Tool/software:
I am picking up on an issue with the TLK110PT phy chip that seems to align well with the following problem report: https://e2e.ti.com/support/interface-group/interface/f/interface-forum/647280/tlk110-half-of-the-devices-unresponsive-above-70-degrees-c?tisearch=e2e-sitesearch&keymatch=TLK110#. What was the resolution for this customer?
Failure mode;
-
When the unit reaches thermal equilibrium with the oven ambient temperature above approx 50 deg C, it is no longer able to auto-negotiate a physical link with a remote partner outside the oven (also with a TLK110 phy).
-
If the unit is powered and allowed to negotiate the link at a lower temperature, it remains connected up to (and beyond) the board maximum ambient temperature of 70 deg C.
-
Similarly, if the unit starts up in the failed mode at an ambient above 50 deg C, it does not renegotiate when the temperature is reduced to normal room temperature.
-
Behaviour when physically unplugging the link is indistinguishable from a power up (failure assumed to occur in part of the early negotiation).
-
When the unit is unpowered but connected, the partner link light flashes slowly (0.5Hz).
Configuration;
-
We are bringing up the unit in hardware bootstrap mode, therefore we believe there to be no reliance on software/initialisation to establish the physical link.
-
Both the unit under test and the link partner are expected to auto-negotiate the link parameters.
What we have tried;
-
We have tried, and eliminated, the recommended advice in the above forum post of pulling JTAG_TDI high via a 1.5K pull-up. Are TI aware of a feature, documented or otherwise, whereby the part may get stuck in JTAG mode above a certain temperature?
-
Additionally, the following problem report was raised by a departed colleague on a related product: https://e2e.ti.com/support/interface-group/interface/f/interface-forum/840325/tlk110-power-on-issue-with-tlk-110
-
Bob Grice's reports suggest a working-fail-working 'window' whereby MDIO comms also fails, my current problem is more of a working-fail 'threshold' where MDIO continues to respond. I do not believe it to be the same problem at the present time.
-
Nonetheless, following the advice in Bob Grice's problem report we have reverified the following;
-
We have checked the input clock, which is generated by a crystal osc and forwarded via a clock buffer, both are within specification. We are able to see RX_CLK, rate appears to be correct.
-
We have increased the ramp-rate for the 3V3 rail to 47us.
-
We are still unable to ensure the clock circuits are powered and stable before applying power to the TLK110.
-
None of the above has resolved the issue at temperature.
-
Other points to note;
-
When the unit fails to negotiate the link after starting up above 50 deg C, the MDIO interface remains functional. A software reset does not cause the link to be re-established.
-
If a valid link partner is replaced by an external loopback plug (TX -> RX), a previously failing unit is able to establish a link with itself regardless of temperature. Are TI aware of any issues with internal clocks moving out of tolerance at temperature?
-
Failing units have been identified from the following suppliers with the following TLK110 date codes (all suppliers have provided a C-of-C);
-
Mouser (TI date-code 1927, Texas? [Perhaps generic country-of-origin] )
-
Rebound (TI date-code 1849, Thailand)
-
Arrow (TI date-code 2228, Taiwan)
-
-
Are TI aware of any thermal-related issues with TLK110 parts from the above date-codes/fab plants?
