LM139A

Ed,

-10mA would not damage the LM139A. However, internal parasitic transistors will turn on and the output may be incorrect. This effect is called output inversion.

Hi Ed,

I second Ron's summary.

The LM139 does not have ESD protection structures on the inputs, so the reverse voltage flows through (random) uncontrolled reversed junctions of the die substrate. As such, there is no guarantee on what will conduct and when.

Devices with ESD clamp diodes has a "controlled" path that we can specify a maximum current. Since the LM139 has a "leaky floor" action when reversed, there is no one device we can characterize to give a specific max current.

If you are expecting reverse voltage on the input, even occasionally, then an external Schottky clamp diode should be used between the input to V- to provide a "controlled" path for the current (and can be safely limited by a series resistance).

Regards,

Ron and Paul,

Thanks to both for the reply. 

Paul-  I do not completely understand your last sentence.  I can see how an external Schottky clamp diode between the input and V- can provide a controlled path but where would the current limiting resistor be placed?  I am guessing between the input pin and the cathode of the Schottky?  And would we size the resistor to limit the current to 10 mA or would it be sized to protect the Schottky?

We have been having a significant number of failures of this part and are having a hard time coming up with the root cause.  The evidence is beginning to point to the inputs being pulled below the negative supply, but we cannot yet say for sure.  We had a few failed parts sent to an external test house for evaluation.  Would someone at TI be willing to take a look at the report and see if they concur with the test house or see if they can derive more information from the report than the test house was able to do?

Hi Ed,

The diode would be right across the input and the V-/GND pin. The resistor would be in series with the input/diode node (before the diode). That way the resistor "absorbs" the extra current as the diode clamps to GND/V-.

Knowing the maximum current and the maximum available voltage, a simple Ohms law calculation would give you the minimum series resistor size.

So if the maximum voltage available was 12V, and the max current was 10mA, then 12V/10mA = 1.2K minimum resistor size. The bigger the better - but there are performance tradeoffs.

5K to 10K is common. But beware that you may need to add a compensating resistor to the other input to cancel out bias current effects if you can't tolerate several mV offsets.

Pulling the inputs below ground is "A Bad Thing". It will eventually degrade the performance, to the point of failure since you are essentially chipping away at the foundation of the device (substrate). External clamp diodes take the brunt of the current. Be sure they are Shottky types, which have lower forward voltage than the substrate diodes.

If fast edges are involved (faster than the diodes can react, or less than <100ns), then add a 10 to 33pF ceramic SMT cap in parallel with the diode to short-out the initial fast edges. Both the cap and diode should be placed as close to the pins as possible for the shortest path.

Regards,

0028.LM139.docx

Hi Paul,

Thanks for the explanation.  Just to be sure I attached a schematic with the resistor and diode inserted as I think you are describing.  The LM139 V+ supply is connected to +15 and the LM139 V- supply is connected to GROUND.  The circuit detects OPEN and GROUND on the MASTER_ON_OFF input, but it is pulled to ground remotely from LM139.

The schematic also shows a capacitor that we have on the input pin.  Is there any way that the capacitor could be the cause of our problem?  At power down it is conceivable that when MASTER_ON_OFF is OPEN that the LM139 -input could be above the V+ pin of the IC.  But the data sheet implies that the input voltage can be 36 Volts max independent of the supply voltage, so this should not be a problem?

Thanks again.

Ed

Hi Ed,

The diode should be on the right side of the resistor, connected directly to pin 4.

One issue I see is that the common mode is violated when the cap fully charges up. The common mode limit of the 139 is 1.5V below V+, or 13.5V in your case. The resistor will pull-up to 15V when the cap is fully charged.

However, this is not a big problem since as long as at least ONE of the inputs is within the common mode range, the output should be valid. This is noted in note 5 on page 6. Just beware that you are violating the common mode spec and you have been warned! :) You could add a 6.8K resistor across the cap to create a voltage divider to limit the upper voltage to 13V (or 10K and 68K to limit power dissipation)

The inputs can go up to 30V, regardless of supply voltage, so that is acceptable - but applying input voltage to an un-powered devices is never a good idea. The danger is that a capacitor can provide a large "dump" current (up to several amps) should something break down. But adding the series resistor will limit that current - so there is another reason for the resistor.

So adding the series resistor should ease your issues. 15V/1mA = 15K minimum. Since your application is not really "precision" (a few mV change does not make much difference), I would use a 50K to keep the currents as low as possible.

Regards,

I came to the forum with this exact question—can the LM139 handle a voltage lower than the negative supply rail if it is current limited—and was happy to find a discussion already existed. However, the answer is a little unclear from the responses above. Ronald’s response makes it sound like no damage occurs with -10mA or less so it’s ok as long as an incorrect output can be tolerated. But Paul’s recommendation to add a clamping diode implies that the device could be damaged regardless of whether or not the current is limited.
(To add to my confusion I looked at Ed’s schematic and I don’t see where the negative input voltage could be coming from, but maybe that was a separate follow up question or maybe I’m just being slow!)
In my application the input can go below the negative supply rail only during power up (~10 ms max) and the current is limited to “only” 150 microamps max. I want to ensure that this condition doesn’t stress the LM139 with many power ups over time. If it does I’ll go ahead and add the clamping diode, but I’d really prefer not to at this point in the product design unless I have to.
So to reiterate and clarify the question: is the clamping diode always recommended/required to protect the LM139 if the inputs can see a voltage below the negative supply rail, even if the current is limited to 150 microamps?

Zach,

150uA is not a stress problem. At that current voltage is about -600mV. Multiply by current and power is just 90uW.
Even if a parasitic NPN transistor redirected all the current to VCC (at 30V). Power would still only be 4.5mW.

Ronald,Thanks so much for the information and for the timely response!Thanks,
Zach