Which is the most accurate method to measure current?

[Flyback]

Hello,

I am required to measure current which can be up to 10 Amps.
Which is the most accurate of the following three methods?….

…one is a non-inverting amplifier, and the other is a differential amplifier.

Non inverting amplifier (Gain 101)
https://i44.tinypic.com/2z9d2f7.jpg

Differential amplifier (inverting) (Gain 100)
https://i42.tinypic.com/e25fo.jpg

Differential amplifier (non-inverting) (Gain 100)
https://i40.tinypic.com/24me3vm.jpg


The opamps will feed into micro ADC's

Here are the schematics in pdf:

 

[ronsimpson]

With a resistor of 0.001 ohms it is easy to have errors due to current passing through wire. A diff amp is looking across the resistor and not at the voltage drop in the wires. The amp can be located at a distance.

With the single ended amps the amp also sees any voltage between the resistor and the amp. If the amp is at the resistor then no problem.

 

[steveB]

You can't use the inverting configuration because you have a single supply topology and inversion would drive the output negative. The two noninverting configurations are pretty much the same. I would think the non-inverting amp is simpler (consider what ronsimpson said about trace resistance, as that will kill you if you are not careful) as long as you know the gain is 101 and plan for that in your processing.

BTW, I think 0.001 ohms is too small to be practical.

 

[Flyback]

You can't use the inverting configuration because you have a single supply topology and inversion would drive the output negative

Thanks but the middle one is inverting, but the input voltage goes further below ground the more current that flows......so basically it is inverting, but really "non inverting", if you like.....given this, do you change your opinion?.....all the above three work, its just a case of which is most accurate?......the amp is at the resistor, i can change it to a 10mR resistor, but no more as that is 1W dissipation at 10A

The two noninverting configurations are pretty much the same

....how can they be the same?......the books (eg "how to use opamps", page 20) tell us that the simple noninverting one doesnt equalise the impedances seen from the two opamp inputs...and therefore there are more errors?
Also, the proper differential noninverting one has less common mode noise ..surely?

 

[dougy83]

They are all pretty similar; gain is higher for the non inverting setup. If you are using the lt1006 you shouldn't have much issue with any of the configurations. Its very low input offset voltage means that the high gain and low sense resistor value in your circuit should actually be ok. The low impedance to the input should mean that the offset current doesn't introduce much error even in the standard non inverting configuration.

Both diff amps should be the same. The non inverting amp will be slightly inferior (maybe100uV error)

 

[steveB]

Thanks but the middle one is inverting, but the input voltage goes further below ground the more current that flows......so basically it is inverting, but really "non inverting", if you like.....given this, do you change your opinion?.....all the above three work, its just a case of which is most accurate?......the amp is at the resistor, i can change it to a 10mR resistor, but no more as that is 1W dissipation at 10A

Yeah, sorry. I missed the fact that the current direction was different. If you have a situation where the measurement ground is not restricted relative to the power circuit, it seems that's ok. I think the non-inverting configuration would be more typical in a single supply system with (approximately) common grounds between the measurement and power circuits.

....how can they be the same?......the books (eg "how to use opamps", page 20) tell us that the simple noninverting one doesnt equalise the impedances seen from the two opamp inputs...and therefore there are more errors?
Also, the proper differential noninverting one has less common mode noise ..surely?

Strictly, I think these are correct statements, but practically I don't think you'll see any difference. Certainly the common mode voltage is very low with such a very small sense resistor tied in right off of ground.

The bigger issue for errors is going to be exactly what ronsimpson mentioned. A 1 milliOhm resistor is just sooo small that it is comparable to typical trace resistances. Hence, personally I would use the non-inverting amp with the sense resistor "bolted" right into a massive ground plane. In other words dont' use any traces on the ground side of the resistor. And, the ground of the opamp circuit should also be tied in very closely there right into the same massive ground plane. Then the input of the non-inverting amp should go directly to the other lead of the sense resistor and not anywhere along the trace that connects it to the power portion of the circuitry. This is the only possible way to get any accuracy with such a small sense resistor, and I would expect that even with these precautions, the error due to drops across the ground plane (and opamp non-ideal behavior) might dominate over these other circuit topology issues.

 

[MrAl]

Hello,

The gain is exactly the same for all three configurations, there is no difference in gain. That goes for the differential, inverting, and non inverting gains. For all three circuits the gain is 100.

The input offset is fairly low, dependent on package though, but it's still higher than we might want to see. For example, even 200uv is a little high if we intend to get any accuracy with a target current of 1 amp. With 1 amp we see 0.001v in the resistor and adding 200uv we get 0.0012v, which means we already see an error of 20 percent. That may not be acceptable depending on the applications. A better choice would be a chopper stabilized op amp, but again it does depend on the application. Another choice is to go with a sense resistor of 0.010 ohms and use a gain of 10 instead of 100 and in this way minimize the error due to offset. For this new circuit, for 1 amp target current we get 10mv across the sense resistor and adding 200uv now only produces a 2 percent error..quite a bit better than before. The difference of course is we will see a 0.1v drop now at the max current of 10 amps which is considerably higher than 0.01v which we see with the 0.001 ohm resistor.

The balanced inputs are usually considered better because of the input currents.

But again, the gain of all three circuits is 100 and that includes the non inverting circuit.

 

[dougy83]

The gain is exactly the same for all three configurations, there is no difference in gain. That goes for the differential, inverting, and non inverting gains. For all three circuits the gain is 100.
...
But again, the gains of all three circuits is 100 and that includes the non inverting circuit.

I'm not sure you came to that conclusion. Gain of a non-inverting amplifier is 1 + Rf/Rr which is 101 for Rf=100k, Rr=1k
There are two differential configurations and a single non-inverting configuration.

The input offset is fairly low, dependent on package though, but it's still higher than we might want to see. For example, even 200uv is a little high if we intend to get any accuracy with a target current of 1 amp. With 1 amp we see 0.001v in the resistor and adding 200uv we get 0.0012v, which means we already see an error of 20 percent. That may not be acceptable depending on the applications.

The LT1006 (the opamp in the schematics) has a typical offset voltage of 20-30uV https://www.electro-tech-online.com/custompdfs/2013/05/1006fa.pdf. This is a 3% error at 1A

 

[MrAl]

I'm not sure you came to that conclusion. Gain of a non-inverting amplifier is 1 + Rf/Rr which is 101 for Rf=100k, Rr=1k
There are two differential configurations and a single non-inverting configuration.

The LT1006 (the opamp in the schematics) has a typical offset voltage of 20-30uV https://www.electro-tech-online.com/custompdfs/2013/05/1006fa.pdf. This is a 3% error at 1A

Hi,


Yes you are right about the inverting gain amp having a gain of 101. I must have had the same circuit load for that configuration as one of the circuits and so it was 100 as well Thanks for bring this up.

But the 20uv typical doesnt do anything for me. If we want to pretend that every amplifier always meets the 20uv spec then that means we ignore part of the data sheet...the part where they specify the MAX input offset. If you want to ignore that, that is of course up to you, but that's not what i want to do...i do not assume some spec's are to be followed closely while others are to be completely ignored, unless of course they can be disproved.

 

[KeepItSimpleStupid]

@Flyback

I think you really need to define the problem a bit better. AC current, DC current, resolution, Max voltage, accuracy, distance, interference sources, where will the sensor be placed, frequency, RMS?, permissible voltage drop, 4-wire, 2-wire, etc.

In any event, I would seriously look at the Hall effect Current sensors from Allegro. There are also a number of IC's designed to measure current on the high side via a sense amp, but without any boundaries at all, it becomes an exercise in futility.