Measure current over time to calculate battery lifetime?

Hello,

We have a product which uses a telit module w/ cellular and GPS and we are trying to tweak the software to get longer battery life out of it. However, before we can really do anything we must be able to accurately measure the current so that we can see how our changes affect things.

The simple way is to jut charge up some batteries and put them in use and see when we get the low battery alert. This just takes too longer. I'm thinking about creating a test "jig" where I could supply the power to the devices and then measure the current using a micro controller. This would be logged to a computer so I can calculate the current drawn from the battery over say a 1 hour period and then determine how long the device would last with those settings. We could also correlate these logs with the system logs to determine current measurements in different modes of operation.

Was thinking about using an arduino and then a TI chip which measures current and voltage via i2c, the INA219... It uses a shunt resistor to measure the current. I was going to use a .1 ohm 1% resistor.

Any suggestion or alternatives you can think of to help me accomplish this?

Might I recommend simply using an instrumentation / differential amp with gain, such as the TI INA196, along with the proper sense resistor? Then you can just use an ADC input on whatever platform you're using (Arduino, etc.) It might take less effort than working with the I2C interface.

As for the resistor, you want to determine what your full scale current range to measure will be. For instance, if you will never measure more than 100mA, and you know that the gain of the INA amp is x20, you would configure the sense resistor close to 1.65ohm so that at full scale measurement of 3.3V, the current being read is 100mA.

1.65ohm * 0.1A = 0.165V x 20 = 3.3V

Or maybe you use a reference of 2.5V, etc. in which case, the resistor needs to be closer to 1.25ohm for full scale at 2.5V.

Through some combination of sense resistor, reference voltage, and range, you can get the results you need. Then in software it's just a matter of simple math to convert back to floating point voltage.

Does this help?

Scott,

Thanks for your reply. The op-amp route was actually my first thought, but then I found that chip and thought it might be the better way to go. It will do sample averaging for you, has calibration registers to take care of the scaling and then you just read current (and voltage) out of it. This removes the need for accurate reference voltages, etc. I figured since it was just some bench equipment (not production) I could likely get away with the $3 a chip.

I do need to think more about the shunt resistance.. at .1Ohm @ 2A will only give me 20mv... I might want a bit more then that. I'll go back and read the data sheet I think they said try and set it up for between 40 and 60mV at full scale.

-jr

For the resistor it is a trade-off between what you can sacrifice in terms of voltage drop to the system and range / precision of the measurement.

As for keeping the voltage drop low, and aside from using the INA219, some MCUs have the ability to introduce gain into ADC inputs. I know Cypress PSoC devices will do this, and some Atmels (which the Arduino is based on, don't know if you have that flexibility on the platform.)

The best advice I can give is to experiment with it. You can always change the sense resistor if you don't get the results you desire.

Thanks. I'm going to stick with the .1 right now which gives me 20mV @ 2A full scale.. I can just series .1 resistors to play with it and zero in on a value which works for the device and range I need. I think the only thing I will loose is some accuracy since each resistor is 1% +/-.. Once I dial it in I'll source the correct precision resistor..

I may actually want to do something higher like .5 ohm yielding 50mA @ 1amp and just let it clip. 99% of the time is going to likely be spent

How about a Hall Effect Current Sensor, such as:

http://www.gmw.com/magnetic_sensors/sentron/csa/CSA-1.html

Why use a microcontroller? Why not just use a multimeter with a RS232 link to a PC to record at regular intervals? Maybe I have missed something?

That is a good point, but the bench meter I would want is much more expensive then the parts and a few hours to put it together and we would want to put a few of these together for some of our testers. My time and salary is already paid for.. capital expenses come hard in a small startup.

Any reasonably priced meter suggestions with RS232?

-jr

How about:

http://www.circuitspecialists.com/csi345.html

I can't vouch for quality of the meter or accompanying software, but the price is nice.

I use a Brymen TBM815, it's not wonderful, but more than enough for my needs and it has an Infrared
coupled RS232 interface. If you managed to get one, it is a good idea to have a good look at the PC logging software before you buy just to make it meets your requirements. The meter that Steve suggested does not look bad at all, especially for that price, but I haven't used it.

Equipment cost versus engineer's salaries is interesting, but off topic and would probably start
a very long thread. I won't go there... (Still it feels like you will be developing your tools rather than
spending the time on improving the product.)

Apologies for the formatting in my above reply - looks like I will be a good candidate to test the new preview feature...