# Long term effect of over-voltage on clamping diodes and ADC reference voltage

I am contemplating building a small voltage/current monitoring circuit that needs to monitor a wide range of voltages and currents. Currents will just be a single low-side shunt.

Monitoring will be done by a number of ADC inputs on an MCU at 3.3V.

Input voltages for both voltage monitoring and current shunt will be 0-24V.

Now, not only do I want to measure the 0-24V (through simple voltage divider), but also the lower portion of that same voltage range (0-3.3V) in greater resolution, so directly connected without the voltage divider. Similar arrangement for the current shunt - different voltage dividers to measure different current ranges at different resolutions.

So I am thinking the simplest input stage for the over-voltage inputs will be a simple resistor and clamping diodes to +3.3V:

simulate this circuit – Schematic created using CircuitLab

Now the 10KΩ will limit the diode current to a maximum of around 2mA, so the diodes won't care much about it. My question is, what will happen to the power supply voltage and, more importantly, the ADC reference voltage (I could use an external reference, but for this I will probably just use Vdd) when 24V is applied for an extended period of time (I'm talking hours or even days here).

Should I be considering adding extra decoupling to the input of the MCU? Maybe an additional small resistance? To the input of the MCU (A), or to the diode->Vdd connection (B)?

simulate this circuit

Would either of those be recommended, necessary, or pointless? Given the small (2mA) current, would there even be an issue to worry about?

(The simulations I have run, though crude [falstad], have all shown Vdd staying at 3.3V, but I don't trust it, and I don't have PSpice).

• What I'd be worried about would be your external clamping diode possibly having a higher knee voltage than the internal one, in which case the current may flow through the internal one instead - many probably can tolerate your 2ma, but perhaps not all. – Chris Stratton Nov 19 '14 at 18:05
• Yeah, I've already considered that. I'll be shopping around for the lowest $V_F$ Schottky's I can find to get around that problem. The MCU in question has an abs max input on any IO pin as VDD + 0.3V, so I'm assuming the internal ESDs have 0.3V $V_F$, so will be looking around for 200mV-ish diodes. – Majenko Nov 19 '14 at 18:16
• Low Vf tends to imply higher leakage, so with 10K in series and a 1M input impedance, you may need to keep an eye on that too. You might out of curiosity sacrifice an MCU to measure the actual forward voltage. – Chris Stratton Nov 19 '14 at 18:19
• Those are just ballpark figures for now. Not sure what the exact impedance is, and the 10K can be changed at will to suit. – Majenko Nov 19 '14 at 18:21
• I'm not sure what Resistor-B does. I've added R-A's to opamp circuits, but only when forced. Does the MCU care if the supply varies a bit? – George Herold Nov 19 '14 at 22:04

If you're already pulling more than 2mA from the supply, then the 2mA from the clamping diode will simply reduce the load on the supply by that much.

From a DC standpoint, it only matters when you normally draw less than the clamping current because then you're backfeeding the supply, which may allow the voltage to rise.

From an AC standpoint, you might need some filtering depending on what the 24V input is actually doing and how sensitive the circuit is to that kind of supply noise.

Other than that, I think the first version is good.

• Supply would be an LM1117-3.3 [compatible], and the MCU draws in the order of 50mA most of the time, so that's not an issue. Monitored voltages could be anything - right now I could do with it for monitoring voltage and current for a (max) 19V 2.5A constant current supply while trying to rescue some dead SLA batteries, so fairly static (17.55V, 164.2mA at this precise moment), so no high frequency components. – Majenko Nov 19 '14 at 17:09

Don't shop for the lowest Vf diodes you can find unless you don't care about accuracy. You've got to trade off accuracy (due to leakage) vs. Vf. In practice Schottky diodes will guarantee that you don't get significant current flowing through the protection network. Something like a BAT54 is appropriate.

Now, pedants will whine that the data sheet says something like 300mV absolute maximum and we're probably exceeding that at low temperatures- but really it doesn't matter because the silicon diode drops in the protection network will similarly increase so no significant current will flow.

• I have found some reasonable ones at 290mV Vf which is just below the 300mV suspected Vf of the in-built ESDs. Panasonic DB2730800L: semicon.panasonic.co.jp/ds4/DB27308_BED.pdf - should do the job nicely. I'm not too worried that much about absolutely precise accuracy, and I can always compensate in software. – Majenko Nov 19 '14 at 23:11
• It typically leaks 2uA at 25°C & -2.5V , only a 1% error maybe you don't care about 1%, but by 85°C it's up to 100uA (something like a 50% error). This cannot realistically be compensated. The BAT54 will typically leak 2uA at 75°C, but maximum is 10x worse so it might work for you, but.. – Spehro Pefhany Nov 20 '14 at 0:01