1
\$\begingroup\$

I am designing a lithium-ion battery operated device. My concern is when the battery voltage drains to near min. operating voltage (3.3V) of the system, the system starts to reboot and struggles to be up. If I don't add any circuit to cut the voltage off, is there any reliability concern? My system consists of only digital and analog ICs, like microprocessors, modem, USB, PA, buck, boost. Thanks.

Edit: The battery has its own protection circuit for over-discharged, overcurrent, etc.

\$\endgroup\$
  • 4
    \$\begingroup\$ If you undervoltage (over-discharge) a Lithium battery, you can severely reduce the number of charge cycles that you can get out of it. If you undervoltage a DC-DC convetrter, you could fry the power transistors. So, reliability concerns vary. A block diagram of your system would help here. \$\endgroup\$ – Nick Alexeev Jan 4 '17 at 23:48
  • 1
    \$\begingroup\$ In the systems I have worked on, there is not usually a UVLO for the battery. Or if there is, it is secondary. Usually the processor senses battery voltage and shuts down the system before unreliable behavior occurs. Also, very early in the boot process, the battery voltage is checked. Before any user visible events occur. If the voltage is not at least 3.4 or 3.5 (or whatever) the processor will immediately return to its lowest powered state ("off"). This eliminates reboot loops. And if they DO occur, the user won't know. It will just loop silently until the protection circuit kicks in. \$\endgroup\$ – mkeith Jan 5 '17 at 8:20
  • 1
    \$\begingroup\$ In alkaline battery powered devices, you must insure that current draw is less than around 20uA once cell voltage drops to 0.8V per cell. So they typically do need some type of hardware UVLO. Otherwise there is some danger of a cell leaking electrolyte which can sometimes ruin the battery compartment or even the electronics. \$\endgroup\$ – mkeith Jan 5 '17 at 8:22
  • \$\begingroup\$ What they all say BUT, > 3.3V is rather high for endpoint voltage on a LiIon cell. What are the actual battery and system voltages when reset looping starts? Modern LDO regulators should be able to approach dropout voltages of around 0.1V. (Vbat-Vsystem) differential. Adding a larger system resevoir capacitor MAY allow lower very low voltage limits (as small dips do not cross the low threshold as easily. \$\endgroup\$ – Russell McMahon Jan 5 '17 at 12:05
  • \$\begingroup\$ @RussellMcMahon, in low-discharge devices (10 hours or more of battery life), you don't give up much if you terminate at 3.3V. Said another way, when discharging at C/10, the voltage will be dropping very rapidly at 3.3V. Of course, we don't know if this is a low-discharge device. \$\endgroup\$ – mkeith Jan 5 '17 at 17:35
1
\$\begingroup\$

I have seen this happen many times during early product development. Usually nothing will permanently break (I've never seen reset loops cause a hardware failure). But there is kind of an understanding that it is undesirable.

Most systems I have seen have battery voltage monitoring by way of an ADC input on the processor. Also, most modern systems have the ability to shutdown under software control. Because of this, there is no need for a hardware UVLO. Here is how a typical system works. The voltages vary.

During operation, if the system voltage reaches 3.5V, you warn the user somehow. When it drops to 3.4V, you unconditionally shut down the system into its lowest power state. (Voltages may vary in your system).

During boot up, very early, before you do anything that would cause a power spike, or be detectable by the user (like turn on a backlit display or play a sound or whatever), you check the battery voltage. If it is below say 3.55V, you do not fully boot up. Return instead to the low-power state. The operation shutdown voltage is LOWER than the minimum boot-up voltage. This way, if the unit shuts down due to low voltage, the user cannot just power it back on.

Also, if it does somehow get into a reboot loop despite your efforts, it will be looping before the battery voltage check which means it will be before any user noticeable events occur. So, the user will never know about the reboot loop. Better not to loop, but if you do, then it is better if the user doesn't know.

Here is another point. If you use hardware UVLO, you have to fully consider what triggers it, how it is released, and what will happen when it is released. For one thing, the UVLO needs to have some hysteresis. If it is triggered at 3.4V, it shouldn't release until a somewhat higher voltage like 3.5V or even more. Otherwise there is a chance you will get your reset loop anyway.

The reason is that the battery voltage sags under system load, then it rebounds when the load is cut off. If the UVLO has no hysteresis, it will just bounce back and forth in a loop as it goes over and under the threshold. But with hysteresis, once it cuts out, it will need a somewhat higher voltage before it reconnects the load. So no bouncing.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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