Accurate Shunt Measurement

[opeleg]

Hello All. I am attempting to design a very accurate, low noise DAQ circuit. I will be reading off a 50mV (400A) shunt and using an ADC and a digital isolator to read the current from a DSP. Attached is a pdf of my prelimary circuit.

Things to note
1) The current is bidirectional (+/- 50mV)
2) I am using an ADUM5401 for the digital isolation. This has a built in dc-dc convertor to produce an isolated supply. I am worried about the noise this might create so I am thinking of using an external supply with an ADUM1300 (no DC-DC).

My first question is, if my signal is isolated, does it make a difference if I use an instrumentation amplifier or not? Does it make a difference if I am using a low side or high side shunt?

Secondly, I decided to go for a Max4208 for the IA. This doesn't use a standard 3 op amp design so I can have better results near 0V signals. Does anyone have experience with these?

Please let me know if you see anything I can do to improve the noise or accuracy of my circuit.
I appreciate any help that is offered, and thank you in advance.

Thanks,
Omer

 

[MikeMl]

What references the (floating?) 5Viso supply to the power circuit whose current is flowing in the shunt?

Why do you think it is necessary to "isolate" the opamp/ADC?

 

[opeleg]

We isolate our measurements because we have our system running in a harsh environment, and for the usual benefits that isolation provides, like preventing a faulty ground loop.

And also, why not isolate?
Omer

 

[MikeMl]

Cost. I was trying to find out if you thought you need the isolation because the shunt is in a "high-side" line.

Answer my other question.

 

[opeleg]

I wasn't sure how I should reference it. Should I just use a capacitor from IN- to GND?

 

[MikeMl]

I wasn't sure how I should reference it. Should I just use a capacitor from IN- to GND?

My point is that you went to a lot of trouble to create a floating 5V supply. You have the two ends of shunt, which have a tiny differential voltage between the ends, but might be tens or hundreds of volts from "ground", call that the "Line" voltage. At a minimum, you somehow have to connect the floating supply so that it straddles the Line Voltage, so as to keep the voltage at the ends of the shunt within the "common-mode-input-range" of the instrumentation amp.

For example, say your Line voltage is 100V. You need to deploy the floating 5V supply so that one end is at 102.5V, and the other is at 97.5V. This will put the voltage at the ends of the shunt near the center of the common-mode range for the instrumentation amp...

One way to do it is to create a resistive voltage divider with two equal resistors. Tie the outside ends across the 5V floating supply; tie the center tap to the LINE. I know you are going to ask what value the resistors should be. I'll give you a clue. The current that flows out of the divider tap is twice the bias current for the instrumentation amp plus the leakage current from your isolated subsytem to ground. The divider has to charge/discharge the capacitance between the floating subsystem and ground at the noise frequencies...

 

[Ubergeek63]

Hello All. I am attempting to design a very accurate, low noise DAQ circuit. I will be reading off a 50mV (400A) shunt and using an ADC and a digital isolator to read the current from a DSP. Attached is a pdf of my prelimary circuit.

Things to note
1) The current is bidirectional (+/- 50mV)
2) I am using an ADUM5401 for the digital isolation. This has a built in dc-dc convertor to produce an isolated supply. I am worried about the noise this might create so I am thinking of using an external supply with an ADUM1300 (no DC-DC).

how much accuracy do you need?

that is a VERY expensive way of doing it... why not just use an allegro current sensor?

https://www.electro-tech-online.com/custompdfs/2010/02/0758.pdf

dan

 

[opeleg]

Thank you Mike for your time and effort in helping me out.
I thought what you are describing I have already done with my 2.5V reference going into the REF pin of the IA op amp.

Also I am still confused, even after a bit of effort, in understanding your following comment:
"The divider has to charge/discharge the capacitance between the floating subsystem and ground at the noise frequencies"

Thanks,
Omer

 

[opeleg]

I'm also posting another configuration that uses a dual supply. I was thinking of trying both, since i'm not sure which one will provide better results.

 

[MrAl]

Hi opeleg,


I dont know if you missed Ubergeeks post or not, but he/she suggested that you check out the Allegro line of current sensors.
They sense 200 amps with a hall effect device and provide isolation all in one package for about 5 dollars USD.
I dont know what kind of accuracy you are looking for though.

 

[MikeMl]

...
I thought what you are describing I have already done with my 2.5V reference going into the REF pin of the IA op amp.
...

