Highside MilliCurrent

[sourcecharge]

Hi all,

I know, uCurrent rules, but for those that use bipolar power supplies and bridge driver (half or full), highside current measurements are required.

This is what I came up with for each PS output:

I made a pcb through hole version using 1k ohm precision temperature stablized metal film resistors (measured to < 0.01%) that I had on stock. 

There are 1k 0.01% digikey surface mountable resistors for about 2 bucks a piece. (notice the extra 10x precision compared to the uC)

The total number of 1k 0.01% resistors need to make 2 Highside MilliCurrents would be 22.
https://www.digikey.com/en/products/detail/stackpole-electronics-inc/RNCF0805TKY1K00/2269700

Even if you get those, you still need extra non precision, but temperature stablized resistors and a couple of 18 turn through hole pots for the output stage. (my design)

The common mode voltage seems to be the catch for highside measurements. 

Postive non inverting bias current can be zeroed out by using the relative function, but a Negative non inverting bias current is opposite. 

The negative non inverting bias current offset voltage can be relatived out if you turn on the oscilations without the power to the input lead follower transistor connected to the highside mC while the PS still connected to the mC.

(The 11 ohms on the input is from the shunt resistance from the current measurement from the mastech multimeters I have.  Also the PS on the output voltage meter is the relative function of the multimeter set at 8.8uV)

This gain mulitplication per stage variation is the lowest in offset voltage in B2spice in order to maximize the bandwidth.

[jbb]

A good thing to have on your bench!

I built something similar a few years ago. I used it for measuring the run and sleep mode power consumption of a microcontroller.

Instead of buying precision resistors, I decided to use an all-in-one INstrumentation Amplifier (INA)… I forget which one exactly. These all in one devices tend to have good internal resistor matching and therefore really good Common Mode Rejection Ratio (CMRR), at least at low frequencies.

The device produces 2 outputs. Firstly, a coax connection for the INA output - essential for viewing waveforms on a scope. Secondly, the INA output also goes into a low pass filter (1Hz ) for readout on a multimeter.

Power was initially provided by two 9V batteries (+-9V supply), but I later switched to a stack of CR2450 coin cells which I had handy.

[sourcecharge]

A good thing to have on your bench!

I built something similar a few years ago. I used it for measuring the run and sleep mode power consumption of a microcontroller.

Instead of buying precision resistors, I decided to use an all-in-one INstrumentation Amplifier (INA)… I forget which one exactly. These all in one devices tend to have good internal resistor matching and therefore really good Common Mode Rejection Ratio (CMRR), at least at low frequencies.

The device produces 2 outputs. Firstly, a coax connection for the INA output - essential for viewing waveforms on a scope. Secondly, the INA output also goes into a low pass filter (1Hz ) for readout on a multimeter.

Power was initially provided by two 9V batteries (+-9V supply), but I later switched to a stack of CR2450 coin cells which I had handy.

I researched INAs and didn't find even one that had a slew rate above 2V/us where as the opa189 or opa2189 has a slew rate of 20V/us.  That and the opa189 has a CMRR of 168db.

On top of that, the offsets on the INAs were way higher too.  Can't beat the precision laser cut internal resistors probably if you are mass producing this, but measuring the resistors manually from a box of 200 or so (what i did), you can find 0.01% matching resistors that are like 20 bucks each normally. 

[Kleinstein]

The main idea with the 3 OP-amp INA / differential amplfier is to have most of the gain in the first stage and only relatively little (e.g. x1 or x2 ) in the 2nd stage. With that there is no more need for super accurate resistors to get good CMRR. The gain of the input stage would improve the CMRR. So one could get something like 40 dB extra with a gain of 100 in the input stage. Accuracy in the input stage would only effect the gain accuracy, not the CMRR.

p.s.
With a high side voltage and still a ground link one would likely need protection for the amplifiers, especially for the case with the voltage at the measurment side is higher than the OP-amp supply.

