TLK110 in unpowered state

Martin Hansel

Other Parts Discussed in Thread: TLK110, TMDSICE3359Hi @all, I am seeing a very stange behaviour of my Eth-Interface: Without powering the board, the Phy seems to respond to an incoming 10MBit-Link-signal! I could isolate this problem to the following state: The negative pulse on Pin13 (RD-) is echoed somehow on Pin16 (TD-), the positive pulse on Pin14 (RD+) has no further effect (as expected). There is no external connection between these Pins (DVM says about 10k). I have 5 identical Interfaces on the board -- the same behaviour... regards Martin

over 9 years ago

0 Patrick OFarrell over 9 years ago

TI__Mastermind 31600 points

Martin,

Could you share a schematic for review? In your testing, what is the link partner on the other end of the cable and what length of cable are you using?

Patrick

0 Martin Hansel over 9 years ago in reply to Patrick OFarrell

Intellectual 390 points

Hi Patrick, thx for your quick response. Schematic: I prepared the significant part for you, but I don't think, that this will clarify the problem. Link-partner: a HP-switch, cable-length: about 3m But: I do not have problems in the normal/powered state! -- Communication is working What I did already: -- I disconnected the transformer on the tx-line (TD+/TD-) When the negative link-pulse with an amplidue of about -1V arrives on Pin13 (RD-), I can see a distorted signal on Pin16 (TD-) with an amplitude of about -0.6V. Martin

Ethernet Port.pdf

0 Ross Pimentel over 9 years ago in reply to Martin Hansel

TI__Mastermind 23675 points

Hi Martin,

I just reviewed your schematic. Have you tried replacing the ferrite bead on the CT of the transformer with a 0 ohm resistor? Also, is there always a 3.3V supply connected to the CT of the transformer or do you disable this when you power down?

Regards,

Ross

0 Martin Hansel over 9 years ago in reply to Ross Pimentel

Intellectual 390 points

Hi Ross, in power-down there is no supply at all, and yes, I also shorted the ferrite. Again, it seems as if there is a kind of electrical connection between RD- and TD-. In the circuit I only can see the received link-pulses on RD+/- and the irregular response on TD-. ALL other pins in the circuit are on 0V and quiet. I'm thinking about a DOA of the Phys, but 10 out of 10?? Martin

0 Ross Pimentel over 9 years ago in reply to Martin Hansel

TI__Mastermind 23675 points

Hi Martin,

What type of magnetics are you using? We recommend for this PHY that you use magnetics with choke. A typical one would be Pulse's HX1198FNL.

Regards,

Ross

0 Martin Hansel over 9 years ago in reply to Ross Pimentel

Intellectual 390 points

Hi Ross, we are using the Pulse J3011, but I don't think that this issue is related to the magnetics. I ordered some new Devices of the TLK... Martin

0 Martin Hansel over 9 years ago in reply to Martin Hansel

Intellectual 390 points

I just assembled a brand new device of the TLK110 -- No change, still this "phantom-pulse" Martin

0 Ross Pimentel over 9 years ago in reply to Martin Hansel

TI__Mastermind 23675 points

Hi Martin,

Thank you for the updates. Is it possible for you to test with voltage still on the CT of the transformer (power down all other pins except CT of the transformer)?

Regards,

Ross

0 Martin Hansel over 9 years ago in reply to Ross Pimentel

Intellectual 390 points

Hi Ross, sure I can try this, is not so easy, but what do you expect by this ? If I have some Voltage on the CT, I will (back-) supply the Phy by the 50Ohm Resistors. Anyway, if the all the Voltage is ON (CT and device-supply), I do not see any unexpected behaviour. But: In the meantime, I think, I was able the show this issue on your eval-board, too. (AM335x ICE BOARD Rev 1.0A) Steps to reproduce: -- unpower to board -- connect an active partner -- trigger on the negative link-pulse on Pin2 of the RJ-Jack (Signal RDN) -- check the signal on Pin6 (Signal TDN) Can you see this, too? regards Martin

0 Ross Pimentel over 9 years ago in reply to Martin Hansel

TI__Mastermind 23675 points

Hi Martin,

Sometimes it is advised to switch the CM choke to the device side and not the cable side. Would it be possible to wire your board with the choke on the device side?

