• TI Thinks Resolved

SN74CB3T3125: Input pull-down resistance

Prodigy 40 points

Replies: 8

Views: 71

Part Number: SN74CB3T3125

We have encountered a couple of instances where we suspect the A port internal protection circuitry degrades over time when connected to an open-drain signal pulled up to 5V through a 5 k-ohm resistor.

We are using the CB3T3125 as a 5V to 3.3V translator.

  • Vcc = 3.3V. 
  • All 4 OE inputs are tied low (to GND). 
  • Since we only using 2 of the 4 portions (ports 1 and 2), ports 3A and 4A are also tied low;  Ports 3B and 4B are floating.

After a period of extended "on time,"  ports 1A and 2A develop a resistance to ground.  In failure scenarios, the pull-down resistance is 1.0 to 1.5 k-ohm, which, when combined with 5 k-ohm pullup, causes the signal level to not rise above 0.9 to 1.1 V.   This is interpreted as a persistent logic-low signal.  A "healthy" device has a measured pull-down resistance of approx. 335 k-ohm.  We have also measured 15 k-ohm to GND on a part, yielding a signal level of 3.7 V.  This doesn't cause a level detection failure, but indicates that the device has been degraded.

Any guidance on possible causes or mitigation techniques would be appreciated.

  • 335 kΩ does not sound correct (1 GΩ would be more plausible). Was this measured in circuit?

    The most likely cause of this kind of damage would be overvoltage, but overshoots are unlikely with open-drain signals. Are there undershoots? Is there any noise? Might there be ESD? Do you have oscilloscope shots?

  • Hi Kirk,

    The 335k Ohm as Clemens mentioned seems very low to what a normal unit should be. The system specs that you described shouldn't cause the issue you are seeing but as Clemens mentioned overvotlage/undervoltages that occur on the line such as transients and so forth. Also I am also curious how the device impedance was measure - were you just doing a voltage divider equation as stated before? 

    Please let me know so we can see what is possibly causing this issue.

    Best,

    Parker Dodson

  • In reply to Clemens Ladisch:

    For a "good" device: The 335 k-ohm was measured in-circuit with a Fluke digital multimeter set to measure resistance, with the red lead on GND and the black lead on the I/O port pin.  With the black lead on GND and red lead on I/O port pin, the resistance doesn't register, so it is above 1 G-ohm.  

    On a compromised device, the 1.0 to 1.5 k-ohm resistance between I/O port pin and GND is bidirectional.  On a port pin which is degraded but not causing a failure, I measure approx. 15 k-ohm with red lead to GND and black lead to I/O port pin, about 230 k-ohm with the meter leads reversed.

    The input signal is switched by the transistor side of an opto-coupler which is used for ground fault detection.  This circuit only goes active when a ground fault is detected, which isn't often.  The circuit is within an enclosure, not subject to ESD events due to touching.  It may be that there is some sort of inductive coupling onto these signals, though there aren't any high power signals nearby on PCB layers.  I don't have any 'scope shots -- I suspect the cause isn't going to be able to be replicated on a benchtop.

    Thanks for taking a look at this puzzler.   

  • In reply to Kirk Treubert1:

    Measurements done in-circuit might be affected by other components.

    When applying a positive voltage to GND, you are measuring the current through the clamping diode between GND and the I/O pin.

    An optocoupler output indeed should be harmless. I can only suspect that there is noise come from some other source (through the output signal, or the power rails).

    (Not related to this problem: for unidirectional downtranslation, you could simply use a buffer with overvoltage-tolerant inputs like the SN74LVC2G17, or 07 if you still need an open-drain output.)

  • In reply to Clemens Ladisch:

    We have SN74LVC125APWR in our library, which appears to be a drop-in replacement onto the PCB footprint for this unidirectional application. 

    LVC125A also has 5V-tolerant inputs when operating from 3.3V supply rail.

    Is there any difference in the input protection, or susceptibility to degradation resulting from over- or under-voltage conditions, for the LVC125A vs. CB3T3125?

    I'm looking for potential fixes which don't require changing the PCB artwork....but any recommendations to improve the operability of the installed device in this application are appreciated.

    Thanks.  Kirk Treubert

  • In reply to Kirk Treubert1:

    Hi Kirk,

    The  LVC125APWR has a similar protection scheme to the CB3T3125 however - the LVC125A has a lower absolute max voltage rating of 6.5V compared to the CB3T device which is around 7V. But the damage seems to be the port to ground connection. Both of the parts are going to behave similarly when the device is undervolted - the I/O clamp current can only handle a magnitude of 50mA through it before it is susceptible to damage - however running it over recommend operating will also degrade it. So if there are any transients with components less than 0V those clamp diodes will begin to conduct. A way to help prevent damage is to add a series resistor to limit current coming out of the I/O clamps. I don't know if switching devices will necessarily help the application because it sounds like there could be a transient issue that causes damage to the protection cell of the ports - which the typical solution is to add a current limiting resistor if possible, but the sizing would depend on how much insertion loss the system can handle.

    That being said,

    How many device failures have you had, and out of how many total? How long are the boards being ran before you start seeing the failures? Are the boards susceptible to undervotlage events? 

    Please let me know !

    Best,

    Parker Dodson

  • In reply to Parker Dodson:

    Thank you for your feedback on the port protection.  So far we've had two similar failures on devices which have been running in fielded systems for several months before developing a "stuck fault" symptom due to the described pull-down anomaly.  Perhaps 100-150 systems have been deployed, with varying amounts of run time.

    The signals from the 3.3V side are monitored by a processor system.  These have been subject to "false positives" due to effects of amplifier switching "noise," but these events are digitally filtered by FPGA process and DSP software.  It is possible that the input ports to the CB3T3125 devices are seeing some negative voltage transients induced by amplifier switching events which aren't all common-mode.

    It seems that inserting some resistance in-line with the port A inputs is warranted whether we use a switch device or an actively-driven device.  Does it also make sense to place some resistance in series with the VCC pin?

  • In reply to Kirk Treubert1:

    Hi Kirk,

    That makes sense - thanks for sharing some more details. I think some of the negative voltage transients could be pulling current through the protection cell eventually causing it to wear out - using the series resistor is probably the best idea.

    I don't think the resistor at VCC is necessary the current is most likely coming from the ground and as long as there is resistance on the ports it should limit the current coming through the port through the clamp and extending its life.

    If you have any other questions please let me know!

    Best,

    Parker Dodson