Hi, guys !
I have this schematics... It gives me a high level when my battery voltage drops under a threshold. For this I need a stable voltage reference... I choose LT1790... The problem is that I want to add hysteresis to this comparator. And the only way I know is by connecting a high impedane resistor to the non-inverting imput. For this hysteresis to work, I need to put a voltage divider to the reference, because if i connect directly the IN+ to the reference it won't work. Ok, but now, the opamp si powered from 12V and it will output 12V when signaled, and that voltage will go to the 5V reference output through the 1Mohm and 47K resistors. In simulation the circuit is working, but I don't know if the LT1790 is safe in these circumstances ? Won't a current flow from the 12V to the 5V ? In the simulation there is 54uA at the output of the reference when the comparator is low, and when the comparator turns high, the curent from the reference output drops to 48uA... But is not negative or else...

 

According to the datasheet, the LT1790 can sink several mA without a problem. With a 1Mohm resistor connected to +12V, that is a sink current of only 7uA so it should work OK.

 

Won't a current flow from the 12V to the 5V ?

No.
A small current will flow to the junction of R5 and R6 when U1's output is high, but it's much less than the current through R5 and R6 from the LT1790, so will have only a small effect on the LT1790's 5V output current.

You can calculate the voltages and currents to see the effect.

 

D6 will cause problems and is not needed.
What is R1 ?
What is R3 ?

R-4 (1M) will not provide very much hysteresis.
How much will the Battery-Voltage fluctuate with normal varying Loads ?

The Gate of U-12 should be tied to Ground with a ~1M Resistor to
prevent the Gate from "Floating" above Ground when the Circuit is not Powered.
Replace R-28 with a 1K Resistor to insure faster FET Switching.

R-6 is not needed.
After it is removed, the Ratio between R-4 and R-5 will then determine the Hysteresis.

The LT-1790 requires a 1uf Capacitor on its Input, and also on its Output.
.
.
.

 

Thank you for your good observations.
The resistor values were set just for test a quick test. I will adjust them accordingly... and put the capacitor too.
R1, R3 are just for test.
R4, actually provides to much hysteresis. 1.33Mohm is needed for 1V hysteresis at the battery voltage.
Wow, I wasn't aware that I can dump R6. It seems it's working without it too. I hated that divider, anyway.

And now, let me explain you what I tried to do with that MOSFET.
That MOSFET will pull-down a Raspberry Pi GPIO pin that is pulled up internaly by a 1.8k resistor, signaling that is time to shutdown the device. The way I choose to drive that MOSFET it was made with a very low consumption in mind. And also to limit the Vgs to a safe voltage (about 5V in this case). D6 drops 5.6V from those 11-12V of the opamp output. Indeed, It would have been a good thing to put a 100k resistor in GS, but I thought that the gate is connected either to 5V or to ground anyway, through that opamp... If I had made a classic protection with the zener placed in GS and a resistance from the opamp in the gate, the consumption would have been too high when the opamp was activated. As it is now it consumes only 6uA. I don't need verry fast switching...
So, what do you think... ? Can it be made much safer without pulling to much current from the opamp (lets say no more than 50uA) ?

 

I thought that the gate is connected either to 5V or to ground anyway

The gate is pulled to near ground through the forward-biased Zener diode.
It's better to add a pull-down resistor (e.g. 100kΩ) at the gate.

But you don't need the Zener since that MOSFET can tolerate 20V VGS.

 

Indeed... But once I burned a series of P-chanel MSFETs (BSS84) connecting the source to +12V and the gate to ground. And they also do have +/-20V Vgs. Only when I limited the Vgs to some 5-6V they worked fine. And I was left with a fear of putting 12V on the gate...

 

Only when I limited the Vgs to some 5-6V they worked fine.

Likely you were getting voltage spikes from somewhere that damaged the gates.

But a better way to protect the gate is to use a series gate resistor with a 5V Zener from gate to source.
The maximum value of the resistor is determined by the gate capacitance and how fast you need to switch the MOSFET.

 

The FET Gate will consume zero Power, ( it acts like a very small Capacitor ).
There is no advantage to reducing the Gate-Voltage.

Eliminate R-3, ( I thought it might be some sort of device that needed Power ).

Or,

Just replace this complicated Circuit with a much simpler one ..........
.
.
.

 

Another version using a LM393 comparator. Use LT1905-2.5 for reference and power from RPi 3.3 volt supply