Designing a Battery Monitoring Circuit: Questions & Challenges

[roeb]

I am currently in the process of designing a battery monitoring circuit. At the moment we are planning to base the circuit around the MAX4081 (bidirectional current sensor) and a microcontroller (atmega48). I am a bit concerned as to whether or not we will actually be able to get anything out of the MAX4081 that will indicate to us the actual status of the battery.

- 24 V lead acid battery
- at times it will be supplying ~10 A

The original plan was to take the current sensor and get the current, we can then integrate with respect to time to get Amp hours. However, I am not convinced that this will actually work.

Say the battery is supplying 10 A and we measure 10 A at an interval of 10 usecs. So yeah we can integrate that and get the amp seconds but does it mean anything? What if the battery is supplying 5 amps at an interval of 10 usecs? We can keep on adding up the small areas but I still fail to see how that would yield anything useful especially if the current is jumping around...

I found a nice article that goes over the basics of this, but unfortunately it still doesn't answer all of my questions... http://focus.ti.com/lit/ml/slyp086/slyp086.pdf


My questions are:

1) Can a current sensor be used for battery monitoring? I'm not going for a terribly high accuracy, but something that is functional would be nice.

2) How do I extract a charging/discharging curve from the current?

I can use the microcontroller just fine for ADC from the current sensor but I am still concerned that we won't be able to get anything useful out of the measured currents...

 

[mgb_phys]

That sounds like an ideal way to measure a battery charge state.
You aren't making any assumptions about the battery (it's volt/amp/capacity curves) you are simply doing current in - current out book-keeping.
How often you sample the current is up to you and depends on how frequently the output current changes in your application, there is a limit to how often it can change based on the resistance and capcitance of the circuit, there is no point sampling faster than this - a sort of Nyquist limit if you like.
Main problem will be noise in the current sensor andthe limits of the cheap ADC in the uC.
You probably want to look at some statistical smoothing technique, either somethign simple like a running average or more complex like a Kalman filter.
Another good feature might be a reset to the integrator when you know the battery is full - perhaps from an external voltage sensor or from measuring a predicted top up current value.