I don't know about FB, but I'd like to build one. I got started at one point with the three terminal sensors used to detect power (from Digikey) but got sidetracked. If you've some advice, I'd love to hear it....
I was just going to suggest a 3 terminal undervoltage sense chip,
one for EACH (presumably series) cell that drives a negative
logic AND gate, with the output of the AND gate driving a
N Channel MOSFET.
The MOSFET would be a N channel, low gate voltage, low 'on'
The 3 terminal undervoltage sense chip(s) would have to trip at
something like 1.1 volts to protect each cell. I dont know if
it's possible to find this yet or not. If not, or if it has
to be custom ordered, perhaps one of National's low voltage
comparator chips would work. The top sense chip would have to
sense a negative voltage referenced to the positive supply
because most likely the chip wont run at only 1.1 volts.
All the sense chip power supplies connect to max plus voltage
and ground (most negative battery terminal). With three series
cells this would provide 4.5v during batt high and 3.3v during
batt low to each sense chip.
The negative logic AND gate would simply be schottky diodes,
one for each sensor chip. The common of all the diodes
connect through a 100k resistor to max battery plus terminal.
This junction also connects to the gate of the MOSFET. When
either of the sensor chips outputs goes low it turns off
the MOSFET. This also provides a convenient node for a low
current on/off switch. Total supply drain during 'off' mode
is 45ua plus a small additional current that the sensors
draw (very small also).
I suspect that because the trip voltage is so low (1.1 volts)
the low voltage comparators would have to be used. They can
all use the same reference voltage if also using a resistive
voltage divider to provide references of 1.1v, 2.2v, with the
top comparator using a lower reference (also from the
divider) and a separate two resistor voltage divider.
Perhaps someone makes a chip with several sense inputs already?
Would be nice :-)
I'm interested in a protection circuit as well, that I could possibly drop into any future bike light mods I do, since they tend to require quite a few (4-6) NiMH cells.
I hacked the low-battery indicator on my bike light to trigger at a higher Vbattery (ended up at 4.17V for 4AA). Basically, it compares two voltages dependent on Vbattery, a resistor and diode and a resistor and resistor. Vrr starts high, but falls faster, and when Vrr < Vrd, the indicator comes on. I'll write up a diagram from my notes sometime next week if anyone is interested.
What are your thoughts on measuring the total voltage versus the individual cells? Total voltage is easier and probably requires fewer components, but wildly-mismatched cells could have problems.
Interestingly, I came here just now to comment on that
very same issue.
As i thought about the details a little more, i realized
that to measure every cell individually would be the
ultimate, but would require differential voltage measurements
instead of the simpler 'from the ground up' type.
This would complicate the circuit quite a bit, but
i guess that's life. For the most part it would be
mostly resistors which could be tiny chip types too.
The other complication is that EVERY tap in the series
connection of cells has to have a wire comming off of it
to the circuit so that each node can be measured. This
can be quite a hassle too :-) but i guess if you want the
best, you have to do it.
This is the same problem i ran into when i wanted to
charge a cordless drill batt pack by doing each cell
individually. Lots of connection wires even for a
modest eight cell pack.
The simpler single voltage measurement might be
good enough for some apps so perhaps a circuit
for single measurements and another circuit for
multiple measurements is in order.
Yes wildly mismatched cells would throw off the simple
single measurement method.
Im going offline again to take a closer look, draw up
preliminary circuits, etc.
I've taken a little more time to look at this and
i've been able to find at least three different
approaches to the problem of keeping an eye on
several cells at once.
Approach #1 is the most straight forward and is going to
work the first time it's turned on.
#2 is a little more tricky.
#3 is the most advanced and therefore may require a little
R&D time on the bench, but uses the least parts for the
#4 (which i didnt show) would require PIC programming,
so i dont know if many people would go for that.
My personal favorite is probably the approach that stems
from sampled data theory (#3), but i like the simple #1
too because there's no chance it wont work. I've even
built and tested something very similar to #1 already
for one Li-ion cell, and it worked the first time it
was turned on too.
