Battery current analysis

[calexanian]

I have been looking for a more automated system for monitoring battery current draw for our products. Here is a typical situation, most likely common these days. Many of our designs are typical. Most of the time the micro running off of 2 or 3 AA cells just sits there asleep drawing a few micro amperes or so. Ever second or so we wake up, handle the timer, or any other interrupts drawing a couple of milliamp for some number of milliseconds or so, and then go back to sleep. Every now and then we wake up and perform a function, be it either run a voltage booster to do something or transmit data, what have you drawing up into the hundreds of milliamp, or even in the 1 to 2 amp range inrush current. That lasts lets say a couple seconds, and then back to sleep. Traditionally we would just mathematically integrate and break all of those various current draws into time domains, multiply all of those various domains sorted by their respective currents, and add them together to get our average current. All fine and good, when things work perfectly. Unfortunately things seldom work perfectly in new product design.

The question.

Is there a meter, or SMU, or something designed to perform this function?

The problems as I see them.

How does this device overcome the range switching of the currents? We are concerned with micro, milli, and up to over an amp, all in the same measurement circuit. Burden voltage? That is critical in these low voltage circuits. we cant just use one current shunt for all ranges because most meters do not have the resolution to cover all of these ranges at once.

Speed. We need sub millisecond resolution to catch all of the varying currents jumping basically from range to range.

Time/Average management. This all needs to be scaled to a time base that makes sense for the application.


Does anybody know of a ready made product for this kind of application or should I just start designing my own kind of special SMU type device for doing this?

Thanks.

[leppie]

This recent thread might be informative: https://www.eevblog.com/forum/testgear/measuring-current-on-a-low-power-iot-device/

[robrenz]

Possibly this?

[calexanian]

I read that thread before. Still asking the question if it exists.

Also i looked into those keysight units, well when they were Agilent anyways. Not really what we are looking for.

[robrenz]

Ok, how about this

Data sheet for above

[calexanian]

Looked at those. Not high enough resolution on the low end. I called a design meeting after my last post with two of my guys (Its good to be a product manager) and outlined how I might see a product like this going together. We think we can develop a system for this use. Might be a good product.

[Gribo]

Assuming a single range, from 10uA to 1A range, you need ~18 bits of dynamic range, that is quite a tall order even for bench DMMs. The current high end DMMs from Keithley and Keysight can perform such measurements, for quite a steep price.

[Zero999]

Assuming a single range, from 10uA to 1A range, you need ~18 bits of dynamic range, that is quite a tall order even for bench DMMs. The current high end DMMs from Keithley and Keysight can perform such measurements, for quite a steep price.

There are ways round that such as an amplifier with a programmable gain so you can change ranges.

[calexanian]

We have already drawn up a preliminary implementation for sub millisecond range switching. It was so easy we are going to have to do a patent search to see if somebody else is already doing it. Once you make the declaration this product is specifically designed for modern battery operated personal/small electronics a whole world of options opens up. I already told the owner that I think we can do it and nobody else has it. We discussed if we want this for just our own use or as a salable product. If we just make it for in house we might decide to open source it, but if we have something really good we may think about making a product. Either way i am kinda excited. I have been working with nothing but our core product line for over a decade here and the chance to make a piece precision test gear sounds like fun.

[nctnico]

As long as you are interested in average current then using an integrator should be doable. Battery fuel gauges (Coulomb counters) do exactly that. Reaching over 100dB dynamic range in an analog circuit is definitely possible.

[Zero999]

We have already drawn up a preliminary implementation for sub millisecond range switching. It was so easy we are going to have to do a patent search to see if somebody else is already doing it.

Be very careful. At the moment, if it's patented and you get caught for infringement, it's not so bad because you can claim ignorance to some degree. If you do a search and it turns out it's patented and hasn't expired, you can get done for knowingly violating a patent.

Even if it hasn't been patented and you patent it, it doesn't mean it will hold up in court. Someone else could have done it before but not patented it and claim prior art.

Disclaimer: I'm not a lawyer.

[calexanian]

We have already drawn up a preliminary implementation for sub millisecond range switching. It was so easy we are going to have to do a patent search to see if somebody else is already doing it.

