# Open Collector Comparator for Voltage Glitch Detection

I recently ran a round of PCBs which included an experimental circuit to detect short glitches in my input voltages. I attached the schematic for reference:

The intent is that C29, in parallel with R27, should hold V- at ~3.1 V even if 3V3 has a short glitch. So a quick glitch on 3V3 should be detected by the comparator and trigger an output which I can read with additional circuitry.

After testing a few PCBs and seeing initial input current spikes and some minor smoking from the board, I've narrowed it down to this chip as a possible culprit. I started taking voltage measurements and saw big disparities in V-: 0.96, 2.82, 1.58, 1.32, 3.96?? Seems like this chip is just completely fried, but I'm not sure why this would have happened.

Does anything stick out as an obvious mistake? Some thoughts/notes:

1. I noticed the same input current jump when I applied only 3V3 to the board, without any power on 5V.

2. Without desoldering the resistors from the board, I can't accurately measure the individual resistor values because of the rest of the board.

Edit 2016-06-03:

Think I got to the bottom of it. I think the chips were actually still operational, and the smoking/initial current surge was coming from somewhere else (to be determined where). The actual problem had to do with my biasing circuitry for V-, and the comparators actual bias voltages.

The datasheet gives the bias current anywhere from 25 - 400 nA.

My "ideal" biasing current would be ~ 310 nA (3.3 V / 1.0634 MΩ).

Considering the wide range of possible comparator bias currents, this could definitely explain the difference in V- readings I was seeing. For my next round of PCBs, I'll spec the ideal biasing current to be in the 10-100 uA range so the comparator bias current becomes insignificant.

Does this make sense?

• "minor smoking" – Spehro Pefhany Jun 3 '16 at 18:52
• Yeah, lol, noticed a brief small stream of smoke from the board when I first plugged in but it stopped quickly after. – Jim Jun 3 '16 at 18:58

Well you could easily have fried the part when applying 3V3 with no power on the comparator. This is because input protection diodes would route the 3V3 onto the unpowered 5V bus and likely exceed the diode ratings and cause possibly catastrophic device failure.

It's probably a good idea to put 100kohm in series with the +Vin pin.

• That's what I was suspecting, but the datasheet implies its not a problem The input common-mode voltage of either input signal voltage should not be allowed to go negative by more than 0.3 V. The upper end of the common-mode voltage range is (VCC +/– 1.5 V, but either or both inputs can go to 30 V without damage. 100k in series with both inputs for better safety? – Sean Houlihane May 24 '16 at 17:22
• Yeah I've just read that so I'm thinking about it..... The problem is that the DS does not explicity say you can have inputs powered when the device is unpowered and this gives me cause for concern. Try 100 k in series with both inputs is advisable now I think. – Andy aka May 24 '16 at 17:27

I noticed the same input current jump when I applied only 3V3 to the board, without any power on 5V.

This might be the problem.

The data sheet is a little confusing but I suggest you pay attention to this part.

The other thing that's more important is the pinout.

Figure 2. Three different packages for the TS391.

Your schematic shows the TS391IYLT. Make sure you haven't installed the RILT version.

• I tried a fresh board to see if this was the case, and applied 5V to power the entire board; I have a regulator providing 3V3 from 5V. But, upon connecting my power supply I saw an initial current jump to ~2+ amps (where steady state is about 3-400 mA). – Jim May 24 '16 at 19:25
• Also, on one of my boards I had previously (probably) fried, I de-soldered the comparator and saw the voltage at the pad for V- go to the expected 3.1 V. So, it does seem like the chip is giving me an issue. – Jim May 24 '16 at 19:29
• See update regarding package. – Transistor May 24 '16 at 20:35
• Good point @transistor ... I looked at the markings on my package and it does seem like the manufacturer used the right one (marking K510, IYLT, same pinout as ILT). However, I have had issues with my manufacturer using counterfeit or "gray market" parts in the past. I checked an old rev of my board with a nearly identical schematic (except it had a 2 pin jumper as an enable on the ouput of the comparator to feed an AND gate) and the voltages were fine. I'll keep digging. – Jim May 24 '16 at 21:30