[sourcecharge]

The main idea with the 3 OP-amp INA / differential amplfier is to have most of the gain in the first stage and only relatively little (e.g. x1 or x2 ) in the 2nd stage. With that there is no more need for super accurate resistors to get good CMRR. The gain of the input stage would improve the CMRR. So one could get something like 40 dB extra with a gain of 100 in the input stage. Accuracy in the input stage would only effect the gain accuracy, not the CMRR.

p.s.
With a high side voltage and still a ground link one would likely need protection for the amplifiers, especially for the case with the voltage at the measurment side is higher than the OP-amp supply.

I was eye balling the INA849.  The output offset was kinda high and the 10x gain error of 0.025% was not appealing although this one had a 35V/us slew rate.  I don't remember why I passed on this one.

I did find in B2spice that putting the gain all in the 1st stage had much higher voltage offset on the output.  I also remembered that lower gain spread out across multiple stages increased the bandwidth.

I found that using a 3x,3x,1.6x had the least amount of offset, with the lowest gain spread across the 3 stages.  Also, the last stage was just a non inverting opamp, that could be adjusted with a trimmer pot.

The common mode voltage or the maximum voltage measureable is 2.5 voltage less than the PS.  I use about +/-10V max to measure so it works out for right now.

[twospoons]

So why not just use TI's INA290? Its a couple of bucks, handles up to 120V common mode, has 1MHz BW for the lower gain version, and CMRR of 140dB.  You are really going to struggle to match that using opamps and discrete resistors.
While I applaud the OP's efforts to 'roll-your-own' (not enough ppl try to do this) some things are much better done on a single chip.

[sourcecharge]

So why not just use TI's INA290? Its a couple of bucks, handles up to 120V common mode, has 1MHz BW for the lower gain version, and CMRR of 140dB.  You are really going to struggle to match that using opamps and discrete resistors.
While I applaud the OP's efforts to 'roll-your-own' (not enough ppl try to do this) some things are much better done on a single chip.

The price is nice, but its a sot-23-5 and I've sworn off using those along with the tsops.

Too small for me, I only like using soic packages as I hand solder them on.

I don't need 120V but it does seem like this would be better input range for a universal highside current meter. 

I think there are current sencing amplifiers that have high common mode voltages too but they use the input voltage to offset and power the chip.

The 168db of CMRR on the opa189 or opa2189 and the low offset is worth it though and I'm happy with what I made.

[Kleinstein]

Having more gain with the input stage should not cause more offset. A downside can be the BW and also the need for extra headroom in the supply.
At least the OPA189 is quite fast, so the speed should not be an issue.
The problem with low gain in the first stage is that the resistor matching for the 2nd stage is than limiting the CMRR.

[sourcecharge]

Having more gain with the input stage should not cause more offset. A downside can be the BW and also the need for extra headroom in the supply.
At least the OPA189 is quite fast, so the speed should not be an issue.
The problem with low gain in the first stage is that the resistor matching for the 2nd stage is than limiting the CMRR.

I can't tell you why B2spice is saying there is more offset when the first stage has all of the gain. 

Maybe you can try it on your spice of choice and check to see if it matches b2spice.

I have a box of old 70s 1k metal film temp stable precision resistors. 

I was able to measure them down to .1 ohm accuracy using a fluke 89 IV and matched them to the exact reading of the multimeter.

I may not have got exactly 3x gain on the 1st and 2nd stages but the required resistors were matched as good as I could get them.

[Kleinstein]

The resistor accuracy in the 1st stage only effects the gain accuracy. With the trimmer for the gain of the 3rd stage this is not relevant anyway. The point is that with more gain at the input the resistor matching gets less critical.

The offset seen in a simulation is only one example. Normally there should be very little there, as much of the offset comes from effects like input offset current or voltages that scatter and are naturally not included in the simulation. There are also effects that can compensate, but real life is more scattering. Normal math tells that more gain for the 1st stage should give a better chance for lower offset too, though not much difference.