TLK106 cable length

Achim,

The TLK106 should indeed be capable of operation beyond 100m of cable.  The functionality you are seeing is not expected. 

Could you share a schematic of your design?  I think a quick review of this would be the best starting point.

Patrick

Hi,

here's my schematic:

Since that is not really readable, again in parts:

TLK106L:

Connection PHY-magentics:

Magnetics and RJ45:

Kind regards,

Achim

Hi Achim,

Could you also provide a register dump as well? There is a feature in the TLK105/6L that does power back-off when 150m cable is not used. I am curious what state this register is in as well as a few others.

Cheers,

Ross

Ross,

unfortunately that is not easily possible. The TLK106L is connected to a Beckhoff ET1100 to form an EtherCat network. It's the ET1100 that is connected to my microprocessor and which I have access to.

So I cannot access the TLK106L's registers directly but only through the EtherCat slave controller. I will need some time (and I am only back in the office on September 8) to write an program that gets the data you requested. I know what you are talking about - but I doubt if the ET1100 touches the power saving registers.

Anyway, I will check and come back to you as soon as I have new information, which will take some time.

Achim

Hi Achim,

Would it be possible for you to solder wires from MDIO/MDC on the micro to the TLK106L? All you would have to do is make sure that the PHY addresses are different and you will be able to directly access the registers. Our EVMs allow for you to change the address of the PHY.

Regards,

Ross

Ross,

unfortunately, it is not possible to directly connect the TLK106 to my micro, since the micro does not support the MI protocol. It's only the ET1100 that is doing this kind of magic...

The layout of my hardware looks like this:

But I managed to read out some of the most important registers of the TLK106... See my next post.

Achim

Ross,

here is a register dump of the registers I deemed important (it's a manual process to read those registers and I can only access 00h...1Fh):

I have done this with a 2 x 30m cable.
(What I do not understand is the 'good quality link'. On the scope, the signal level looks pretty low, especially the very short ones.)

Achim

Hi Achim,

Could you run the register dump again with the cable length you are experiencing errors with?

When you mean "pretty low" what are the levels? Could you provide a screen shot of the scope measurements you are seeing?

Regards,

Ross

Ross,

here are two diagrams showing signals directly after the transformer, after 30m, and after 60m.
 To make things better comparable, I have manually shifted the one signal by 150ns, to align both curves.

This is after 0m (500mV/div) and after 30m (200mV/div):

This is after 60m and after 30m (both 200mV/div):

If you need 90m as well, let me know.

I am using cheap CAT5e cables, since this is what my customers will use, I am afraid.

Achim

Hi,

I managed to access registers beyond 1Fh using the "MMD access control register (Register 13)" and the "MMD access address data register (Register 14)", which TI call "Register Control Register (REGCR)" and "Address or Data Register (ADDAR)".

For testing, I read out the first three registers, which I can also read out directly. Both methodes deliver the same values:

Address 0000: 3100
Address 0001: 786D
Address 0002: 2000
Address 0003: A211

Thus, the extended register access seems to work.

Now for the most interesting register: '00AEh = Power Back Off control Register

Reading it yields: 8204h

From the datasheet the binary value should be 1000 000x xx10 0000 with 'xxx' denoting the power level. Decoding my value: 1000 0010 0000 0100. Thus, the power level (marked green) is 000 = 'normal operation', which is the hoped for value.

But, there are other bits (marked red) which do not correspond to the power-on values. In the data sheet, they are marked as 'reserved' so I do not know what they stand for.

Additionally, writing a 8244 (power level 001 = Level 1 (up to 5m cable between TLK link partners)) should reduce transmission power, so that only a 5m cable is supported. However, with my 60m cable, everything still works... Somehow this does not work.

Now, I am leaving for the weekend. Hopefully you guys are doing a better job then I do!

Achim

Hi Achim,

Thank you for all the information! Let me cycle back with you after I talk with a few of my team members.

In the meantime there are cable diagnostic registers that you can configure to measure link quality and various other parameters.

REG 0x001E (Cable Diagnostic Control Register)

Bit[15] = 1 (starts to execute cable measurements)

Bits[9:8] (Link Quality Indication, see page.65)

REG 0x0170 (Cable Diagnostic Specific Control Registers)

TDR specific implementations

Regards,

Ross

Hi all,

here is an update on the permissible cable length. I tested with both my hardware as well as with TLK106LCUSEVM, which is essentially the same (TLK106L + HX1166).

2 x 30m CAT5e --> works
3 x 30m CAT5e --> fails
90m CAT7 + 30m CAT5e --> works
90m CAT7 + 2 x 30m CAT5e --> fails

So, the question is, why does TLK106 not work with CAT5e cabling???

