TLV6710: TLV6710 operating at >1.7V at input pin and getting latched

Hello Anthony,

Your schematic did not attach. You must use the "insert image/media" button (little box with a plus next to the paper clip) to insert a picture.

If it is a PDF, use the paper clip to attach a file.

Hello Anthony,

From what it looks like, if the diode is conducting, you should have  around 5V on the input. Is there something else on the SW_BAT_OFF_N line that could be clamping?

If not, then the input may be clamping for some reason. Abs Max is 6V - so it should be able to take it (though it may not be functional, but should not latch). Be sure to look at all the pins for a transient (inputs, outputs, supply). Transients going below ground can cause a latch-up.

Without knowing the source and size of the transient, it is hard to tell what is causing it. Is it drawing high supply currents? If it is a true latch, it should be drawing large supply currents (mA's).

I would have you try try the sister 18V part, the TLV6700? It has a wider 6.5V input range vs. the 1.7V. It can drop-in replace the TLV6700. But the 24V pull-up knocks it out...

It does not draw a lot of current when the pin6 gets stuck high.  The part does not get damaged and will work fine once VDD is removed.  I disconnected all the loads on the SW_BAT_OFF_N signal and it still gets latched high after a transient.  

The 5V zener diode clamps at 3.5V because it is biased by a 10Mohm resistor.

I forgot to mention that there is a .01uF cap across the zener.  This cap made the part MUCH MORE STABLE during the transient.  Now it happens much less often.  Last night it happend everytime and then this morning it happend first thing but now I can't get it to latch.  I was going to probe all the pins and send them to you to see if there was something weird on the VDD or the input but I couldn't get it to fail anymore.

The fact the the .01uF made a BIG BIG difference in the latching seems to indicate that the input pin is the issue that causes the latch.  I was hoping that there wiould be something obvious to you about the 10M or the zener that made the part latch high.  Then as I was reading on this forum about the 1.7V requirement on the input I got more nervous about having a 5V zener on the input.  I was afraid that the part just doesn't work as a comparator anymore if you exceed 1.7V.  It might be nice if you asked the designer this question.  What is the issue if there is 3.5V on the input?  Does the part become unstable and could it latch? 

Another thing I could try is swap the anode and cathode of the 5V zener so that the clamp is 0.7V instead of 3.5V.  This could give us a clue that the part does not like the 3.5V input.

Thank you for your inputs and I'll get back to you ASAP.  Please keep this ticket open.

BTW,

This part is very cool because it has a wide voltage range, ultra low power consumption, a 400mV reference and two comparators in a small package.  It is very useful!  Your datasheet focuses on using it only as a window comparator but there are dozens of other possibilites.  I'm not using it with fixed resistors on the input but as a brute voltage comparator.  My point is that the data sheet is so constraining that I wonder if the IC designer cut corners with the part to keep it only as a window comparator with fixed resistors on the input.

Hello Anthony,

As you said, you could reverse the Zener, or replace it with a "regular" diode (or two) - anything to keep the voltage below 1.7V. You just need to get over 400mV.

Adding the capacitor will soften any transients if they are coming in off the Vbat_reg. Are all the resistors near the input pin? Are there long traces from the input pins to the resistors? A long trace can act as an antenna if it goes by any nasty EMI points, such as coils, relays or transformers.

If Vbatt is only going as high as 12V, you could just make it a voltage divider, replacing the diode with a resistor. Say 1Meg that would give you 1V. And maybe a cap across the resistor. Or also parallel the diode to act as a back-up clamp.

BTW, I do not know your application, but "it happened the first time" usually points to some part of the circuit that needs to charge-up or draws high currents upon first power up after a long period., such as a power supply with large bypass caps. Instead of power-cycling few seconds...try spacing it out by a few minutes or hours.

Hello Anthony,

My guess is the "constraint" is because the original designers had to make a compromise between maximum supply voltage, input offset and maximum input range. Adding additional input range would have added more complexity, compromised accuracy and supply current...and added die size for the small package.

With a 400mV trip point, they expected that the majority of applications would be after a voltage divider centered around 400mV. 1.7V gives a 4.25x overload range.

While the initial application was as a window comparator, it can be used as two separate conventional comparators sharing a reference, as you have done. We may make that more clear on the next datasheet revision..

We do have newer devices with internal references with full input range, such as the TLV4011, TLV4021 family and TLV4062 family - but they are 5V devices. Your 12V supply and 24V pull-up preclude the use of those...unless you use an external MOSFET and <=5V supply.

Paul,

Thank you very much for your correspondence.  What I think would be useful for guys like me, that use this for other than a resistor divider reset circuit, is that you give us a little more information about this IC's performance outside of the 1.7V max. input range.  I understand this may be a small percentage of the IC's volume usage and therefore not worth your time, effort or datasheet pages.  

But in the meantime if you get a chance to have lunch with the guys at Forest Lane in Dallas who may have experienced latch up of the output to high impedance then please let me know.

In the meantime I will keep you posted on my end with further testing.  I appreciate your inputs very much.

Anthony

Thank you very much.

Here are two more thoughts about the latching:

1) Apparently the IC has an internal power on reset and UVLO circuit:

I think the designer may be able to look at these two circuits and see if it is possible that one of them is getting latched.  When it was latched I left it overnight and in the morning it was still latched.

2) The latching happens very frequently without a capacitor on pin 4.  This is a big clue.  Now that I can't seem to get it to latch (where it was doing it regularly at one point) I could remove the .01uF cap and probe the VDD and pins 3 and 4 and 1 and 6 to give you more information.  I remember probing pin 4 during latching and there was some high frequency (10-100MHz) stuff on it prior and during the latching that went below 400mV but would it latch?  I promise to get you more plots so we can get to the bottom of this.

I also checked the layout and it looks good with the diode, resistors and bypass caps near the IC, however the trace to R37 is long and the connector J9 has a switch and long wire assembly connected to it.  In general I think that since this is a rough environment with possible 60-100 current surges, then this part is not the right part because it is prone to latching.

I will drop in the lower voltage part as you recommended and see what happens.

I am violating your datasheet so this is all my doing.

Best,  Anthony

Hello Anthony,

Perhaps I was not clear. The lower voltage device (TLV6700) is only good up to 18V, supply and pull-up. It cannot handle the 24V pull-up.

It suspect you are getting coupled-in surges from the long trace.

This is a micropower device. It has no defenses against 200MHz - and can turn into a crystal radio. If this is true, you could add 10-50pF to ground on the output to shunt any RF to ground. If there is some nasty EMI floating around, also try adding ~100pF to the inputs to GND to shunt any EMI.

Is it possible to splice in R37 nearer to the comparator as an experiment? I assume R37 is near the connector to the "outside world". Moving the resistor near the comparator allows it to absorb the transient.

If the switch trace is long, and there is a chance of the trace radiating to other parts of the board, then R37 could be "split"...50k near the connector and 50k at the comparator. That would minimize the transients from both the external and on-board.

But the main test would be to lower the 3.5V. Turn the Zener diode around and see if things get better...