The following is an Overview of three approaches to
undervoltage protection of NiMH cells in series
Not including the MOSFET which is optional...(see below)
Straight forward differential amps and open collector comparators...
Parts per cell: 1 chip + 4 resistors
Other parts: one voltage ref diode + 3 resistors
Voltage difference reference diode...
Parts per cell: Ref diode chip + 3 set resistors + one-half op amp chip
Sampled data theory approach...
Approximate parts for up to eight cells:
a. 5 resistors + op amp or voltage ref diode
b. two CMOS analog switch chips (16 pins each)
c. CMOS binary counter chip
d. oscillator (single chip or chip + 4 resistors + cap)
e. small cap
Approximate parts for up to sixteen cells:
a. almost double of that for eight cells.
For all approaches:
Note that the MOSFET is optional because there is always the
possibility that instead of turning off the circuit there
could be a single red LED turned on, which would indicate
to the user that the light should be turned off as soon as
Another possibility is to have a red LED for each cell...
where the ones that light show which cells are low.
This would be pretty neat :-) and very functional.
Comments/ideas/suggestions or just what sounds best to you?
Yup, that's where I went, but gave up. I didn't try the analog switch version (neat idea), but got hung up with the need for taps in the battery pack and the 'negative voltage' issue at the top of the stack. That got me back to reading overall battery voltage (probably OK for me, as I was looking at 3 cell packs).
A solution for larger batteries might be breaking them into three (or so) cell blocks and summing the outputs.
I agree that the output can be handled in a number of ways, including having the output start to flicker (that is go from full bright to dimmer in a regualr pattern) rather than have it die flat out.
I dont think i'd use the 311 because it consumes a bit
too much power for battery operation. There are many
low cost chips out there that use way less then 1/10th
the power of the 311. Even the common LM339.
What i need to know now is a few things about your
preferences and the number of cells, etc., ...
1. What kinds of packages do you prefer?
2. What size is the flashlight (or other device)?
3. Can you find a source for the parts (such as LMC7221)?
4. How many cells does your light (or device) take?
There are some packages that are very small, such as
SOT23-5 so im wondering if you've worked with these
or just the usual dip packages.
What you tell me here will make a big difference on the
parts selected for the circuit, although the circuit
itself probably wont change that much so i'll start
drawing it up.
I wonder how it will work as a system? That is when the power switch is off. I see issues if either the V+ or ground leads are broken but the rest of the battery stack is still hooked up.
Neat idea. If you bump up the needed drive to shut down (lower R12) you could add a 'battery's dead' LED in series, or maybe even individual LEDs to tell which cell had died (probably need to be red?)?
Anyway, any advice on my original idea (three terminal 'supply OK' chips driving the FET)?
Thanks for the chip lookup. I like that comparator quite a bit.
Max voltage is 6 volts so it's only good for maybe three cells,
but there are a lot of apps using three cells.
The SOT23-5 package isnt that hard to work with either.
I'd say the 2202 chip would work in this app quite well.
Thanks for bringing that power on issue up...i forgot to
show the power switch :-) Basically, any low current
switch (10ma is good enough) wired to short out D1 to
turn the light (or other device) off.
Multiple LEDs: yeah i like that :-) One red led per batt :-)
I thought i said something about the p/s monitor chips...
I built one already with a 2.63v supply monitor chip and
it worked great, but i couldnt find any chips that would
sense 1.1 volts (or even 1.0 volts) needed for NiMH or
similar cells. Mine was for an Li-ion recharge cell.
I guess if you could find a 2.2v power supply monitor
chip you could use it for two cells in series, if you
dont mind not doing EVERY individual cell like we are
doing here. Perhaps there is a chip out there we
havent found yet.
In any case, the circuit is really simple... output of
p/s monitor chip goes to MOSFET gate. MOSFET source goes
to ground, drain in series with load. You can kill power
with a low current switch by killing the power to
the chip alone (ok, with added pulldown),
which makes it possible to control high
currents with a tiny switch (a side benefit of using these
I hope this is what you meant...
