LMH6629: Protective Diodes Capacity: will they discharge large input capacitors?

Hello Hariton,
The absolute maximum rating is for 10mA maximum input current. We do not run any tests on how much electro-migration will occur based on time AND current. We only do the test for a continuous current which in this case is 10mA.

The BAT68 looks like a good diode. Being a Schottky, it will start protecting the amplifier faster than the internal diodes on the LMH6629.

Is this capacitor on the amplifiers non-inverting input? You will need a resistor to GND on the non-inverting input since the amplifier needs a low impedance path for its input bias current to flow.

-Samir

Hi Samir,

1. Yes, we are taking care of bias currents. Its the inverting input, so the Rf is the one that flows it. Unfortunately, it does not discharge our capacitor because load becomes high impedance at power down. A similar circuit is on Figure 59.

2. As you suggested, we will try to stay in the 10mA range that will flow from the input to the ground. If LMH6629 would consume 10mA or less when +1.9V applied to its V+, then we would not need an external diode (1.9V=2.5V-0.6V forward voltage of internal diode). Spec only shows consumption current above 2.7V, figure 44.

But for transition analysis we would need similar consumption data around 1.9V. Do you have such info please?

Hariton,
Can you please send me a schematic of your circuit?

-Samir

1. Here is our simplified schematic. It is identical to the one on page 27, figure 59 of the spec. R3 is effective load, which is the positive input of a second OPA.

2. We run the test. LMH6629 died after about 30 power cycles. It stopped responding to the inputs and the output became +4.4V.

Now we are going to add an external protection diode HSMS-2820-TR1G between anode(+) of C1 and VS. But it is not going to satisfy the Absolute Maximum Ratings (page 4), because it says "Analog Input Voltage −0.5 to VS". The diode is going to keep it at about VS+0.5V each time the power goes down.

3. I am curious, which one is incorrect: the schematic 59 on page 27, or the Absolute Maximums? Each time the Schematic 59 is powered down, the Absolute Maximums are going to be violated for a short time, even if we will add the protective diode, and with any value for decoupling capacitor C1.

Hariton,
1. The circuit is configured in a gain of 6. If your COMP = HI, the circuit is going to be unstable...minimum stable gain is around 10V/V in a non-inverting gain configuration. If COMP = LO then you are okay. Also, the feedback resistor is quite a bit smaller than typically used. If the feedback resistance is bumped up from 50ohm to 240ohm and the gain is maintained, that puts the input gain resistor at 48 ohms. This resistor will help to further limit the current from the capacitor as it discharges.

2. One thing you can do is to add another resistor between the junction of R1 and R2 and the actual input of the amplifier. This will also help to provide further current limiting.

3. The ABSOLUTE MAXIMUM ratings give the limits to which the part can be stressed. Figure 59 doesnt make any assumptions about powering down the circuit. The designer needs to ensure safe discharge of the circuit prior to turning power off. Current limiting resistors and sufficient clamping would have to be added externally if any risk is expected.

-Samir

Hi Samir,

>1. I thought the default is low (spec page 20 , paragraph 7.3.1). But anyway, its a good point: to make it sure and ground it just in case.

>2. Very good idea, thank you. But it will increase thermal noise: 10mA at 2V require 200 Ohms, which is a significant source of the noise.

>3. Capacitor at inputs are common, I don't think many people will take special measures to discharge them each time. It would be so much easier for us if Absolute Maximums would allow VS+0.5 for inputs, so we can use clamping diodes as a simple measure, or for low currents the internal ones (if they would allow about 20 mA peak currents). The spec says -0.5 to VS sharp, which is almost unrealistic to achieve.

4. It would be nice, if the spec will state what is the maximum capacitance value the inputs can safely discharge at the power down, so we don't have to guess or forget about this problem at all.

Hi Hariton,
You should be able to go around 700mV beyond the power supply rails. I am not sure why the datasheet was done this way for the LMH6629. I apologize, I did not look at the datasheet closely before, I just assumed it had a diode drop limit incorporated.

With regards to the discharge limit, it is dependent on the peak current and the rate of discharge. Its not just a simple test of single event stress, but you would have to do a life test to see what happens when you constantly repeat this at a certain rate and what happens as the device ages. Since there are multiple variables present, I believe we leave this to the user to determine the particular condition and protect against it. Because of these variables we stick with continuous current.

The noise contribution from a 200ohm resistor is around 1.7nV/rtHz. Is this the noise you are concerned about?

-Samir

Hi Samir,

1. Regarding discharge limit: The only external parameter is the input capacitor value, assuming power supply is isolated from the rest of the board, which is easy to achieve. Peak current and duration are defined by the LMH6629 internal properties, thats why I was asking about its current consumption at 1.9 volts and max allowed peak input currents. It would be just easier for us, users, to have such capacitance value in mind from the spec when we are designing protection mechanisms. It does not have to be 100% accurate, just an estimate. Right now we have no idea about its range.

If max allowed peak input current is 10ma and current consumption at 1.9V is greater than that (which most likely the case), than it means any schematic that has ANY capacitor connected to the inputs should have an external clamping diode, which renders the internal diodes useless, which I don't think was your intention.

2. Rg in the negative input loop should not be counted for protection, because as you said large values around 200 Ohms will generate 1.7nV/sqrt, which will cancel the main benefit (low noise) of LMH6629.