UVLO circuit design for NiMH battery system

[_Vendetta_]

Hello EEVbloggers,

I wanted to design an under voltage lockout circuit for some project. The system is powered using a 3 NiMH cells connected in series. The circuit needs to cut off the power when the voltage of cells drop to about 2.8V~3V and connect it back only if the voltage is roughly > 3.6V. The NiMH cells always bounce back to 3.4-3.5V when they are fully discharged (at least in my case). therefore, i want to mitigate that effect.
So my idea was to go with an hystersis based circuit (see the circuit in the attached picture below). The idea is technically to use a rail-to-rail low voltage op amp (LMV358 my case) to do the trick since i would need the output to drive some TTL stage ahead to disconnect the battery. The simulations were fine. However, when i tried to implement it using a regular LM358, the hysterisis was a bit off as the cut off occured at about 3.2V not 2.8V~3V as wished and i do not currently have a low voltage rail to rail Op amp.

So my questions are:
- Does the rebouce effect of the cells mess with the thresholds of the hystersis loop ?
- Are there any alternative solutions (using the minimum of components) to implement this ?

Thank you in advance,

[mazurov]

- Are there any alternative solutions (using the minimum of components) to implement this ?

Depends of what you are planning to do with the battery. For example, if your system requires a stable power rail many DC-DC converters already contain UVLO circuitry. The most you'd have to do in this case to use it is to calculate a voltage divider (2 extra components) and some of them don't even require any external parts (0 extra components).

[magic]

A few remarks.

Perhaps some Li-ion protection IC could work?

TL431 is a big power hog and it seems you will keep it powered during shutdown to wait until the battery is recharged. Research low power substitutes, I know they exist but can't give concrete part numbers.

TL431 power consumption decreases when its reference input is less than 2.5V. So I recommend connecting it as below, to further reduce power consumption in shutdown. The low power variants may or may not have this property too, so RTFM

With some trickery, it may be possible to do away with the opamp completely. See below. This requires a logic level MOSFET. TL431 can only pull its cathode to ~2V, which doesn't really leave much Vgs when supply is down to 3V. Look for 431 alternatives with 1.25V rather than 2.5V reference voltage - I think these could work (and yes, they exist).

In the example:
R1/R2 is sized to turn-on the xx431 chip at 3.6V input.
R4 reduces the threshold to keep it going down to 3V input.
R3 limits current drawn by the 431. Current needs to be at least the minimum spec in the datasheet (1mA for original TL431).

[Peabody]

Isn't the problem that the simulations use a rail-to-rail opamp, but the LM358 used in the physical circuit isn't rail-to-rail?  With positive feedback, the ouput of the opamp in simulation will be either at ground or at the battery voltage.  But in practice the LM358 outputs don't extend to that range, particularly on the high side.  So it seems that would affect the hysteresis points.  It seems your circuit should work with either type of opamp, but not with the same resistor values.

The other thing that might affect the switching points is input offset voltage.  You can simulate that by adding a voltage source of +/- a few mV in series with the non-inverting input, and see if that changes anything.

Also, you might look at the MCP6041 or MCP6141 opamps, or the equivalent MCP6541 comparator with push-pull output and some internal hysteresis.  These are all very low power parts - 1uA quiescent current, and Digikey actually has some in stock.

I wonder if something like this actually requires the precision of the TL431.  Perhaps a zener, or even an LED, would work well enough.

[magic]

The original circuit may work wrong with LM358 because of input common mode range too.

Something like LM385-1.2 is infinitely better than any LED for power consumption.
The input needs not be rail to rail if the reference is lower (like 1.25V).
And an opamp isn't needed, just a comparator.

LM385-ADJ looks like it could replace TL431 in my circuit above, if they still make it.

[_Vendetta_]

- Are there any alternative solutions (using the minimum of components) to implement this ?

Depends of what you are planning to do with the battery. For example, if your system requires a stable power rail many DC-DC converters already contain UVLO circuitry. The most you'd have to do in this case to use it is to calculate a voltage divider (2 extra components) and some of them don't even require any external parts (0 extra components).

