# Is there an alternative economical low-voltage reference to a single cell?

This is a simple Li-ion battery monitor circuit I designed around a low-voltage LMV331 comparator, which can operate down to about 2.7 V.

I think it's best for a comparator to work with median voltages, so I'm using a 1.5 V alkaline watch battery as my voltage reference.

The behavior is that when a Li-ion cell is connected to V+, the LED briefly flashes if the voltage is around 4 V or more, because the 50 μF capacitor will charge to above 1.5V quickly. If the voltage is lower, the capacitor will take longer to charge, so the single blink of the LED will be visibly slower. Ultimately, when the Li-ion cell is around 3.2 V or less, the LED will stay on indefinitely, indicating it is a good time for a charge.

The resistor values surrounding the 50 μF capacitor can be varied depending on what voltage you wish to signal that a charge would be wise.

My question is about the wisdom of my decision to use a small 1.5 V alkaline watch battery as my voltage reference.

When powered, this comparator draws an immeasurably small current, but presents some load when not powered, hence the 1 MΩ resistor at the cell output is needed to limit current into the unpowered LMV331 to slightly under 1 μA, so the cell should last a good long time but not forever.

That said, I'm sure everyone can see my distaste for using a cell as a voltage reference, knowing its longevity is less than infinite. But I'm not sure what else to use to offer such a low-voltage reference economically.

I know there are "bandgap" low-voltage references, but they are expensive. Obviously, a simple Zener diode / resistor would be a poor choice, as they need significant current to form a stable reference, and I've found a series of 2 or 3 ordinary forward-biased silicon diodes doesn't work well either.

So, is my circuit the best that can be expected, or is there another part I should consider?

• You can buy semiconductor reference IC's. For example ISL21080. Perhaps the 1.5 V 0.5% reference. Typical Iq is 600 nA, and max is 1.5 uA. But if it is too expensive then I guess some type of battery might work OK. Pick something with a long shelf life. You might be able to use a red LED, also, if you cover up the lens with black paint. Feb 26, 2022 at 3:33
• thanks! I'll look into that one! Most of the Analog Devices references that also require very little current were more expensive for the single part than my whole circuit pictured here! Feb 26, 2022 at 3:50
• @Randy If you are just looking for a simple and cheap comparator, with hysteresis, that can be built with just two BJTs (but it depends on the need.) A 3rd BJT can drive the LED. No need for a battery. I don't see hysteresis there in your circuit. Do you care? What's the range of operation vs ambient temperature? And would you consider a device that blinks the LED at a rate proportional to the voltage, instead? (I've several different approaches in mind.) Finally, I'm guessing you don't really care about the current draw from the Li-Ion? Because this last idea can operate on 2uA.
– jonk
Feb 26, 2022 at 5:39
• @jonk I've seen that circuit you are describing, that blinks the LED. It's not what I'm looking for. My circuit is patterned after the behavior of a musicians device I use, whose LED behaves the way I've mimicked. It is that it is a permanent part of the device, offering a short visual test when the device is turned on, and continues to monitor its battery. In other words, the circuit's current draw is sufficiently small it can remain on drawing next to nothing, unless of course the voltage level drops too far. The circuit. Fortunately the lack of hysteresis does not cause problems in my case Feb 26, 2022 at 7:21
• @Randy Good enough. Thanks.
– jonk
Feb 26, 2022 at 7:23

This is a comparator circuit that I favour for establishing voltage conditions on a Lithium battery: -

Note that it uses a very low power comparator (LPV521) and a very low power voltage reference (LT1389).

Circuit taken from my answer here.

The quiescent current of the LPV521 is typically 300 to 400 nA and, the quiescent current of the LT1389 is about 800 nA. I've also listed other op-amps and references that should be good too.

Compare this with the quiescent current of your LMV331 which is over 100 μA and hopefully you'll see that there are much better choices to be made for this design.

• Thanks Andy. I'm reasonably satisfied with the LMV331 for the comparator, but I do like the specs on the Analog Devices LT1389, and actually was one of the parts I considered. I'll probably sample one to try (though there is apparently a LONG wait time at Mouser for these). I had just hoped there might have been a more economical solution. They are over \$8 (USD) at Mouser, which is bout 4X the cost of my whole circuit. :-). Feb 26, 2022 at 20:31
• What about the MAX6006? Feb 26, 2022 at 21:40
• Wow! I'd have never thought a simple shunt zener reference could be that stable at such a low current! Why are the zeners I had on hand(like 3V In4732) that stable?! They drifted all over the place as I dropped the voltage when i didn't feed them at least 10mA! Feb 27, 2022 at 8:34
• It's because it isn't a simple shunt zener; there's more under the hood than a simple zener @Randy Feb 27, 2022 at 10:33
• Right. Well thanks again buddy! I'm going to try to get the MAX6007 (preferably SOT3). Its 2.048. Even without simulating it seems to make sense that its going to easier to get my desired behavior targeting a slightly higher voltage. Now if only I can find someone with stock! Feb 27, 2022 at 15:52

Why wouldn’t you use a single NPN as a bandgap reference? It would work well enough for what you’re doing here, and it would be even less expensive than a battery.

