Nano to Amp, current measurment

[ampdown]

Hi guys!

I am doing my bachelors project. And I am building a current measurement device where you have digital feedback from embedded systems. Hard to explain in short, but if and when I finish I will share details. It is simular to Dave's uCurrent, but with additional information when you are optimizing your firmware.

I am measuring current using multiple shunt resistors, which are multiplexed with ADG1604(analog mux), to achieve low burden voltage. Then the voltage drop on shunt resistors is amplified using an instrumental amp INA110(very good in many parameters). The device is like a test instrument, where it is automatically controled with MCU.

The current range is from few nA(5-10nA) up to an Amp. I know for high current region, I need additional switching shunt, cause the drop is high on analog MUX's. The maximum voltage for the device which is measured is 12V.

Do you guy have any thoughts is there a better way to amplify or measure the current? Your feedback will help for my theory part and over all device performance.
I have heard about transimpedence amp's, but I am not sure they would work on this large amp range.

[David Hess]

How much dynamic range and accuracy are you expecting in each range?  High current shunts suffer from poor accuracy do to their relatively high temperature coefficient.

Transimpedance amplifiers are useful where the lowest possible, basically zero, burden voltage is required.  At very low currents, an integrator is better than a transimpedance amplifier.

You mention the INA110 so does this have to be a differential measurement like with high side current monitoring?

[hli]

Looks as if you want to re-do the "EnergyAware commander" functionality from EnergyMicro (now part of SiLabs). You can look at the schematics for the EFM32 series of boards, which can measure  from low µA (or high nA) up to mA. (And then correlate current consumption to the line of code currently executing).
For switching the shunts you can use low-resistance FETs. I remember that I saw a auto-range version of the original µcurrent somewhere which does this (Google yields nothing, but I have the schematic printed out somewhere).
One problem to be aware of: when you do low-current measurements and decoupling caps in the DUT will screw up with your timing. The current you measure is not exactly what is used by the circuitry you want to measure.

[timb]

Looks as if you want to re-do the "EnergyAware commander" functionality from EnergyMicro (now part of SiLabs). You can look at the schematics for the EFM32 series of boards, which can measure  from low µA (or high nA) up to mA.

TI has similar functionality on their MSP430 LaunchPad boards; they call it EnergyTrace+.

The way they do it is actually pretty clever! Basically they use a small inexpensive MSP430, a MOSFET and capacitor as a charge transfer device. Essentially, the switch opens for X milliseconds, transferring a known amount of charge into the output capacitor, then closes. This is logged. A comparator monitors the voltage on the output capacitor and when it goes below a certain point another charge cycle happens.

Essentially, because they know how the amount charge transferred in each burst, they can figure out the current consumption by calculating the time between each cycle. It works from nA up to hundreds of mA. It's basically a columb counter. It links up with the onboard debugger and will show you, on a graph, the energy usage related to various parts of your code during execution. The whole thing is pretty slick and just works.

[David Hess]

TI has similar functionality on their MSP430 LaunchPad boards; they call it EnergyTrace+.

The way they do it is actually pretty clever! Basically they use a small inexpensive MSP430, a MOSFET and capacitor as a charge transfer device. Essentially, the switch opens for X milliseconds, transferring a known amount of charge into the output capacitor, then closes. This is logged. A comparator monitors the voltage on the output capacitor and when it goes below a certain point another charge cycle happens.

Essentially, because they know how the amount charge transferred in each burst, they can figure out the current consumption by calculating the time between each cycle. It works from nA up to hundreds of mA. It's basically a columb counter. It links up with the onboard debugger and will show you, on a graph, the energy usage related to various parts of your code during execution. The whole thing is pretty slick and just works.

TI managed to patent it also which pisses me off.  The same idea of using charge balancing to monitor supply current from a switching power supply was published 20+ years ago in ED, EDN, or some other electronics trade magazine.  While theoretically that makes the idea ineligible to be patented, the patent office only searches limited publications for prior art.  TI of course did not bother to list the previously published article as prior art.

Charge balancing has the virtue of having very wide dynamic range capability and the frequency output is very digital friendly.

[timb]

TI has similar functionality on their MSP430 LaunchPad boards; they call it EnergyTrace+.

The way they do it is actually pretty clever! Basically they use a small inexpensive MSP430, a MOSFET and capacitor as a charge transfer device. Essentially, the switch opens for X milliseconds, transferring a known amount of charge into the output capacitor, then closes. This is logged. A comparator monitors the voltage on the output capacitor and when it goes below a certain point another charge cycle happens.

Essentially, because they know how the amount charge transferred in each burst, they can figure out the current consumption by calculating the time between each cycle. It works from nA up to hundreds of mA. It's basically a columb counter. It links up with the onboard debugger and will show you, on a graph, the energy usage related to various parts of your code during execution. The whole thing is pretty slick and just works.

TI managed to patent it also which pisses me off.  The same idea of using charge balancing to monitor supply current from a switching power supply was published 20+ years ago in ED, EDN, or some other electronics trade magazine.  While theoretically that makes the idea ineligible to be patented, the patent office only searches limited publications for prior art.  TI of course did not bother to list the previously published article as prior art.

Charge balancing has the virtue of having very wide dynamic range capability and the frequency output is very digital friendly.

Yes, I agree, it pisses me off as well. I actually independently came up with the idea about 3 years ago, even going so far as developing a prototype board. I was considering turning it into a product, however about that time TI released their first LaunchPad with EnergyTrace+ and I noticed they had a patent on it, which royally pissed me off.

During the research and design of my unit, I came across several examples of prior art, just like you said.

I have a friend who used to work for the patent office, I talked to him about getting TI's patent invalidated but he said it would take years and likely cost thousands of dollars out of my own pocket (since TI has lawyers on retainer for this sort of thing).

My other option was to just release the product and hope TI never notices, then fight it if they do. I ended up just abandoning the whole thing.

[David Hess]

In any event, I consider this problem trivial and previously solved in multiple ways unless wide dynamic range is required and in solutions like in curve tracers, multiple current shunts are used to achieve that.  Instead of a current shunt, my usual solution involves a designed in common base/gate transistor stage operating as a fast but simple transimpedance amplifier so a low impedance is provided to the powered circuit but the charge balancing solution is probably better in a general sense.

It it useful to think of this problem in different terms when monitoring power or current into a power supply decoupled circuit which I believe TI's charge balancing implementation brings up.  Think in terms of energy used during specific time intervals.  Energy per unit time is of course power.