I'm designing a very-low-power, i.e. less than 5 uA device, that requires external inputs. These inputs need to be protected against over-voltage. The allowed range is about -0.1 V to 3.0 V. If necessary, I can live with a top voltage that is limited at a value as low as 2.8 V and as high as 4.0 V, so at least there is some flexibility.
The whole device has a power source at a somewhat higher voltage and a step-down converter to a system voltage at 3.0 V, so I could in general just add a protection diode to that 3.0 V system voltage, accept that the limit is increased to at most 4 V and be done with it. Sadly, I can not limit the current through this diode to the point, where the device would just sink that, as the device itself only consumes about 5 uA and there may be up to 1 mA of current through that diode:
So, instead I would prefer to have some other means to limit that voltage. The typical approach here is to use a Zener diode to ground, bypassing the system itself. Of course that also fails, as there are no Zener diodes with a sharp enough knee in that voltage range to ensure that there is voltage limitation above, say, 3.5 V and very low current consumption below 2.8 V, which is kind of required for this circuit to function. I also can not add a resistor divider before the input and then use a Zener diode with a higher Zener voltage, because that would create a pull-down which in conjunction with an additional pull-up, would greatly increase current consumption at the permissible resitance range.
So, the last solution I see is some more complex method of limiting the voltage. The approach I prefer is some sort of shunt regulation. I did not find any shunt regulators sufficiently low-power for my needs, like the ALD111933, which is a trimmed-threshold FET pair. With it, I could create a circuit like this:
Do you see a particular problem with this? There is no need to filter out extremely fast transients, another mechanism will take care about that.
Are there any alternatives? I thought about low-power operational amplifiers, but they have such low bandwidths, that even with the transient filter, they would be too slow.
Also, the only really acceptably trimmed FETs seem to be from that manufacturer. Are there others, that I just can not find?
Edit: I've kept the actual signal on the input hidden, sorry for that. The reason for it is, that it can be quite diverse. In general, it should be a digital or analog signal between 0 V and 2.8 V; analog signals at a bandwidth of less than 100 Hz, digital signals up to 10 kHz. There are no protection diodes in BUF1, which has an absolute maximum at to 4.3 V. In case of errors that may occur during normal operation, I have to expect voltages up to about 8 or 9 V. In case of severe errors, that should not occur, but would be nice to withstand, the voltage may rise to up to 30 V at the input. I can also scale R1 appropriately, selecting low-leakage diodes. BUF1 also has very low leakage current, so R1 would be around 200 kOhm.
Edit: Another small clarification: I'm not just interested in blocking transients. It may be necessary to allow a over-voltage to be applied over the duration of days until the fault can be fixed.