi Cecilia,

    Reason: It is a firm requirement from our customer. It is not something we can change or argue about since it is a government procurement. It is not a new requirement. The government have been purchasing with this requirement for many years.

    Applications: All kinds of normal IT equipment such as computers, routers, printers that need to have a fiber optical interface instead of galvanic.

    Use case example: A cisco router with 16x RJ45 switched interfaces with speeds 10/100M or 10/100/1000M to be interfaced with media converters. In the other end may be computers, printers etc. As in any normal IT environment, with the exception that this is interconnected using. fibers instead of twisted pair. Some equipment is new, some is old and provided by other manufacturers who have had this requirement when they provided the old equipment. Some new equipment will be provided as well by other manufacturers.

    I think the requirement comes from that the government wants to assure that a 1000M link is established when the fiber link SFP supports 1000M and 100M when the SFP or supports 100M. There is no reason to negotiate regarding speed.

    I have problems understanding the rational for your NO answer, i.e. I think the data sheets states that it should work: The data sheet clearly says in table 17 and 18 that the copper side has autonegotiation (AN) with one single speed advertised if we set ANEGSEL_0 and _1 to 1. In 100M there is no AN on the fiber side and in 1000M we can, according to table 18 in the data sheet, disable AN by selecting forced mode by setting ANEG_DIS = 1.

    Section 9.4.10.9 (Auto-MDIX Resolution) states that there is no dependence between Auto-MDIX and AN in 10/100 which may indicate that there is a dependence in 1000? I fail to see any logical reason for dependence since AN on the fiber side is a part of the PCS layer while AMDIX is a function below both the PMD layer and the AN for the copper side.

    We would need technical details, not just a simple NO when the data sheets suggests that it works.

    Thanks,

    Tommy

  • 0 in reply to Tommy Pedersen
    Prodigy 40 points

    Hi Cecilia,

    We would please appreciate a more detailed answer. The data sheet suggests that it should work since the autonegotiation for the copper side would only advertise one speed, which is equivalent to transparent mode according to my understanding. Am I wrong?

    Thanks,

    Tommy

  • 0 in reply to Tommy Pedersen
    TI__Genius 9920 points

    Hi Tommy,

    Yes I can get you a more detailed answer. I will be looping in our systems engineer who can help with your inquiry. 

    Thanks,

    Cecilia

  • 0 in reply to Cecilia Reyes
    TI__Mastermind 18425 points

    Hi Tommy,

    Based on your original post and reply from April 30 it looks like you plan to support the ability to switch between the 100M/1000M media converter modes while keeping LED_0 and LED_1 strapped high. With both LED's strapped to '1' you are ensuring that only 100M copper speed gets advertised when the system is in 100M media converter mode and only 1000M copper speed gets advertised if the system is in 1000M media converter mode.  Auto-negotiation and Auto-MDIX will also stay enabled (unless disabled via software). In this case, your requirement to ensure the copper interface always links up to the speed of the fiber interface will be met.

    The datasheet statement in section 9.4.7 was added to inform users that if multiple speeds are being advertised by the copper side when the PHY is in media converter mode, then the user needs to ensure that the copper side resolves to the same speed as the fiber side. This case can happen when user's have not strapped LED_0 and LED_1. If there is a mismatch in the speeds between copper and fiber side then the PHY will not transfer data correctly. 

    -Regards

    Aniruddha