Ultra-low noise battery over-voltage circuitry ...

[gentlevoice]

Hi All,

I have an apparently simple need/circuitry that has been teasing me for a while now I hope one of you may be able to help with a suggestion here ..

As it is I am in need of an "over-voltage protection circuitry" for serially connected li-ion batteries. Some of the criteria for this circuitry are:

1. It must be ultra-low noise especially when there's no over-voltage on the li-ion batteries. I.e. particularly when the batteries are not charged the over-voltage circuitry must be ultra-low noise also down to DC.

2. The exact over-voltage should be adjustable (in tens of millivolts). So that appr. 30 mVs over-voltage leads to maximum current drain. Response time is not important - it can be several seconds.

3. Maximum current drain should be 200 mAs.

4. Idle current consumption (i.e. when no charge/overvoltage is present) should be as low as possible - and no more than 0.2 mAs.

5. I would prefer a simple circuitry as I will need to make some of them.

6. They need be "stackable" as more Li-ion batteries are serially connected and each of them need an over-voltage protection.

I've been experimenting with the attached circuitry which I reckon fulfill most of the requirements, however, it is unstable temperature-wise. That is - when a sufficient over-voltage is present it correctly drains current to lower the over-voltage BUT there's a thermal runaway so that it doesn't switch off again when the over-voltage is no longer present. Different transistors lead to the same result, unfortunately (2N3904/3906; MPSA18/BC560).

I've also looked into IC-based solutions but have not found an IC that fulfills the requirements and does not need either programming, is not overly complex, or costly.

So, if one of you may have a suggestion as to how the attached circuit may be amended to be stable temperature-wise - or have a suggestion for a different circuitry I'd appreciate your feedback.

Cheers & a Happy New Year to all

Jesper

 

[Mooly]

Would a single low power, low voltage opamp used as a comparator (or a comparator itself) be suitable ? There would be no temperature variability and you should be able to get the switching points to your desired spec using hysteresis to ensure it didn't hover in no mans land.

Haven't tried it, but that would be my starting point.

 

[gentlevoice]

Hi Mooly,

Thanks for your feedback ... ;-)

Would a single low power, low voltage opamp used as a comparator (or a comparator itself) be suitable ? There would be no temperature variability and you should be able to get the switching points to your desired spec using hysteresis to ensure it didn't hover in no mans land.

It could be usable - I've considered this myself but my main concern has been the noise level of opamps - the ones I've seen often have quite some noise at low frequencies ... Yet maybe you have a suggestion?

BTW I like you Avatar

Cheers,

Jesper

 

[Mooly]

Thanks

I can't see any opamp used as a comparator modulating the supply with noise tbh. Same goes for when it flips states and loads the cell with your 20 ohm load.

 

[Osvaldo de Banfield]

There are lots of tiny very low power drain opamps and comparators, mainly CMOS or BicMOS. Search in the LT, or ST pages. I can't see the problem with the disturbations in the power lines as you maintain capacitive load in the opamp/comparator circuits as low as you can, and still, a small RC filter in power bus will suffice.

Happy and prosperous new year to you too.

 

[gentlevoice]

Hi again Mooly,

I can't see any opamp used as a comparator modulating the supply with noise tbh.

Hmmm... no, you may be right. Worth considering ... also searching for a low-power opamp I came across some that have a shutdown pin - which would altogether solve the issue with low current drain when not charging.

I think I have enough ideas now to move on so no need to reply further.

Thanks again Mooly ;-)

Jesper

 

[gentlevoice]

Hi Osvaldo,

I think our posts just crossed eachother - as I write in my reply to Mooly reckon I now have the ideas/info I need - which also seems to be confirmed by you.

Thanks for replying and for the New Year greeting ;-)

Jesper

 

[Mooly]

Good luck

 

[Elvee]

Hi All,

I have an apparently simple need/circuitry that has been teasing me for a while now I hope one of you may be able to help with a suggestion here ..

A TL431 shunt regulator very easily fits the bill: here is an application example.
The sim shows the behavior for voltages from 3.55V to 3.65V (the limit is set at 3.6V).

On its own, the TL431 can handle 100mA, but this can easily be increased to any arbitrary value with the help of an external PNP.

Under quiescent conditions, the consumption of the whole circuit is <1mA, and no noise current can be injected (there are lower power versions if 1mA is too much).

The limit is very tight, stable, and easily adjustable.

If required, the circuit can be fitted with an indicator, and it could also be slown down if necessary

 

[Osvaldo de Banfield]

I'm wondering which levels of voltages do you have available to power the comparator, the level of the signal(s) and its frequency(ies) to be measured, and the current do yo expect to waste in it, how to see if I can offer or suggest a possible candidate.

 

[Elvee]

(there are lower power versions if 1mA is too much).

The TLV431 is an example of low power version: with it the quiescent current can be down to 0.1~0.15mA.

One just needs to scale the circuit's resistors by a factor ~10, and also adapt the potential divider, since the reference voltage is 1.25V instead of 2.5V

 

[gentlevoice]

Hi again - appreciate your helpfulness here ;-)

@Elvee: Thanks a lot for drawing up a circuitry and outlining the solution you suggest. As it is I've actually already tried a TL431 (did some measurements), however, it seems to have some low frequency noise that I really would like to avoid in this application. And since the TL431 doesn't switch off entirely I reckon the noise will still be there to some extent through the 680 ohm resistor even if the cathode goes relatively low and the PNP is in an off state ...

