Battery undervoltage lockout

[LukeW]

I want to isolate a battery from the load for undervoltage protection. The quiescent current from the battery should be zero or as low as practical once the battery has reached the cutoff state.
The voltage threshold would be say 13.6-14 volts (4S lithium) but I want a trimmer in there for fine tuning anyway. The BOM should be as simple and compact as possible.
I think high-side switching of the load would be nice, not low-side. It should be "auto reconnected" without a latch that requires a manual reset button to turn the load back on.
Would like to have a bit of hysteresis, to prevent oscillation as the battery voltage rises back up slightly when the load is removed.

This attached example is a quick simple '431 arrangement with a voltage divider, driving the Pch-FET gate. A parallel branch is added to the voltage divider high-side from the output rail, which adds a bit of hysteresis.
Would you do it differently? Any ideas or recommendations?

[Ian.M]

A TL431 is *FAR* too current hungry, and needs too low a resistance divider chain. 

I think you can misapply a LT6650 micropower reference, using its buffer amplifier as a comparator and MOSFET driver.   However as this is *NOT* the usual mode of operation I suggest getting one and doing some tests to see if real life matches the simulation.



R4 & R3 are to provide approx 0.2V of hysteresis.  Rs is only for plotting the quiescent current, its effectively zero.  The quiescent current should be under 10uA, 20uA worst case.  As the LT6650 is rated for 18V operation, it can tolerate the maximum charging voltage that will ever be applied to a 4S pack.

LTSPICE simulation attached with a sweep of the preset.
Edit: R3, 4K7 is meant to be labeled R5.    The actual R3(a,b) is a 10K pot to trim the setpoint.

[IanB]

You could also take a look at this thread where I asked a similar question:

https://www.eevblog.com/forum/projects/designing-low-voltage-battery-protection-circuit-good-reference-designs/

Some good ideas were provided there.

[Ian.M]

Yes, but the >12V battery voltage would be a major problem for the circuits in the thread you linked. 

[mos6502]

Replace the TL431 with a TLV431, and you can get a current draw of less than 500uA. Should be fine for any lead acid battery.

Another option would be an LM10, which draws 270uA typical. It's a pretty expensive part though.

[LukeW]

I think you can misapply a LT6650 micropower reference, using its buffer amplifier as a comparator and MOSFET driver.   However as this is *NOT* the usual mode of operation I suggest getting one and doing some tests to see if real life matches the simulation.

That looks pretty nice  Will have to order some silicon.

Q: assuming that I did want a 431 based solution (or TLV431, which would be better) and I was willing to accept the hit on the quiescent current, what's the best way to add some hysteresis to the 431 circuit?

My original post shows an extra resistor added to the 431 circuit to provide the hysteresis, but I haven't tested this resistor on actual hardware, and I'm not sure if it's the right approach.

[amyk]

A 4S lithium pack? Do any of these fit?

http://www.sii-ic.com/en/semicon/products/power-management-ic/lithium-ion-battery-protection-ic/

[Ian.M]

Some more thoughts on the LT6650 circuit - If the battery voltage drop under load is more than the hysteresis, you get pathological behaviour up to and including high frequency power oscillation, and it can also burn out its MOSFET by operating in the linear region. 

Replacing the 10Meg hysteresis resistor with a small signal N-MOSFET (e.g. 2n7002) driven by the same gate signal as the power P-MOSFET, to short out a few K of resistance at the bottom of the divider chain is a safer way of implementing hysteresis as it isolates it from the load capacitance.  Also passing the load current through a low-ohm resistor at the bottom of the divider chain adds positive feedback to increase the hysteresis and compensate for battery internal resistance, preventing oscillation and burnout.

I've attached a new LTSPICE schematic implementing this.  R4 is the compensation resistor, and R5 is the hysteresis preset.  Adjust the threshold (turnoff) with no load, then the hysteresis preset adds up to 1V to that for the turnon setpoint.

Due to the very low value of R4, you may choose to use a PCB track.  You can also put a 100R preset across R4 and take the bottom of the divider chain to its wiper to implement adjustable load compensation.

[LukeW]

Some more thoughts on the LT6650 circuit - If the battery voltage drop under load is more than the hysteresis, you get pathological behaviour up to and including high frequency power oscillation, and it can also burn out its MOSFET by operating in the linear region. 