For my case a DC-DC converter would be an overfit for my application. Plus, i would not want to deal with high frequency signals at least for now. Thank you for the head up though !

A few remarks.

Perhaps some Li-ion protection IC could work?

TL431 is a big power hog and it seems you will keep it powered during shutdown to wait until the battery is recharged. Research low power substitutes, I know they exist but can't give concrete part numbers.

TL431 power consumption decreases when its reference input is less than 2.5V. So I recommend connecting it as below, to further reduce power consumption in shutdown. The low power variants may or may not have this property too, so RTFM

With some trickery, it may be possible to do away with the opamp completely. See below. This requires a logic level MOSFET. TL431 can only pull its cathode to ~2V, which doesn't really leave much Vgs when supply is down to 3V. Look for 431 alternatives with 1.25V rather than 2.5V reference voltage - I think these could work (and yes, they exist).

In the example:
R1/R2 is sized to turn-on the xx431 chip at 3.6V input.
R4 reduces the threshold to keep it going down to 3V input.
R3 limits current drawn by the 431. Current needs to be at least the minimum spec in the datasheet (1mA for original TL431).

Assuming a quiescent consumption of ~5mA, my application tolorates that. I am aware of having lower voltage references (at 1.235V) but the price point is way higher than rather a simple TL431 temperature stable solution. Your circuit looks cleaver, but i could not see the hysterisis in it? 

[_Vendetta_]

Isn't the problem that the simulations use a rail-to-rail opamp, but the LM358 used in the physical circuit isn't rail-to-rail?  With positive feedback, the ouput of the opamp in simulation will be either at ground or at the battery voltage.  But in practice the LM358 outputs don't extend to that range, particularly on the high side.  So it seems that would affect the hysteresis points.  It seems your circuit should work with either type of opamp, but not with the same resistor values.

The other thing that might affect the switching points is input offset voltage.  You can simulate that by adding a voltage source of +/- a few mV in series with the non-inverting input, and see if that changes anything.

Also, you might look at the MCP6041 or MCP6141 opamps, or the equivalent MCP6541 comparator with push-pull output and some internal hysteresis.  These are all very low power parts - 1uA quiescent current, and Digikey actually has some in stock.

I wonder if something like this actually requires the precision of the TL431.  Perhaps a zener, or even an LED, would work well enough.

Depending on my calculations, the inverting points would occur roughly at about:
V+=R2/(R2+R3) * ( VOH - 2.505V) + 2.505V
V- =R2/(R2+R3) * ( VOL - 2.505V)  + 2.505V
Where VOH is the saturation high out of  the op amp and the VOL is the low output (roughly 0.2V something for  the LM358 ?).
The thing is that i do not currently have any problems with the high hysterisis threshold. The lower threshold is the problem.

[Buriedcode]

The LM358 can only really output close to the negative rail with a very light load >10k.  Your output already has a 10k load in that schematic, and with R3 being 5k, will load it further.  VOL can rise dramatically when more than about 80uA.  Did you measure the output of LM358 when its low?  the LMV358 will definately get much closer to ground with that load on its output.

Secondly, as others have also pointed out, the common mode input range of the 358 includes the negative rail (ground) but not really within ~1.2V of the positive rail.  If your battery voltage is 3V, the voltage from the divider is 3 * (270 / (270 + 100)) = ~2.19V.  The 358's input would top out at roughly 1.8-2V here.  Note the LMV358 has the same input voltage range - however in my experience it tends to be slightly wider, and works within about 1V of the high rail, that might explain the difference also.

Ultimately, best to test it with a real LMV358    There are much lower power CMOS opamps such as microchips offerings that have medeocre specs, but for their intended purpose are more than adequate.  Also, this is really a job for a comparator, for which there are also many low power options.