Hi,

is there anybody out there at TI who can answer the question, why TI's PHY does not comply with their specification? Notably, why do the TLK106LCUSEVM work only up to 60m of CAT5e???

Thanks,

Achim

Hi Achim,

I have talked with the design team and they agree that our TLK106L device should operate correctly up to 150m cable. They are wondering if you could provide us with a reading of REG 0x0218? This will allow us to calculate the SNR of your system.

Regards,

Ross

Hello Ross,

are you sure about register 0218? According to the data sheet for the TLK105L/TLK106L (http://www.ti.com/lit/ds/symlink/tlk106l.pdf, revised November 2014) there is no such register, only 0215 and 021D. Typo? You meant something more like 0185h?

Achim

Hello Ross and Achim,

I'm playing with power saving modes of TLK110 which is similar to TLK105, and 106, I think. I'm using my custom tlk110 board, ~50m Cat5e cable and a router on the far end for this test.

I have written a short program, that sets all valid PWRBOCR[8:6] bits sequentally(10sec each), and during this I probed ethernet signals with a RIGOL DS1052E (single ended probes, 50MHz analog bw, 1Gs/s). I noticed, that this register works exactly the opposite way, as described. Here are the single-ended measured Vp-p values:

0b000->1.7V
0b001->1.3V
0b010->1V
0b011->0.7V
0b100->0.4V(connection lost)

I don't have any big-endian/little endian issues, because for example PHYSTS register is works as expected.

In my opinion if I have a longer cable, I have to drive it stronger, and not weaker. Therefore, if I'm right, of course that 0b001 works with 90m cable for you, becouse it's the highest power power saving mode.

I suggest you, to try all valid values on your own TLK106 device, and measure what is happening.

Is it possible that the TLK106(TLK110) adjusting drive strength without reflect this in the appropriate registers? (Maybe some not documeted registers does this exactly.)

I have also a different question:

What is the function of VRCR[3:0] bits? The datasheet only says, that

"Voltage Regulator Control This value should be ignored on read. To write to this register,
perform a read followed by a write with the desired value."

It is OK, that I have to read this first, if I would like to write it, but there is noting about the meaning of these bits...

Regards,

Attila

There is more to the chip than what's inside the manual... These are my readings for the unknown register 0218 and some more registers (for a 60m CAT5e, which still works):

0218 --> 0040

0155 --> 1D91

0180 --> 0000

001E --> 3100 (link quality is good)

Then I started TDR while the connection was still in use and got the following:

001E --> 0103 (link quality is good, diagnostic = done, process = failed (of course...))

Then I tried another one:

0090 --> 7C81 (so, bit 8 in the CR1 is '1' --> TDR AUTORUN is enabled!?)

But, when I disconnet and reconnect the cable, I still get an 0103 in register 001E...

Hmm, this doesn't really help, does it?

Achim

Attila,

Relative to the initial post, this seems like a related, but slightly different topic.  Am I interpreting that correctly?  If so, would you be agreeable to my splitting this into a separate post?  Perhaps a post on TLK110 Power Back Off and Cable Length?

Patrick

Hello Patrick,

The affected registers are the same on all these three chips, but If it shold be different topic, I can agree with that, of course.

Attila

Hello Attila,

your findings aboute the 'power back off' is interesting. For me, this seems not to apply, since my PHY is in mode 000 which is the only one where manual and chip agree.

But, I measured only slightly more then 1V peak-to-peak! This is one more question for the Texas guys: Is this correct? I even put up scope traces on this forum but no-one bothered to comment!

(And no, I cannot use TDR, since I have to access the PHY through another chip, which gets its data from the PHY - remember, this is EtherCAT! So there is no way to unplug the cable)

Achim

You are both so quick.  I am impressed with your response times to these posts.  At this point, I am afraid I will cause more confusion if I split the post.  Perhaps the best approach is to open a new post and cross reference that new post. 

I agree that there may be common interest and common learnings from the two posts.  In the event that the threads diverge, I think it will be cleaner to have two threads.

Attila,

Would you mind opening a new post and re-posting your original inquiry?  I can then add posts to cross reference the two threads.

Thanks,

Patrick

Patrick,

Or maybe we can rename this post to "TLK106 and TLK110 Cable length and Power Back Off modes" or similar?

If it's not possible, you can do, what you suggested.

Attila,

I would prefer to have two threads, if you are willing to open a new post.

Patrick

Make it two threads. It's the same for the TLK106:

Achim

Patrick,

A made the new thread, you can clean up the mess, what I've done.

