TLV3701-EP: Burns after several hours

sedat unalacak

Part Number: TLV3701-EP
Other Parts Discussed in Thread: TLV3701-Q1, OPA171-Q1, , TLV3701, TLV3702, TLV2464, SN74LVC1G66-Q1, TS5A3159

I use TLV3701-EP to convert open-gnd connection to 5V logic level. I have 4 of these comparator circuits on the card, and after several hours (3-6 hours) one of the comparators burn, and then circuit continuous to work well once I replace the burnt TLV3701-EP with another TLV3701-EP, until another one burns. I try to figure out why. Supply is single 5V, inputs never exceed 5V. So overvoltage should not be an issue. All two input pins and output pin are connected to rest of the circuitry with series resistances in the order of tens of kohms in between, so I don't think overcurrent is an issue. I measured the 5V supply and it is clean.I thought powering sequence might be the problem, because one of the inputs might being pulled up to around 4.5V at power up before 5V supply rises, but datasheet claims that VCC+5 volt is acceptable at inputs. Also this pull-up operation is done with a 560kohm resistance, so TLV3701-EP should be able to clamp overvoltage even if it occurs, because input current limit is 10mA. So I don't think powering sequence is an issue. Then I thought maybe ESD is the problem, because there is not any information in datasheet of TLV3701-EP. But then I realize TLV3701-Q1 datasheet includes information about ESD protection. I assume TLV3701-EP and TLV3701-Q1 have the same structure and TLV3701-EP datasheet does not include ESD information because it is old (please correct me if I am wrong). So I dont think that ESD is the issue either.Then I replaced four TLV3701-EP's on the board with OPA171-Q1 and tested, circuit worked without any fault for 24 hours (I know that using an opamp as comparator is not a good practise). I am trying to find why TLV3701-EP comparators which are initially fine burn after several hours, any help is appreciated.

over 5 years ago

0 kai klaas69 over 5 years ago

Guru 140200 points

Hi Sedat,

can you show us a schematic?

Kai

0 sedat unalacak over 5 years ago in reply to kai klaas69

Prodigy 80 points

It is not included here, but output signal is connected to two nodes: one is input of a logic gate powered with 5V and there is a 10kohm resistor in series between input of logic gate and output of comparator. Second connection is to a voltage divider through a 162kohm resistor. Also 28V supply is kept clean with a TVS diode.

0 kai klaas69 over 5 years ago in reply to sedat unalacak

Guru 140200 points

Hi,

what does "open-GND" mean? What is its potential? Where does it come from? Could you clamp this node against ESD by the help of a TVS?

Kai

0 sedat unalacak over 5 years ago in reply to kai klaas69

Prodigy 80 points

This is a circuit to convert gnd/open signal to 0-5V digital logic. Gnd/open pin is either pulled down to ground or left open (high impedance) by the device we are interfacing. So its potential is either zero or not connected and irrelevant. We use Vishay's this TVS (www.digikey.com/.../en. We work in a ESD protected environment, and there are tens of IC's on the same pcb but we only have problem with TLV3701-EP, so I dont think that it is ESD. Btw is there any any difference in terms of self ESD protection IC between TLV3701-EP and TLV3701-Q1? Becase datasheet of TLV3701-EP does not tell anything about ESD but datasheet of TLV3701-Q1 does.

0 Paul Grohe over 5 years ago in reply to sedat unalacak

TI__Guru 53846 points

Hello Sedat,

What do you mean by "burnt"? Is it heating up/smoking/"poped"? Or just becomes non-functional?

The EP and Q1 versions use slightly different die. The "EP" is handled by a different group, so they will need to speak to the differences (I'm guessing they are very slight - but that can change ESD perfromance).

Since all the I/O pins have some series resistance, the only pin with possible damaging currents is the supply pin.

Are you *sure* there are no glitches on the supply? Particularly at turn-on and turn-off? Even several seconds after power-off?

Watch the supply turn-on and turn-off with the scope on "roll" - digital scopes rely too much on the trigger, and may miss a transient (especially slow ones).

Set the scope to 500ms/div and auto trigger (or whatever causes the screen to "roll"), and set the acquisition to "peak". Turn-on and turn off the power and watch for glitches. Change the on-off times or "bump" the supplies.

I would also look at the output, too, especially if it goes across the board or through any connectors.

On the "dead" units, what are the resistances of the pins to the GND pin? Any below 1k or shorted? Particularly the output and V+ pin.