Be very careful. At the moment, if it's patented and you get caught for infringement, it's not so bad because you can claim ignorance to some degree. If you do a search and it turns out it's patented and hasn't expired, you can get done for knowingly violating a patent.

Even if it hasn't been patented and you patent it, it doesn't mean it will hold up in court. Someone else could have done it before but not patented it and claim prior art.

Disclaimer: I'm not a lawyer.

Ugh... Patents.... Insert nightmares here.... I have ben on both sides of that stick. Both sides suck.

[dannyf]

Agilent has a patent on this. Google seamless current measurement and the guy behind it.

[calexanian]

No. Its this patent actually 8115474.  They are not implementing it right though. This patent could easily be circumnavigated, and, with the new laws is probably not enforceable, but thats a matter of who has more money to throw at lawyers. That is a game nobody wins.  Typical big company. Patent something so nobody else can do it, but don't do it right yourself. Oh well. Can't stop me from making one for myself.

[electronic_eel]

How are you planning to get the microsecond resolution? When you switch on a stepup regulator, you'll typically see very short and sharp current spikes when the output capacitor is switched on. If your time resoultion is not fine enough, you'll miss most of this spike.

Just another idea:

integrate the measurement circuit with a power supply. The power supply can have a constant, but rather large current measurement shunt which can be designed to allow measurements down into the nanoamps. The power supply regulation can compensate the voltage drop on the large shunt.

Then have multiple opamps and ad converters sampling the voltage all the time. The opamps have different amplifications and implement the different ranges. Because you have several ad converters sampling all the time you don't lose any measurements during range switching.
You can then stitch together the results in the digital domain.

[calexanian]

It's actually pretty easy. So easy Agilent figured it out in 2007!  Single or multiple shunts, multiple front ends feeding individual a/d converters. Each one representing an individual range. Software looks at each value from them and determines the current draw as fast as whatever system can analyze it. Pretty dumb simple. 3 12 bit DAC's can cover sub microamps to amps with no analog switching. Couple that with a SMU style power supply that ensures proper supply voltage across the DUT to handle the burden voltage in that range. It becomes a surprisingly hefty supply for even a modest DUT and some firmware to generate your metrics and there you go. A $10,000 Agilent meter with a $2,500 module worth of test gear. Parented of course, because any competent design engineer could spit that out in a few weeks, and they can't have that now can they?

[electronic_eel]

It's actually pretty easy. So easy Agilent figured it out in 2007!  Single or multiple shunts, multiple front ends feeding individual a/d converters. Each one representing an individual range. Software looks at each value from them and determines the current draw as fast as whatever system can analyze it. Pretty dumb simple. 3 12 bit DAC's can cover sub microamps to amps with no analog switching. Couple that with a SMU style power supply that ensures proper supply voltage across the DUT to handle the burden voltage in that range.

You wrote that they are not implementing it right. In what product and what is the problem?

[zapta]

OP, have had similar need as you described (measuring and optimizing battery life time over varying conditions).  The Monsoon power monitor (~700) does a very good job at it (both controlling the voltage and monitoring and reporting the current consumption to an attach computer).

https://www.msoon.com/LabEquipment/PowerMonitor/

For a more user friendly solution check Power Play on github (a OSH project). It uses a coulomb counter for accurate current integration, has a small OLED display and onboard analysis with a very simple operation (a single push button). It can also report the raw data to a computer over USB/Serial for logging and further analysis. It's is also compatible with the Arduino IDE and easy to hack for your specific needs.

https://github.com/zapta/power-monitors/tree/master/pmon_3v8

[calexanian]

It's actually pretty easy. So easy Agilent figured it out in 2007!  Single or multiple shunts, multiple front ends feeding individual a/d converters. Each one representing an individual range. Software looks at each value from them and determines the current draw as fast as whatever system can analyze it. Pretty dumb simple. 3 12 bit DAC's can cover sub microamps to amps with no analog switching. Couple that with a SMU style power supply that ensures proper supply voltage across the DUT to handle the burden voltage in that range.

You wrote that they are not implementing it right. In what product and what is the problem?

The way their product works is you set your time increments for your samples and let them run until you are out of memory. Then you can integrate and average. This is ok for doing short term testing. It only has so much (Very little) memory and when trying to catch things over long periods of time is hard. Honestly I do not believe I can properly explain without being in person with a white board how these tests should actually be performed.