Regards,

Ross

0 Martin Hansel over 9 years ago in reply to Ross Pimentel

Intellectual 390 points

Hi Ross, from the EMC-sight, there might be a difference between the different placing of the CM-choke, but I cannot see the influence to the current problem: "If there is a negative pulse on RDN, it is echoed to TDN, if the device is unpowered" Could you already reproduce this issue on your ICE-Board ? Regards, Martin

0 Patrick OFarrell over 9 years ago in reply to Martin Hansel

TI__Mastermind 31600 points

Martin,

Since the part is unpowered, I would expect any signaling on TD to be distorted and unbalanced. The common mode choke should effectively blunt this signal.

Patrick

0 Martin Hansel over 9 years ago in reply to Patrick OFarrell

Intellectual 390 points

Hi Patrick, as far as I can understand, the TLK110 REQUIRES the CM-Choke on the device-side ? Martin

0 Patrick OFarrell over 9 years ago in reply to Martin Hansel

TI__Mastermind 31600 points

Martin,

Yes, we recommend that the common mode choke be on the device side.

Patrick

0 Martin Hansel over 9 years ago in reply to Patrick OFarrell

Intellectual 390 points

Hi Patrick, if the CM-Choke is used to solve this TLK-Issue, then it should be a requirement and not a recommendation... Regards Martin

0 Tim Iskander over 8 years ago in reply to Martin Hansel

Prodigy 220 points

Martin, Did the CM choke help with your issue? We are seeing a similar issue, except plugging a (completely) powered off board into a 1GB switch causes the switch to A) think it has a 10MB link, and B) severely disrupts traffic on the network (like devices are unable to even ping each other).

0 Martin Hansel over 8 years ago in reply to Tim Iskander

Intellectual 390 points

Hi Tim, sorry, but we didn't try this, because we are using a jack with integrated magnetics.

0 Patrick Ruf over 7 years ago in reply to Ross Pimentel

Genius 4770 points

Hi Martin, hi Ross,

Any news on this topic? We see the same behaviour on our target board (TLK110). I also see the behaviour on a TMDSICE3359.

Best regards,
Patrick

0 Ross Pimentel over 7 years ago in reply to Patrick Ruf

TI__Mastermind 23675 points

Hi Patrick,

The TLK110 requires the use of a CMC on the device side to eliminate the risk of having an unpowered device establishing link.

Kind regards,
Ross

0 Patrick Ruf over 7 years ago in reply to Ross Pimentel

Genius 4770 points

Hi Ross,

We now have tested a RJ45 jack with integrated magnetics which has a common mode choke on both sides (please see picture attached). The connected switch still establishes a link to the unpowered TLK110. Can you please tell us what the exact schematics of the magnetics should be?

Thanks and best regards,
Patrick

0 Ross Pimentel over 7 years ago in reply to Patrick Ruf

TI__Mastermind 23675 points

Hi Patrick,

Is the MAC IF driving? Are you driving any signaling into the digital pins or the clock interface pins?
I wonder if you are powering up the device from an unexpected path.

The transformer you show should work to eliminate the issue. Do you have pin 2 and 5 tied to 3V3 rail when powered up?

Kind regards,
Ross

0 Patrick Ruf over 7 years ago in reply to Ross Pimentel

Genius 4770 points

Hi Ross,

No there is no power to the TLK110 from an unexpected path. All the circuits are unpowered. Just the LAN cable is connected.

Yes, pin2 and pin 5 is tied to 3V3 rail. But not directly. There is a ferrit chip bead inductor in series (FBMH1608HM601) with 600Ohm @ 100MHz (please see picture attached, please note picture shows old RJ45 connector type).

If you can give me a direct contact I can share the whole schematic.

Attached are the signale that I have measured at the PHY (RD+, RD-, TD+ and TD-, please note the different scaling between receive and transmit signals).

Thanks for the support.