0 sedat unalacak over 5 years ago in reply to Paul Grohe

Prodigy 80 points

There is no smoke, it becomes dysfunctional. I dont have the burnt material and pcb now because I had changed comparators with opamps and moved on, but I measured resistances. After the "incident" actually comparator continuos to work, but with great hysteresis around 25V as if there is a very little hysteresis resistance. This circuit is designed to turn on and off around 15V and 4V input, what after it is damaged it turns on and off around 29V and 3V input, bur still works. I measured the resistance between IN+ pin and output (i.e hysteresis resistance) but resistance was the same as non-damaged comparator circuits. Actually I measured all possible resistances between 5 pins in both damaged and undamaged circuits, but there was not a difference. Some pin pairs did not have a proper stable resistance due to diodes, but even they had similar readings on ohmmeter. I took all these measurements in absence of power of course, so I guess somehow the resistance between IN+ and out pins decreases after power up.

28V supply is protected by TVS diode and we use EMI filtering at supply for MIL-STD EMI tests, and we attain 5V with a LDO so it is very unlikely to have glitches at supplies. I checked 28V and 5V carefully but I did not chech them after power off, I will make the tests you suggested in first chance I get.

As I said in previous posts output goes two points on the board and does not go outside the board. One point is the input of a logic gate powered with same 5V, and there is serial 10kohm between them. The other point is, output is divided with a resistive voltage divider and than divided value connected to an input pin of an opamp powered by the same 5V supply. In this voltage divider output of comparator is connected to 162kohm. So like the circuit you see in the schematic above, there is no chance of overcurrent on output side.

Dead units have high resistance to ground. I measured resistances on board (not independently after desoldering it from pcb) and even then resistances were high, close to the resistance of resistors connecting them to ground.

TLV3701-EP is a useful comparator and I want to use it in our other projects as well, so it is important for us to trust it and understand the root cause of this problem. I thought that EP is a more comprehensive quality standard than automotive grade, am I right or they are two irrelevant things? Thanks for your help.

0 kai klaas69 over 5 years ago in reply to sedat unalacak

Guru 140200 points

Hi Sedat,

I would not blindly trust that R35 will protect the TLV3701 against ESD. 8kV ESD according to the human body model can easily arc over a tiny SMD resistor, as 8kV can give a spark of 8mm! I have seen some circuits which have been destroyed by ESD in exactly that way.

Also, an 2kV ESD specification is not enough to fully protect the TLV3701 against ESD. 2kV might be enough for mounting and handling but not for protecting in a rough industrial or automotive environment.

So, in any case I would mount a TVS from the "GND/OPEN" line to "5VDC.GND". If you cleverly mount this TVS at the edge of PCB you can help to shunt the ESD arround the whole TLV3701 circuitry. Do never let ESD enter your PCB!!

Kai

0 Paul Grohe over 5 years ago in reply to sedat unalacak

TI__Guru 53846 points

Hi Sedat,

Hmmm...very strange.

Is it the same position that keeps blowing, say U7, or is it scattered between the four?

Does this happen before or after the EMI tests? On a "fresh" board? Are you doing any burn-ins first?

Can you tell exactly when the fault happens? You mentioned several hours.

What else shares the 5V line? Anything inductive? Relays? Switchers? Do the I/O lines pass by any large EMI emitting devices such as switching power supply transformers or RF TX?

Keep in mind that LDO's cannot sink current, so the LDO output can be pulled above the output voltage with little fight. Since this is a 16V device, it would have to be a biiig transient. And since the TLV3702 has reverse battery protection to -16V, it would also have to be a huuuge negative transient.

If you are seeing shifts in the offset voltage, which would cause the change in thresholds, that would indicate something is happening to the inputs. So if you are hitting it with ESD/EMI, then most likely the pulses are getting around the divider.

As Kai said, small SMT resistors are no match for ESD pulses. Fast edged ESD transients can easily pass through the fractions of a pF across any resistor. Most small SMT resistors will break down at 100-200V. When measuring (or expecting) high voltage, it is common to break the resistor into several series resistors (such as four 160k resistors) to spread out the voltage gradient and also break-up the series capacitance.

The above also goes for conducted/radiated EMI tests. 100's MHz to GHz can easily find it's way past a simple divider when the Xc is lower then the R.

Adding a small (10-50pF) HIGH VOLTAGE ceramic capacitor *AT* the point where the signal enters the board can "take the edge off" by shunting the high frequency energy directly to (safety/chassis) ground.

0 sedat unalacak over 5 years ago in reply to kai klaas69

Prodigy 80 points

Sorry I couldn't understand, what is R35?

To more complete, I share the circuit with protections. As you see I already use a TVS from GND/OPEN to 5VDC.GND. TVS is (https://www.digikey.com/products/en?keywords=SMBJ30A-E3%2F52GIDKR-ND) but I am not sure about its ESD performance. 10 ohm series resistor is pulse proof (https://www.digikey.com/products/en?keywords=541-3248-1-ND) and there is a 470pF at the edge of the board connected to chassis. Am I missing anything? What can I do to improve?