Here's a sim of the idea (try it here):

It's set to light the LED when the voltage is below 3V. You can adjust this. Here's the thing: although the bandgap shifts a bit (about 10mV over the range of interest) the divider can be adjusted to compensate for that, so the threshold will always be the same. Your battery approach can't do that; the reference voltage will shift as the battery ages, ultimately it will fail.

If you're looking for an ultimate low-power solution for battery state-of-charge, Maxim Integrated has something for you here: https://www.maximintegrated.com/en/design/technical-documents/app-notes/6/6378.html

tl; dr: the MAX40000 / 400001 are ultralow power comparators with a built-in 200mV voltage reference. Here's how that would work (simulate it here):

• I think I have tried this and was dissapointed at the difference in the reference voltage produced, over the span of 3 to 4 V. It didn't seem like much, but I now realize the sensitivity of trying to measure the small span of voltages needed to monitor a Li_in cell. I'll give it another try. After all, the slight degradation of the alkaline cell voltage over time is equally a concern. Feb 26, 2022 at 3:47
• Depending on what you have lying around, how about using LEDs in series - one blue LED with a series resistor will start to light at about 3V (add a normal diode if it’s coming on at too low a voltage) and maybe a couple of red LEDs in series for 4V. These will vary somewhat with temperature but that may not be too much of a problem depending on how you want to use the device.
– Frog
Feb 26, 2022 at 4:46
• Try simulating it to see what it does. As it is, you could compensate the non-ideality of the NPN bandgap by adjusting the voltage divider. Feb 26, 2022 at 5:19
• @Frog - The problem with LEDs, Diodes, Zeners etc is that I'm trying to create a reference voltage that does not change over the span of voltages I'm trying to monitor, and and these devices simply won't offer that without significant current. You can see where a continuous battery monitor that draws excessive current is not a good design. Feb 26, 2022 at 7:25
• I added a simulation. Short take: the NPN bandgap varies only about 10mV over the range of interest. That's way better than you will get over the lifetime of your 'reference' battery. Feb 26, 2022 at 8:12

I've found a series of 2 or 3 ordinary forward-biased silicon diodes doesn't work well either.

Why not?

Over a 10 degree C temperature range (more than reasonable for indoor use), the forward voltage of one diode (or diode-connected transistor as in another answer) changes 20 mV. That is less than 1% of the minimum condition you want to detect, and way less than the variability of the other circuit voltages due to using 5% tolerance resistors.

I recommend using a single diode as the reference. Recalculate the two resistors in the voltage divider for a trip-point of 0.6 V, put a pot equal to about 10% - 20% of their total value in the middle, and connect the wiper to the comparator input. Connect a used battery that represents the voltage you want to cause the continuous LED-on condition, and adjust the pot.

The diode's Vf will be dependent on the value of the bias current through it. The less current, the closer the diode's operating point will be to the "curviest" portion of its conduction knee. With a very low bias current, the diode's Vf will change as the battery voltage decreases. This is not an actual problem, because the circuit is a single-operating-point detector. The adjustment calibrates out all circuit inaccuracies at the trip point.

However, since this circuit has by your definition an intermittent connection to the battery, you can increase the diode bias current to something over 1 mA to minimize this aspect of its behavior without decreasing battery life enough to matter.

Update: With a low reference voltage, you might get enough adjustability in the pot without a fixed value resistor between the bottom of the pot and GND. IOW, the pot replaces the 10K resistor (reference designators - !) in your schematic. For example, a 10 K pot and a 39 K upper resistor.

Also, I forgot to mention this before: With the much lower trip point, the size of the timing capacitor for approximately the same circuit behavior will be reduced. First guess, 22 uF.

Update - 2: A nominal 1 mA current through a 1N4148 signal diode will bias it in the middle of the steepest (least-curvy) part of its transfer curve. The average of three different manufacturer's datasheets indicate a Vf of approx. 0.63 V. A 2.7K bias resistor should work well.

This circuit concept behaves exactly the same as the original: the LED blink time increases as the tested battery voltage decreases. The only significant change is that there is an adjustment for the most critical function - indicating that a battery is "too dead to use" with a continuous-on LED. This circuit allows that threshold to be set much more accurately than with two 5% resistors. A side benefit of this is that you also are correcting the threshold voltage for the exact value of the diode reference.

• Thanks. One point you may have missed, or I didn't make clear, is that the behavior I'm looking for is NOT a single "yes no" indication. In my circuit, the single flash is fast when the voltage is higher, and progressively slower as the final voltage is reached. At that point the LED just stays on. Any reference that changes as the voltage changes, and is more "curvey" at low currents, won't be an improvement. Feb 26, 2022 at 19:30
• I didn't miss it. See Update - 2. Feb 26, 2022 at 22:12
• Appreciated. I think I'm going with Andy's recommendation on the MAX600x series. I'll be using the MAX6007, which is trimmed for 2.048 volts. Besides the fact that targeting a slightly higher voltage makes my timing slight more predictable, these "shut" regulators seem to be able to operate with little drift at just over 1uA, which gives me the useful option of having my monitor circuit be permanently on along with the device being powered. Thanks though. Your insight was helpful to me! Feb 27, 2022 at 16:16