TM4C1294KCPDT: Ethernet Link Issue

Hi,

To clarify on some of your questions.  We built 10 identical boards and have noticed that 6 out of the 10 are having some intermittent issues establishing a link.  I have tested the intermittent boards with different ethernet switches and noticed that some work better than others.  Also I should add that we now believe all boards were working fine at one point but somewhere though our testing they went bad.

I have been able to track down the issue a little farther since last night. All of the failing boards have lower resistance from 3.3V to ground measuring less than 200 ohms where as good boards are like 7K. 

The issue has been traced back to pins 53, 54, 56, and 57 on the Tiva MCU ethernet interface (RX-, RX+, TX-, and TX+). We lifted these pins up on the PCB and the resistance between 3.3V and ground returns to normal.  I also replaced one of these Tiva processors with a new one and the resistance and current draw returned to normal. Seems like something it causing the RX and TX lines to get shorted out inside the processor.  Have there been any reports of ESD sensitivity or anything else that could be causing this issue? 

Hi,

  If you look at the device datasheet, TM4C129 System Designs Guideline and LaunchPad schematic they all have a ESD protection array between the transformer and RJ45. Do you have the that?

How did you verify the pulses?  Oscilloscope.  My problem is the pulses are getting degraded in amplitude.  Do you have any pictures of your capture.

Hi CW,

Foot note to Charles great suggestions, perhaps you can add 3v3 TVS diodes in place of 4x-22pf on TX/RX differential pairs. OnSemi makes 1ns reactive 3v3 TVS with solder tabs, little bigger than 0402 cap. I really like these 18KV ESD devices!

/cfs-file/__key/communityserver-discussions-components-files/908/ESD-TVS-3.3v-5Bvd-16kv-SMT_2D00_SOD23-OnSemi-ESD9C3.3S_2D00_D.pdf

hi,

Clock Work said:

How did you verify the pulses? 

             Yes i am using tektronics scope to verify pulses.

Clock Work said:

Do you have any pictures of your capture.

          As a company policy i can not upload photos of degraded signal right now. But i will try to find a way to upload photos.

Clock Work said:

My problem is the pulses are getting degraded in amplitude.

           After studying  few documents i found out that Rbais resistor affects the voltage levels of differential Tx lines. So it may be one of the cause for degraded pulses. Please verify

regards,

digvijay

Digvay,

I used an oscilloscope to verify.  The boards were all good at one time but then the signals got degraded.  We are fairly certain that it is ESD as we have taken many boards that were working and were able to damage them by numerous power cycles. I am using 4.87K for rbias.

Try to measure resistance from 3.3V ethernet termination plane to ground.  Our failing boards measure around 200 ohms where a good board is 7000 ohms.  

Failing Board Picture:

Good Board Picture:

Respected sir,

                     I am also getting resistance of 230 ohm between VCC and GND.  I was able to establish link (malfunctioning) by changing Rbais. But i am facing another problem. The voltage on TX-P and TX-N line is not symmetric. On TX-N line it is giving 1.08 V which is right. But on TX-P line it is giving voltage of 720 mV. They should be same. When i opened tracks by removing component on track and measured resistance with respect to ground it was giving 1K for TX-P and 42K  for TX-N. Can you please verify this with faulty board.  I am planning on populating second card and then testing this issue.

Regards,

Digvijay

Secondly the 20pf caps TX/RX pairs may be suspect in allowing outside sources of EMI from ground to enter the pins. Toshiba makes a tiny bidirectional TVS 20KV (DF2B5M4SL) pads on bottom/side wall. Forget about soldering them by iron, solder paste & hot air perhaps may work!