When designing an ultra-low-power device that is powered by a primary lithium battery, is it necessary to include over-voltage protection?

Specifically, I am using a Microchip (Atmel) SAML21J which specifies a supply voltage of 1.62-3.63V.

I am using Lithium-Thionyl Chloride (Li-SOCl2) batteries with an open circuit voltage of 3.65V.

In practice, I've never worried about the small potential over-voltage in testing on the bench, both because the voltage immediately drops slightly with load, and presumably the microcontroller can tolerate short-duration small over-voltage. However, as I look at other components with a max voltage rating of 3.6, I wonder if there is value in adding, say, a Zener diode to clamp the voltage.

The factors against adding protection in this case are 1) cost, and 2) potential battery life decrease.

I'm mainly concerned that brand-new batteries installed during production could cause damage to some percentage of finished units.

  • \$\begingroup\$ Using So as a chemical symbol for sodium? Is that the standard name for this battery chemistry‽ \$\endgroup\$
    – Hearth
    Commented May 7, 2019 at 23:07
  • \$\begingroup\$ I believe he means Lithium-thionyl chloride (LiSOCl2). \$\endgroup\$
    – TimWescott
    Commented May 7, 2019 at 23:10
  • 1
    \$\begingroup\$ A zener diode won't bring you joy -- they don't have a sharp transition as they reach their zener voltage, and if one actually worked it'd do so by severely discharging the battery. I'd go shopping for a low-dropout 3-terminal 3.3V regulator. \$\endgroup\$
    – TimWescott
    Commented May 7, 2019 at 23:12
  • \$\begingroup\$ @TimWescott Lithium-thiolnyl is what I meant, thank you. I'll edit. \$\endgroup\$
    – JYelton
    Commented May 7, 2019 at 23:13
  • \$\begingroup\$ @TimWescott The larger question is do you think a 3.3V LDO is necessary? I haven't actually had components fail without some sort of regulation on this design, but I'm relatively new to high-volume battery-powered device design. \$\endgroup\$
    – JYelton
    Commented May 7, 2019 at 23:15

4 Answers 4


I'd suggest you DO need some form of regulation.

Looking at the datasheets for the batteries, the nominal voltage is 3.6V BUT it is temperature dependent. This datasheet from Minamoto shows the terminal voltage rising to about +3.8V at 70°C.

enter image description here

If you can guarantee that everything on your board will tolerate 3.5V, then you could use a low dropout linear regulator such as this Ablic LDO. This has only 53mV dropout voltage and should follow the battery down below that point. It's available up to 3.5V.

  • \$\begingroup\$ +1, the S-1317 LDO looks like a very nice device. Thanks. But how do you know that it will follow the battery down? Is there any undervoltage threshold? \$\endgroup\$ Commented May 8, 2019 at 0:48
  • \$\begingroup\$ @Ale..chenski I don't have experience with this particular LDO, but others I have used provide close to the dropout voltage since they are simply a P-Chan FET turned on. They work all the way down till the FET starts to turn off. That is typically in the 0.9-1.2V range. If this unit is similar you'd expect it to work in the same range ….below 3.6V however it is NOT regulating of course. \$\endgroup\$ Commented May 8, 2019 at 0:52

What datasheets mean: "operating range" translates to "if you're inside this range, we guarantee it'll work like we say". "absolute maximum" translates to "if you're inside this range, it won't be permanently damaged".

If you're between the operating range and the absolute maximum, then you can expect nasties like more than advertised currents, parts not working right at the extremes of temperature, parts maybe not meeting timing (although usually CMOS is faster at higher voltages). If you're just a hair over, then you may be OK, particularly if you don't have to push it really hard.

If you're not above the absolute maximum ratings, and you're buying a whole lot of parts, you probably have the leverage to ask an applications engineer for the company just what you can expect.


First, you can't design a cheap and low-consuming limiter that will be accurate to 20 mV difference.

Second, you need to check manufacturer's QA methods and specifications on their definitions of "absolute max" ratings. All max ratings are derived from accelerated tests at elevated temperatures, and then extrapolated on normal operating conditions using certain assumptions and theoretical formulas. And the assumptions usually are on quite conservative side. More, the max ratings usually imply "prolonged" exposure to max conditions, so episodic fresh battery overexposure for few seconds may have no effect. It could be that the rating is defined for 10,000 years of service, and exceeding the rating by 5% would lead only to 1,000 years according to their aging models (hypothetically speaking).

Lastly, if you are really concerned with reliability and making a high-volume device, you might be able to negotiate special sorting batches from the manufacturer, since there is always a statistical spread. Or if you are happy with reduced lifespan ( like 1,000 years instead of 10,000), you can get corresponding assurance from them.


Table 46-1. Absolute Maximum Ratings
VDD Power supply voltage 3.8 V max

p 1008 of 1155 ref

  • \$\begingroup\$ Thank you. I went looking for this and missed it. While it's good news for the main microcontroller, other components that stop at 3.6V (a Microchip RN4870/71 BLE module for example) are still candidates. \$\endgroup\$
    – JYelton
    Commented May 7, 2019 at 23:39
  • \$\begingroup\$ then go search for the Abs Max specs \$\endgroup\$ Commented May 7, 2019 at 23:45
  • \$\begingroup\$ So, will not die, may not work :-) \$\endgroup\$
    – Russell McMahon
    Commented May 8, 2019 at 9:22
  • \$\begingroup\$ @RussellMcMahon, it likely will work but might not meet all specifications at highest or lowest temperature on some rare corners of silicon batches. It might exceed the stated leakage or standby current by 10-15% or something. If you have serious benefits from not doing any voltage conditioning, you should/can request reliability data from manufacturer, and calculate your risks. The max margins are usually engineering estimations times extra safety margins imposed by business people and technical marketing. If you exclude corner temperatures, the voltage margin might be much higher. \$\endgroup\$ Commented May 9, 2019 at 3:15

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.