Replacing the 10Meg hysteresis resistor with a small signal N-MOSFET (e.g. 2n7002) driven by the same gate signal as the power P-MOSFET, to short out a few K of resistance at the bottom of the divider chain is a safer way of implementing hysteresis as it isolates it from the load capacitance.  Also passing the load current through a low-ohm resistor at the bottom of the divider chain adds positive feedback to increase the hysteresis and compensate for battery internal resistance, preventing oscillation and burnout.

I've attached a new LTSPICE schematic implementing this.  R4 is the compensation resistor, and R5 is the hysteresis preset.  Adjust the threshold (turnoff) with no load, then the hysteresis preset adds up to 1V to that for the turnon setpoint.

Due to the very low value of R4, you may choose to use a PCB track.  You can also put a 100R preset across R4 and take the bottom of the divider chain to its wiper to implement adjustable load compensation.

Wine/LTSpice under Linux is being a pain to get working for me at the moment.
Would you mind sharing a schematic pic? Thanks.

[Ian.M]

No problem.  Its worth persevering with WINE ad LTSPICE though.

[Zero999]

Replace the TL431 with a TLV431, and you can get a current draw of less than 500uA. Should be fine for any lead acid battery.

The TLV431's maximum voltage rating is 6V. It's possible to use it in circuits which run off higher voltages but other components are required to keep the voltage across with it below 6V.

[nfmax]

Here's my take on the undervoltage lockout circuit theme. This is still a work in progress, designed to fix my home intruder alarm system which currently runs its backup battery down to death if a power cut lasts too long. It uses a 12V, 1.2 A-hr valve regulated lead acid cell. There is one pair of leads connecting it to the panel, it appears to be used in float mode.

I apologise for the scruffy hand-drawn circuit, but if you can read it you will see an LM10 configured as a comparator, controlling a BUZ11A FET in 'backwards' mode as the switch (since I had one handy). The turn-off threshold is about 10.9V, the turn-on about 12.1V. Assume the FET is off to start with. The charge current from the alarm panel flows through the body diode of the FET and the battery charges. When it reaches 12.1V terminal voltage, the FET turns on, shorting out the diode - the battery now sees pretty much the full original charging voltage, and operates in float mode.

Now assume the mains supply fails and charge power is removed. The battery supplies load current through the on FET, with the LM10 and sensing network taking about 300uA additional load. When the battery discharges to 10.9V, the LM10 turns the FET off. The voltage at the FET drain now starts to rise toward +10.9V (assuming there is some capacitance in the panel to slow it down a bit) and when it reaches about 6V or so the CMOS Schmidt trigger switches, driving the outputs of the 4 paralleled stages which power the LM10 low. The LM10 and sensing network are now powered off, and the current drain from the battery is essentially zero. Even if the battery voltage recovers above 12.1V (unlikely) the FET cannot be turned on again. The only way to reset it is to apply charging power to the FET drain. This switches the CMOS Schmidt trigger back, reapplies power to the LM10, and once the battery charges to 12.1V, turns the FET on again.

The 1k0 resistor in series with CD40106 pin 1 is protection to limit the pin's input current when the FET drain voltage is negative of it's source (FET off, battery charging from the alarm panel through the body diode). The 4u7 capacitor on the LM10 supply is there to slow the rate of voltage rise so that the LM10 has time to turn on (which takes about 150us) and drive its output low before the LM10 supply reaches the FET threshold voltage. The LM10 output seems to always power up high at turn-on. With 100nF here, the FET is glitched on for about 250us - this doesn't seem to do any harm, but I thought it best to avoid it.

The circuit also acts as a 'battery seal', since it will power up in the off state when connected to the battery. The circuit is only really suitable for 12V batteries, as although the LM10 will take 40V, neither the CMOS VDD-max nor the FET VGS-max allow it.

Any comments?

[LukeW]

http://www.hobbielektronika.hu/forum/getfile.php?id=234807

There's also this one that Google turned up for me - courtesy of our Silly Chip friends.

Interesting architecture - a bit more complex in terms of BOM, but probably comes out cheaper.
The jellybean Microchip comparators, opamps, LDOs etc are very competitively priced compared to most expensive LT stuff.

The low-side switch is not optimal, IMO, but it depends on your load.