The power switch is not shown on the schematic (yet).
Wire a low current switch across D1 for the power on/off
The actual value of all the 10k resistors depends on
the recommendations of the manufacturer who makes
the cells you intend to use this with. Check to see
what the lowest voltage allowable is and use this
as a guide:
10k for 1.10 volts
15k for 1.05 volts
20k for 1.00 volts
All 90k resistors are really 91k (2%) or 90.9k (1%).
Use either all 1% or all 2% types.
That's great...i hope you can build this circuit
without too much trouble.
You bring up another very good point Doug, as usual :-)
The rationale here is that the draining current when
the light is off is compared to the self discharge
current of a typical NiMH cell. If this draining current
is a small percentage of the normal self discharge then
the drain current is considered acceptable, perhaps even
very acceptable. Now it just so happens that using one
LM339 chip turns out to be approximately 1/8 (that's
one eighth) of the self discharge of a 2000mAh NiMH cell,
so it's deemed 'acceptable'. Using the other comparator
chip specified with the schematic brings even that down
to a very acceptable level. Put another way, the self
discharge initially looks like about a 900 ohm resistor
while the comparator chip looks like around 7500 ohms.
The question then is how much faster do the cells drain
down with 900 ohms in parallel with 7500 ohms as compared
to the 900 ohms alone. Using the CMOS comparator doesnt
even show up on the chart :-)
Of course the reference diode networks bring this up a
little too, about 50ua each. If that's not good enough
then a constant current source would have to be included
for each ref diode. The current decreases quite a bit
when the batterys drain down too.
While what you say might be true in some cases, it's not for AAs. I read the graph as 20% loss in 20 days at 25C from self discharge. A 339 is close to one mA open circuit. That makes them about equal. A second 339 will double that, of course. Half or a third the standby lifetime depending. IMO significant.
Perhaps a small signal PNP could be rigged to shut the 339s down (although I'm not sure of the effect of the inputs at higher voltages is off hand.....).
I may be talking straight out of my bum-hole here since I haven't really looked into undervoltage monitoring before, but what about something like the MAX6709/6714 IC?
They operate down to 2v and have 4 inputs for voltage monitoring and the datasheet schematics show the use of an absolutely minimal number of components. I can't seem to find them on digikey or mouser, so availability may be an issue here.
The package is a 10 pin uMax, so soldering would be tricky for those without a steady hand, but this looks like it could be the backbone for getting the job done very easily. The datasheet circuit also shows low voltage indicator LEDs for each input which seems to be one of the "nice touch" features discussed here as well.
Thanks for bringing that up. I was looking at the supply current
per section, and there are 4 sections to one IC chip so that's
800ua per chip. That's about half of the self discharge, but
even though that's still somewhat acceptable i dont like another
spec on the LM339 for this particular app... The common mode
input range isnt all that good for what we are doing here,
while the LMC7225 has input rail to rail operation. This makes
the 7225 greatly preferred over the 339. Input current is
only around 5ua per chip, which knocks the current down
to very acceptable levels and at the same time puts the
input common mode range in a better area. This means
the LMC7225 should be the only chip used and the LM339
shouldnt be used at all for this circuit. The LP339
looks promising too, but again the common mode range
is crap :-)
With all nice changes comes some downfalls, and this
update is no exception. The downside is the max
number of cells that can be used with this circuit
and the LMC7225 will be 5 cells, not six or more.
I'll update the schematic as soon as possible to
reflect this change.
Hey that's a nice chip, thanks for bringing it up.
If you find any more chips that look promising, please
let us know.
The only problem here is that when monitoring a number
of cells wired in SERIES we want to measure the cells
individually, and that really requires a differential
input voltage measurement. See, our circuit is really
just a "differential input to single ended output"
circuit. The 'differential' part means we need a
measurement of both tabs on an individual battery,
not just the positive tab. This requires a voltage
sensor that has two inputs (not including ground).
The Max chip has just a single input for each
channel so we cant really use that chip here, although
for other apps im sure it could be quite handy
(such as a single cell monitor) so i think we should
keep it in mind.