Achim,

We are measuring a relatively fast signals. Are your probes in x10 mode? If I use X1 mode, I measure 1.3Vpp in no power save mode.

Thanks, Attila.  We can continue with discussions on TLK110 and Power Back Off on your new thread at:

http://e2e.ti.com/support/interface/ethernet/f/903/p/381706/1344546.aspx

We can continue discussions on TLK106 and cable length on this thread.

Patrick

Hi Achim,

I am not surprised that you are reading 1Vpk-pk. It is listed in our spec found on page 75 (SLLSEB8B). See image below:

Your scope images confirm that the device is transmitting properly. We do expect the signal to degrade as you go further down the cable, but this should not be causing link issues. I will continue my discussion with our design team and get back to you. 

Regards,

Ross

Hi Achim,

This is a register that is not listed. It gives us a value that we convert from HEX to DEC and put into an equation to find the SNR. Based on the value you provided we know that the SNR for 60m is great. 

I have a few questions from the design team for you:

1. Is it possible for you to test the link with a single 90m cable?

2. Can you provide 0x0218 readings for cable lengths above 60m?

3. You mentioned that your 3rd party board was able to establish link on all your different scenarios. Could you provide that PHY part #?

4. Can you run these tests with force 100Mbps, force Full-Duplex and force either MDI or MDIX depending on your cable?

5. Can you force the link partner and the TLK106 into 10Mbps mode just to see if you can establish link for 90m? This will greatly aid us since 10Mbps is extremely robust and reliable and will give us a baseline.

Regards,

Ross

Ross,

thanks for the clarification. So the TLK106 is right on spec.

However, I may not have been too clear in my post. My point was, that voltage levels increase when setting decreasing power levels. Like Attilla found out. I have repeated may measurement in the new thread:

http://e2e.ti.com/support/interface/ethernet/f/903/p/381706/1348276.aspx#1348276

Achim

Hi Ross,

1. Is it possible for you to test the link with a single 90m cable?

--> I only have 3 x 30m CAT5a cable and one 90m CAT7 cable.

2. Can you provide 0x0218 readings for cable lengths above 60m?

--> I will give it a try.

3. You mentioned that your 3rd party board was able to establish link on all your different scenarios. Could you provide that PHY part #?

--> It is a DP83849IVS with Pulse HX0068ANL (there are already error frames)

4. Can you run these tests with force 100Mbps, force Full-Duplex and force either MDI or MDIX depending on your cable?

5. Can you force the link partner and the TLK106 into 10Mbps mode just to see if you can establish link for 90m? This will greatly aid us since 10Mbps is extremely robust and reliable and will give us a baseline.

--> Unfortunately not. A Beckhoff ET1100 is connected to the TLK106 to run EtherCAT, which only allows 100Mbps. And the ET1100 does all the setup, so there is no way to set 10Mbps.

For layout reasons, I have connected the PHYs Tx output with the Rx-cable and the Rx input with the Tx of the cable. Thus, I have to enable Auto-MDI/MDIX (which EtherCAT dictates anyway...).

What I see on the scope when I connect a 90m line (i. e. 3 x 30m) is that it successfully sets up a connection but after a very short time drops it again and the line gets silent. This happens again and again, maybe with a 1...3 sec periode.

Have a nice weekend.

Achim

Hi Achim,

I just confirmed in lab what you have been seeing for the power backoff register. "Normal operation" (000) is giving me the proper output voltage, and when I set bits[8:6} from 000->011 the Vod lowers to similar levels you read. 

I will cycle back with the design team, but I would think that there is an error in the datasheet and 011 should be associated with 5m ... 001 should be associated with 100m.

Regards,

Ross

Hello Ross,

I tried to answer your question: "Can you provide 0x0218 readings for cable lengths above 60m?". However, that does not work.

I have to access the registers of the TL106 via an Ethercat Slave Chip (ESC), the ET1100 from Beckhoff. The latter I can only access through the Ethercat Master, which sits on the other side of the Ethernet network. So the master sends a 'read register' command to the network. The TLK106 receives this and hands it over to the ESC. This one executes the command and reads the appropriate registers of the TLK106. Response is the same way, only in reverse.

Now, with a 3 x 30m = 90m cable I do not get a proper connection. This prevents me from reading any registers when there is no stable connection.

Sorry about that. And I do not have a workaround...

Achim

Hi Achim,

What are the voltage levels of the device on pins PFBOUT and RBIAS? 

Regards,

Ross

Hello Ross,

I finally got around to do the measurements. Since I have two PHYs, there are two of them:

First PHY:

PFBOUT = 1.56V

RBIAS = 1.21V

Second PHY:

PFBOUT = 1.54V

RBIAS = 1.23V

Achim