0 sedat unalacak over 5 years ago in reply to Paul Grohe

Prodigy 80 points

It is scattered between four.
We didn't run any EMI test with this particular board. These are fresh boards and their first tests under normal electrical and environmental conditions.
Error time is random. Some endured more than 10 hours some burnt in an hour. Average error time is around 4-6 hours.
There is nothing inductive or relay powered by 5V. 5V is used to supply some tlv2464 opamps, 3 logic gates and these comparators. Also there is one TS5A3159 switch and one SN74LVC1G66-Q1 switch powered by 5V. These two switches, comparators, logic gates and all tlv2464 opamps have their bypass capacitors. There is not a transformer or digital communication on board.
As I mentioned we did not do any EMI tests on the board, but I will keep in mind your suggestions about ESD. I posted the circuit including the protections as a reply to Kai's post. As you can see there there is a TVS diode and capacitor conected to chassis. Capacitance value is bigger than what you suggested, could it be the problem?
Nature of this error scares me. What if I apply these rule of thumbs like using several smd resistors, using smaller capacitance etc. and end up with a circuit that endures 1000 hours instead of 5, which will pass the tests but fail eventually? Is there any way to make sure that ESD will not be a problem?
Thanks a lot

0 sedat unalacak over 5 years ago

Prodigy 80 points

I read www.ti.com/.../slva680.pdf which claims that impedance and inductance from ESD source to chassis (or gnd) should be low. I use a common mode choke (www.digikey.com/.../en for common mode filtering at supply side, but of course it has differential inductance also. So, are connecting TVS to a ground separated with a common choke and using a series 10ohm bad practise? Note that chassis is connected to analog ground that both TVS and TLV3701 use with a resistor and capacitor, so common mode choke is NOT between TVS and chassis. Also, even if this is a bad practise, TVS is connected between gnd/open and ground which TLV3701 use, so it still should be protecting TLV3701, am I right? Or would connecting TVS to chassis be better?

0 Paul Grohe over 5 years ago in reply to sedat unalacak

TI__Guru 53846 points

Hello Sedat,

Is there *anything* that would be causing large transients across the board during that time? Is *anything* turning on or off? (power supplies?). Parts do not just "die" of boredom.

I would do the "peak scope" measurements on each input pin during power-up and down, and several seconds - up to a minute - after power down to watch for last-gasps.

The TVS's you have selected have a max clamping voltage of 38V. If the transient is capacitively passing through the divider, that is well above the 17V Abs Max.

Where are you placing the common mode choke? Common mode choke implies a balanced signal. If you "ground" one side of the choke, you render it ineffective.

The "ground" between the source and circuit should not have a choke in it. With a choke, you then create a high impedance (break) in the ground at high frequencies (or fast edges - like ESD), so now you can have a large AC differential between the grounds. Normally a capacitor is placed across grounds to provide a direct AC path, the opposite of what the choke would do.

I think this may be a layout issue. If you want, I can contact you at your registered email address and you can send layouts directly to me to review.

But we really need to determine what is the trigger and/or where the errant signal is coming from.

Where does the ground for C41, C42 and D8 return? These will have the largest ground currents during a transient and need to go directly to system/power ground. Trace the ground currents from the protection caps/didoes to the main ground in the layout and remember that the voltage along that trace can be raised several volts during peak current.

What happens if you remove the 560k resistor? Do they keep dying? If so, try also shorting R36. Remove R33? Remove R37? If one of these makes a difference, that will give you a clue as to the ingress point. If you remove them all, and they are still dying, then that points to either the output trace or supply being the cause.

0 Kirby Kruckmeyer96848 over 5 years ago in reply to sedat unalacak

TI__Genius 17115 points

Hello Sedat,
In your schematics, you show all your grounds as 5VDC.GND, but you have a 28V supply on IN-. How is the 28V supply grounded? It is on the same plane as the 5V supply?

0 sedat unalacak over 5 years ago in reply to Kirby Kruckmeyer96848

Prodigy 80 points

Hi,

Yes, that 28V pull-up and 5V regular share the same ground plane.

0 kai klaas69 over 5 years ago in reply to sedat unalacak

Guru 140200 points

Hi Sedat,

I would add a TVS add every neuralgic point of the TLV3701. Even at the output of TLV3701 a TVS could sit, isolated by an isolation resistor, though, to take the parasitic capacitance of TVS into account.

Kai

Amplifiers

Amplifiers forum

TLV3701-EP: Burns after several hours