Applying the floating 2.5V to the REF input does not guarantee that the inputs of the 4209 are within their allowed Common-Mode range. The Common mode range allows differential sensing with the inputs down to 0.1V below Vss and up to Vdd-1.3V. You have two options. Tie floating ground to either end of the sensing resistor; or tie floating 2.5V to either end of the sensing resistor. Which end you tie to it will alter the sign of the current sense.

See if you can figure this out. If you need more discussion of this, write back.

Also, look up the ZXCT1009 family of "High Side Current Monitors"

 

[MrAl]

Hi,


Yes the Zetex line isnt too bad, i have used a few in the past for various things that usually require some sort of current limiting.
The accuracy may or may not come into question though depending on the application.

 

[opeleg]

Sorry I missed your post Ubergeek. I have looked at hall effect sensors, but they don't provide the accuracy i need (16 bit). I don't think its a very expensive method. Accurate hall effect sensors can be very expensive as all.

Mike, I understand the common mode voltage range of the IA. Also, for some reason, Zetex doesn't state very good accuracy on their datasheets (1-2%)
So you don't see a problem in the way I tied the floating ground in my last schematic?

Thanks,
Omer

 

[Ubergeek63]

Sorry I missed your post Ubergeek. I have looked at hall effect sensors, but they don't provide the accuracy i need (16 bit). I don't think its a very expensive method. Accurate hall effect sensors can be very expensive as all.

Mike, I understand the common mode voltage range of the IA. Also, for some reason, Zetex doesn't state very good accuracy on their datasheets (1-2%)
So you don't see a problem in the way I tied the floating ground in my last schematic?

Thanks,
Omer

I have no clue why you would need 16 bits but i doubt you will get it. 250 micro ohm at 10 watts does not tend to be a very temperature stable resistor. to get that kind of accuracy you will need at least to measure the resistor temperature as well and compensate for the temp characteristics of the resistor.

dan

 

[MikeMl]

...
Mike, I understand the common mode voltage range of the IA. ...
So you don't see a problem in the way I tied the floating ground in my last schematic?

You HAVE NOT solved the COMMON-MODE problem! You have created a "floating subsystem", which if you measured its capacitance to earth ground would be on the order of a hundred pF (or more). The only thing that can CHARGE that capacitance is the input bias current (few pA) of the Inst Amp. Calculate the time constant of 100s pF and few pA.

In order for the Inst. amp to correctly amplify the differential voltage across the shunt, the floating supply HAS to BE AT the SAME POTENTIAL as the line the shunt is in. That requires a direct connection between the floating supply and the line! All the REF input does is center the output of the InstAmp into the middle of your ADC range.

 

[Ubergeek63]

In order for the Inst. amp to correctly amplify the differential voltage across the shunt, the floating supply HAS to BE AT the SAME POTENTIAL as the line the shunt is in. That requires a direct connection between the floating supply and the line! All the REF input does is center the output of the InstAmp into the middle of your ADC range.

don't forget input voltage ratings... the specified amp loses over a volt and a half at both rails on the input. technically it is not needed with newer opamps' performance. you can use a $2 part and get good drift and accuracy if you do not mind feeding in a gain constant.

dan

 

[opeleg]

Ubergeek: I am not worried about temperature too much. First of all I am using a very large shunt that doesn't heat up very much, and second of all I believe I can calibrate for temperature inconsistencies. I am more concerned about fluctuations due to a low measurement signal with noise.

Mike. I just wanted to make sure you are referring to the LAST schematic I posted that has a capacitor tied to each line. I'm only asking because you mentioned the REF input again and I understood what you said about the 2.5V not helping and thats why I reposted the schematic. Also, how did you get to 100pF of capacitance to earth ground from the floating subsystem? I will repost my changes soon.

Thanks again guys I really appreciate your help.
Omer

 

[opeleg]

I am thinking of going with another opamp (Maxim 4208). Its an indirect current feedback IA as opposed to a three op amp design which doesn't do well at the rails like you mentioned.

I posted my latest schematic. I also added a butterworth filter after the IA. What do you think? The cutoff frequency is 600hz for the filter.

 

[MikeMl]

Ok, you are now centering the differential voltage in the common-mode range. The anti-aliasing LPF is a good idea. The ~100pf comes from all of the incidental capacitance between the floating subsystem and the non-isolated side stemming from the capacitance between the primary and secondary winding of the transformer used to create the floating power supply, the capacitance across the ADUM1401, and all of the strays... You have now effectively shunted that ~100pF with 2.5K (your 5Ks in parallel).

 

[opeleg]

I have finished the pcb for this design. Thanks for all your help. I will post updates on the results when I have something working.

Thanks again,
Omer