[magic]

Assuming a quiescent consumption of ~5mA, my application tolorates that. I am aware of having lower voltage references (at 1.235V) but the price point is way higher than rather a simple TL431 temperature stable solution. Your circuit looks cleaver, but i could not see the hysterisis in it?

Those must be some big batteries because 5mA will kill fully charged AA cells in two weeks

Hysteresis is by R4. When the PMOS pulls it up, feedback voltage increases and keeps the 431 turned on.
I just noticed there is a small problem: R4 is not open circuit in shutdown, it pulls the divider down, depending on load characteristics. Well, either take it into account or add 1N4148 in series with R4.

Some of those low power 431 may be much cheaper than LM385. The point is, it can eliminate the comparator/opamp if you also find a 1.8V logic level PMOS. (NMOS in case of LM385, the whole circuit is upside down then).

[_Vendetta_]

The LM358 can only really output close to the negative rail with a very light load >10k.  Your output already has a 10k load in that schematic, and with R3 being 5k, will load it further.  VOL can rise dramatically when more than about 80uA.  Did you measure the output of LM358 when its low?  the LMV358 will definately get much closer to ground with that load on its output.

Secondly, as others have also pointed out, the common mode input range of the 358 includes the negative rail (ground) but not really within ~1.2V of the positive rail.  If your battery voltage is 3V, the voltage from the divider is 3 * (270 / (270 + 100)) = ~2.19V.  The 358's input would top out at roughly 1.8-2V here.  Note the LMV358 has the same input voltage range - however in my experience it tends to be slightly wider, and works within about 1V of the high rail, that might explain the difference also.

Ultimately, best to test it with a real LMV358    There are much lower power CMOS opamps such as microchips offerings that have medeocre specs, but for their intended purpose are more than adequate.  Also, this is really a job for a comparator, for which there are also many low power options.

Yes, Your explanation makes sense. I will buy a LMV358 and redo the test !

Assuming a quiescent consumption of ~5mA, my application tolorates that. I am aware of having lower voltage references (at 1.235V) but the price point is way higher than rather a simple TL431 temperature stable solution. Your circuit looks cleaver, but i could not see the hysterisis in it?

Those must be some big batteries because 5mA will kill fully charged AA cells in two weeks

Hysteresis is by R4. When the PMOS pulls it up, feedback voltage increases and keeps the 431 turned on.
I just noticed there is a small problem: R4 is not open circuit in shutdown, it pulls the divider down, depending on load characteristics. Well, either take it into account or add 1N4148 in series with R4.

Some of those low power 431 may be much cheaper than LM385. The point is, it can eliminate the comparator/opamp if you also find a 1.8V logic level PMOS. (NMOS in case of LM385, the whole circuit is upside down then).

I think that a 432 may also do the trick sense it only requires a cathode current of 100uA which is 10 times less than 1mA of the 431 for a reference of 1.25V instead of 2.5V.

[Peabody]

Back to the question of simpler, lower cost options, I just want to mention that a microcontroller may be a better choice than analog circuits for stuff like this.  A cheap little 8-pin job will typically have a built-in 1.1V reference and an ADC peripheral, can sleep at  extremely low current, take up less space, and wouldn't waste any current in resistor dividers or feedback circuits.  If you can find any in stock.  The AVR parts (see edit), for example, will let you do ADC with Vcc selected as the reference, but measuring the 1.1V internal.  So as battery voltage drops, the ADC result goes higher.  There's no need to divide down to 1.1V, so no current used for that.

May not be appropriate here, but future readers might need to be reminded of this option.

Edit:  I may be wrong about this.  The Atmega328P has this feature, but I don't know if any 8-pin parts do.  The Attiny85 does not.

Edit2:  No, I was right the first time.  It looks like the ATTiny25/45/85 parts do let you sample the reference bandgap.

[magic]

Maybe you could program the brownout detector to do the work?

There are dedicated "voltage detector" ICs which do exactly what you would expect while consuming 1µA or less current; they are to TL431 what comparators are to opamps.
They typically have little hysteresis so some tricky feedback network would need to be created to get more.