Elvee ... unless you have some other noise data related to when the TL431 goes low I will prefer to go will the opamp with hysteresis option. In either case - many thanks for considering and drawing ;-)

@Osvaldo: In this case the li-ion cell voltage will be appr. 4.05 volts. This voltage also supplies the opamp with operating voltage.

Cheers to you ;-)

Jesper

 

[trobbins]

Gentlevoice, what is the reason for the "low noise" aspect of your BMS ? Is it just that some BMS loads may use a duty cycle controlled load (such as a fixed resistor), which may then cause an ever so slight voltage disturbance ?

Aren't there any commercial BMS modules that do what you want?

 

[gentlevoice]

@Elvee: Just noticed that our posts also crossed. The TLV431 looks interesting, yet again, if the low frequency noise is still there when the '431 goes "off" it won't be feasible ...

Ahhhh, electronics

Jesper

 

[Osvaldo de Banfield]

Which is the noise voltage target? Why your design is so critical in this aspect?

 

[gentlevoice]

Hi Osvaldo & trobbins ... in reply to your question about the noise voltage I would like to achieve the full benefit of using batteries which is a noise level of minus many dBs. When charging the battery it's less important but when feeding the circuitry from the batteries alone I'd like no other disturbances ...

Cheers,

Jesper

 

[trobbins]

A normal BMS would not do cell balance loading whilst there is an external load on. So that would just leave your connected over/under voltage circuitry, which wouldn't operate in any non-linear sense until under-voltage occured.

I suggest that the parallel loading of the o/u sensing circuitry would be under your noise floor. What equipment do you have for noise floor measurement, and what does your battery and connected equipment measure (without any u/o circuitry)?

 

[Elvee]

@Elvee: Just noticed that our posts also crossed. The TLV431 looks interesting, yet again, if the low frequency noise is still there when the '431 goes "off" it won't be feasible ...

The off-state current is 100nA max.

Let's imagine a wildly pessimistic situation, where the noise modulation of that current is only 40dB down.

This leaves you with 1nA noise, very, VERY worst case, which you can easily filter down to 10pA with a simple RC within the detector, since you're not interested in speed.

Do you really think you are going to notice such a noise level?

And what means (setup, instruments) did you actually use to measure the quiescent current's noise in the first place?

 

[peufeu]

When drawing current from the batteries, their voltage will quickly fall from the fully charged voltage (usually 4.2V) to a lower value. Even if it only drops a few mV, this will be enough to fully turn off the shunt regulator, leaving only the TL431 bias current, which is small and constant. So, no problem.

If you are really paranoid, you can add a low-RdsON MOSFET in series with the shunt reg, and turn it on only when charging. This has the added benefit of bringing down the idle current drain to near zero when the system is not in use.

 

[gentlevoice]

Hi all,

Thanks for considering & replying. About battery noise measurements the diyaudio user Gerhard made some measurements on various batteries - a link is here:

http://www.hoffmann-hochfrequenz.de/downloads/NoiseMeasurementsOnChemicalBatteries.pdf

I have reason to believe that the batteries I use are at least as good noise-wise as the lowest noise versions he measured.

The equipment I use to measure noise is an AD7760 EVB which has a shorted noise-level of about - 153 dB relative to appr. 3.4 volts. In practice it is not perfect though as in my setup for some reason there is some low frequency noise when measuring. The actual setup can be seen here:

http://www.diyaudio.com/forums/powe...listened-li-ion-batteries-13.html#post4117303

However, even when drawing up to 0.8A from the batteries in question there's no discernible difference between when the batteries are connected to the AD7760 and the AD7760 input is shorted. I will draw a maximum of 11 mA in this case.

@Elvee:

Do you really think you are going to notice such a noise level?

I've built a DAC (DDDAC variant) which basically is fully battery powered. The individual batteries are charged from very low noise PSUs. When I switch off the PSUs (only an NPN emitter connected to each battery) it is actually quite easily discernible that the PSUs have been switched off. More quietness, calmer, more detailed and immediate sound (to my ears) without the PSU.

@peufeu:

When drawing current from the batteries, their voltage will quickly fall from the fully charged voltage (usually 4.2V) to a lower value.

Just FYI: According to batteryuniversity.com one of the factors that greatly reduce Li-ion battery life is high charge voltages. Another diyaudio member (can't remember who right now) once wrote to me that he had good experiences float charging li-ions at 4.05 volts.

Noise in the "off-state" of the TL431: According to figure 12 on p.10 in this TLV431 datasheet ( http://www.ti.com/lit/ds/symlink/tlv431.pdf ) the noise of the TLV431 at 1 mA cathode current is about 275 nV/sqrt Hz at 10 Hz. I can't find a similar curve for the TL431 but assume/reckon that it would be in the same region. And 1 mA is close to the off state (as specified in the datasheet)

you can add a low-RdsON MOSFET in series with the shunt reg, and turn it on only when charging.

... Thanks for the idea. I also have thought about it myself but got away from it again. Will add a bit of complexity though ...

Although - as I think about it now - I don't see the TL431 as a suitable candidate I would actually prefer using it (I have some already) if the very-low-frequency noise can be reduced. Any ideas here ... (will also ponder myself) ... ? Otherwise I think the opamp with a filtered voltage reference may be the most feasible solution.

Cheers,

Jesper