Well, in the case of a single measurement, you
can simply eliminate all of the batts and their associated
subcircuits leaving just the top section (batt 2 and it's
The only problem with using a single measurement to watch
say two cells is that the cells have to be matched, and
have to drain down at almost the same rate. This is
pretty hard to find in actual practice. Say we set
the circuit to trip off at 2.00v (two cells in series
hoping to trip at 1.00v for each cell).
If the lower cell is 1.4 volts and the top cell drops
to 0.9 volts the total is 2.3 volts, still way over
the trip point. Although we dont want any cells
operating below 1.00v the top cell ends up operating
down to 0.6 volts in this case, which is not good
at all. That's why we are using individual
measurements instead of a single measurement for
all the cells.
Remember though that the preferred comparator
now is the LMC7225 chip, not the LM339. We're going to
have to drop the LM339 completely and im going to take
it off the schematic.
Well, from what i've seen in NiCd batteries even with
a very carefully controlled charge it is more likely
that they run down at different rates then together.
I guess the reason for this is that the cells have a
slightly different storage capacity so one runs down
before the other even though they are both charged
to the same exact level to begin with.
One additional point is that among the sets ive seen,
although one cell runs down before the other the second
cell is close behind, meaning one cell follows the other
by some minutes (such as 5 minutes). With a setup that
only detected a single voltage for two cells, this would
mean one cell would be slightly undervoltage for maybe
5 minutes or so. Unfortunately i dont have any data
that relates run time undervoltage to cell lifetime, so
you'd either have to contact a battery manufacturer or
take a slight risk when using the single measurement method.
It's entirely possible that very little damage is done
when running only a few minutes undervoltage and it might
be worth a try, but i cant tell you to do that because
i simply dont know exactly if it will damage the cells or
not, and if so, what exactly the extent of the damage is.
I think there are people out there doing this already
so there might be some merit to it.
An experiment would, of course, be easy to set up.
You would charge two cells (probably individually)
to the same level and then wire them in series.
Measuring the voltage of each cell individually, run
them down at the same rate as your flashlight would
While the voltage drops, wait for one cell to reach
the cutoff point and note how long it takes the
other cell to decrease to the point where the series
voltage is under your cutoff v times two.
For example, say you want 1.1v cutoff. This means 2.2v
for two cells. When you drain them down, one cell might
reach 1.1v while the other is 1.15v, which of course still
isnt a total voltage of 2.2v yet, but one cell is now
about to run below the cutoff point so note the time.
After a little while longer, the higher cell goes down to
maybe 1.12v while the lower cell goes down to 1.08v .
Now the series combo is down to 2.2v so note the time
again. Subtracting the times will show the time that
the lower cell has been running while undervoltage.
Since 1.08v (in this case anyway) isnt very much lower
then 1.10v i wouldnt expect much damage to the cell.
On top of that, if it's only been running for 5 minutes
like that i cant see a problem with the single cell
The only other thing to consider next is what happens
once the cells begin to age after many charge cycles.
If once cell gets 'worse' then it was, then it may be
running at a low voltage for an extended period of time,
but it's hard to say just how long that would be, and
since the cells aged anyway it might not matter as much.
There is always the chance that the higher cell ages
differently then the lower cell and 'catches up' with it :-)
The choice is really yours, and what you are willing to
try out. Perhaps you can monitor the cells once every
two months or so to get an idea how well they are
tracking each other.
It should go without saying that they should both be
fully charged, and if you have good enough control
over the charge for each battery you can perhaps adjust
the charge in the higher voltage battery to make it
track the lower one a little better. You can also
put an adjustment on the voltage detect circuit
to compensate for a battery that has consistantly
higher voltage then the other(s).
Now you have an idea what this problem entails and why
we are using differential voltage measurements instead
of just a single top battery voltage measurement.
Oh that's good...maybe we'll hear from some other people
thinking about low voltage battery protection circuits
I guess there might be a thousand and one designs for
this sort of thing :-)