Limiting sensor input voltage without biasing measurement

I need to measure precise voltage in the range from 0.375 volt to 10.1 volts from a source with 100ohm impedance. Maximum voltage however, may be well above 10.1V, so I need to be sure I do not exceed the ADC maximum. The ADC I am using measures from 0V to 4.096V with an absolute input maximum input of 5.3V. Impedance is about 15 Mohm. Target accuracy is on the order of 100uV.

I tried using a resistor voltage divider with a combined resistance of about 1k ohm to reduce the 10.1V input to about 4.0V, and added a 5.1V zener diode as protection against excessive input voltage. However, at 4V divided voltage, the zener has enough reverse leakage to drop the measured voltage by 10mV.

Is there a better way of using zeners?. I simply put it in reverse between the divided voltage and ground. Alternately, I could use an opamp to scale the voltage instead of a resistor divider, but that adds parts and complexity.

Any suggestions?

• Instead of describing it in words, include the schematic! Oct 9, 2019 at 9:48
• "Maximum voltage however, may be well above 10.1V, so I need to be sure I do not exceed the ADC maximum." - you need to decide what it is. Oct 9, 2019 at 9:56
• silicon diodes with Ideality Factor of 1 (schottky diodes from Hewlett Packard were close to 1) will increase their current 10:1 as voltage across diode increases by 0.058 volts, or 1,000:1 for 3*0.058 or 1,000,000:1 for 6*0.058 or about 0.35volts. Thus silicon diodes should be ample to clamp, with a much steeper clamping action than the Zener. Oct 9, 2019 at 10:11
• Maximum voltage however, may be well above 10.1V - what does that mean? You cannot design anything until you put an upper limit on that number. Could it be 20V? 100V? 10kV? More? If your signal range is actually ~0-10V, you're simply using the wrong ADC. 0-10V is a very common range for ADCs and they are readily available in all forms in that range. Why do you insist on using the wrong ADC, plus a hack, when you could simply select a more appropriate converter?
– J...
Oct 10, 2019 at 11:54

Zeners have a very 'soft' characteristic. As you discovered, they start drawing current well before their advertised voltage.

The best way to limit overvoltage into an ADC is to use a silicon diode to a fixed voltage. Usually this will be the ADC rail, especially if you put a series resistor between the diode and the ADC to limit any current drawn, but you could use a biassed zener to provide the fixed voltage if the rail is not suitable.

While general purpose diodes like 1N4148 are fairly low leakage, there are specialist 'low leak' diodes like BAS116. Here are some figures I took a while ago, to characterise diodes used for ADC protection.

Reverse leakage, and effective resistance around 0v at 15°C.

diode    reverse leakage   slope resistance over +/- 10mV
-----    ---------------   ------------------------------
1N4148      4nA  (5V)            30Mohm
BAT42      35nA  (5V)             1Mohm
BAS116    <10pA  (30V)         >>20Gohm (10pA was measurement resolution)


The BAS116 reverse leakage typ/max spec at 25°C is 3pA/5nA, and 3nA/80nA at 150°C

Here's the forward conduction current of BAS116 versus 1N4148
(it might have been a quite leaky 4148 sample, below 300mV it conducted more than a 1N540x)

 Vf     BAS116   1N4148
----    ------   ------
300mV    40pA     1.3uA
450mV    45nA     27uA
640mV    16uA     1.7mA


One way to get a low leakage clamp is to employ a BJT emitter-base junction. For example,

simulate this circuit – Schematic created using CircuitLab

Typical peak current at 4.2V is about 30nA leading to an error of 3uV with a 100 ohm source resistance.

It clamps to typically about 5.2V with 12V applied with power dissipation of about 350mW, which is acceptable at moderate ambient temperatures.

Of course you'd need additional circuitry if you intend to protect against negative voltages.

consider something like this

simulate this circuit – Schematic created using CircuitLab

Thanks all for your great answers suitable for a variety of situations.

In my case, I am trying to use items already on hand. Hence, the ADC with a restricted range. Also, the ADC works from a 5V supply, so originally I did not have a voltage above 10.1 as reference. For future reference, I see that it is dead simple to chose an ADC with an appropriate range and then limit input with low leakage silicon diodes up to the ADC supply.

Given items I have in hand, I am revising this to use LM258 op-amps as a fractional amplifier front end and including a 12V supply level for the op-amps. The LM258 data-sheet says "Because the inputs have no internal diodes to VCC, the input voltage can exceed the VCC voltage. If this occurs, the input will block current flow due to a reverse-biased diode forming in the input PNP transistor." Max input voltage for the op-amp is 36V which should cover all possibilities.