Best regards,
Patrick

0 Ross Pimentel over 7 years ago in reply to Patrick Ruf

TI__Mastermind 23675 points

Hi Patrick,

The link is because of the parasitic path that Patrick had pointed out in earlier posts.

It had been found that CMC between the device and magnetic reduces the risk of link in an unpowered state.

From what you have shown though is that the CMC is not helping to eliminate this issue.

You might need to try another RJ45 with CMC on the device side.

Kind regards,

Ross

0 Patrick Ruf over 7 years ago in reply to Ross Pimentel

Genius 4770 points

Hi Ross,

Try another RJ45 is easier said than done. Finding one that fits with all the other parameters is difficult enought. Especially using integrated magnetics. As the same effect is on the AM335xICE V2 board I suspect that TI has no solution using a RJ45 with integrated magnetics that enables Auto-MDIX as well.
Is there no solution to modify the circuit around the PHY in some way?

Best regards,
Patrick

0 Martin Hansel over 7 years ago in reply to Patrick Ruf

Intellectual 390 points

Hi Patrick, maybe a LAN-Switch like: www.ti.com/.../ts3l110.pdf might help... Martin

0 Patrick Ruf over 7 years ago in reply to Martin Hansel

Genius 4770 points

Martin, Ross,

What about having a Schottky diode from TD+ (anode) to 3.3V (cathode) and one from RD+ (anode) to 3.3V (cathode)? In the unpowered state this diode will short the pulse to the 3.3V rail. In powered state the diode will do nothing (except having a bit of a capacitance, around 6pF). I have tried this already with two BAT85 diode. It works with both the unpowered and the powered state. But I'm curious what the experts will say to this solution. Is there any downside?

@ Ross: Can you tell me how the pins TD+ and RD+ look like inside the TLK110? Do they have diodes to the supply rail? If so what kind of diodes?

Edit: I have now seen that with this modification I will only get a 10MBit/s connection to the switch.

Thanks and best regards,
Patrick

0 Martin Hansel over 7 years ago in reply to Patrick Ruf

Intellectual 390 points

Hi Patrick,
iirc, it's the negative pulse, who's the troublemaker (the positive pulse is simply a "mirror-pulse" due to the transformer).
You could try it with your BAT85 at the RD- and TD- and GND.
Martin

0 Ross Pimentel over 7 years ago in reply to Martin Hansel

TI__Mastermind 23675 points

Hi Patrick and Martin,

I am not able to share the internals of our device. However, there is protection circuitry on the TD and RD pins.
Unfortunately a CMC has been the only external component identified to help eliminate the issue.

Kind regards,
Ross

0 Patrick Ruf over 7 years ago in reply to Ross Pimentel

Genius 4770 points

Hi Martin, hi Ross,

@ Martin: A Schottky diode from RD- (cathode) to GND (anode) and TD- (cathode) to GND (anode) does the trick. With this modification I cannot see a 10MBit/s link when the TLK110 is unpowered and I have the 100MBit/s link when I switch-on the board. Also 10MBit/s connection is still possible and Auto-MDIX is working fine.
My first attempt failed because I did not take into account that the signals at the PHY go beyond 3.3V rail (probably because of DC bias of center tap).

@ Ross: Is it possible that TI check this solution as well?

Best regards,
Patrick

0 Patrick Ruf over 7 years ago in reply to Patrick Ruf

Genius 4770 points

Hi Ross,

It would help to hear the opinion of TI's TLK110/PHY experts to this solution. Maybe there are objections to this solution we have missed.

Thanks and best regards,

Patrick

0 BO over 7 years ago in reply to Patrick Ruf

TI__Genius 12760 points

Hi Patrick

let me check with the experts and come back.

Regards, Bernd

0 Ross Pimentel over 7 years ago in reply to Patrick Ruf

TI__Mastermind 23675 points

Hi Patrick and Martin,

This approach of connecting to the minus line looks to be a good approach. 6pF added capacitance should not hurt the signaling too much.

Kind regards,
Ross

Interface

Interface forum

TLK110 in unpowered state