Project: Multiple Current Shunt Measurement

Thread Starter

ralphg

Joined Sep 4, 2011
10
I have been working on this project off and on for 18 months and have decided to finally seek some help. The project goal is to accurately measure and record my power system performance parameters to a PC database.

My off grid power system consists of:
Trace SW4024 Inverter with internal 120A battery charger (I1)
80A MPPT PV charge controller (I2)
50A MPPT PV charge controller (I3)
Battery bank 24VDC 600AH

R1 is a 50mV 500A DC shunt
R2 and R3 are 100mV 200A DC shunt

Amp board: (see attached .PDF and .SCH)
U1 is a TI Quad OPA4277A opamp driven by a +15/0/-15 PS. U1(D) input shorted and not shown on drawing.

The box at the right (R10-13) represents the DATAQ DI-148U "single ended" +-10VDC ADC board inputs. USB output.

The circuit works fine on spice and on the bench. In the power shed (with real components), it's a totally different animal. I am seeing inconsistent voltage output swings at U1A when I3 increases. I am sure there are common mode, and ground loop issues in play.

This monitoring system worked consistently and accurately with I1, I2 connected, and R1 being the only shunt in the system. When I expanded the system to include I3, R2 & R3 the opamp output went crazy when I2 and I3 change.

in the real world, with a system such as this , can a single ended input ADC board even work. Will a differential input ADC solve all the problems. How do i improve/solve the grounding/common mode issues inherent with this low voltage system.

I have been lurking on this board for years and would like to thank you all for your unselfish contributions.

Ralph
 

Attachments

wayneh

Joined Sep 9, 2010
16,399
Photos of your build could be helpful. Your problem could be as simple as too much current in too narrow/long a wire, causing small voltage errors where you don't want them.
 

Crispin

Joined Jul 4, 2011
88
Not sure if it will solve your issues but I've got a similar thing going and just kept "getting it wrong" with my shunts. (biggest issue was not following advice offered here to the T and shoddy parts :( )
I ended up using a MAX4172 which is just the ticket.
http://pdfserv.maxim-ic.com/en/ds/MAX4172.pdf

It's a high-side shunt amplifier and made easy work of using the shunt. Well worth the £1.something I paid for them.
 

mcasale

Joined Jul 18, 2011
210
I have worked on a similar measuring system. You really need to go through all the pieces of your "error budget". As mentioned, IR drops can cause a serious error. It looks like the common-mode inputs on the op amp should be okay. Also, the errors due to the offset voltage shouldn't be too big.

The biggest suggestion I can think of is to change the amplifiers to a difference amp with resistor dividers on the + and - inputs. This should convert the differential signal to single-ended, and the DataQ box should be able to handle it.

Also, how many bits of resolution is the DataQ box, and how many do you need? Make sure all the GROUNDs (battery, power supplies, and DataQ) are connected together at ONE place. Otherwise, you get ground loops and IR drops.
 

Thread Starter

ralphg

Joined Sep 4, 2011
10
Gentlemen, thank you for the suggestions. Images will be uploaded soon
If at all possible I would like to keep the shunts on the low side. Over the years I have seen a few electricians carelessly touch a screwdriver in the wrong place with a 480V surprise.

I will try the differential amp setup. The ADC is 10bit, and the opamp input offset was well below the resoultion of the board, when the system was functioning, and in good calibration.

I have and will try a single Instrument Amp with a higher CMRR.

Thank you
 

Thread Starter

ralphg

Joined Sep 4, 2011
10
Images and updated Sch pdf attached.

I calculated the R2 I2 wire resistance at .0014 Ohm, and R3 I3 at .00427, both excluding the shunts. mcasale, is this what you are referring to by "error budget". Could you please elaborate.

thatoneguy, I am not thinking of starting over, just yet! lol. I explored the hall effect option in great detail. 1st negative was cost. 2nd was the high energy consumption, and last, the only sensor I could find that would fit over a 0000 wire, in the correct range, is out of production. I will be using a hall ACS758x200B for the AC main sometime in the future. I have blown the PS energy budget tho. Got tired of frying the cheap SMPS DCDC converters, so I built a bullet proof dual linear PS.

In the last day I have disconnect the inputs from R2,R3, and using two 9V batts (+9/0/-9) to eliminate possible supply issues. With connections as shown on the sch, I2, I3 disconnected, I1 charging through R1 at 16 amps, the input at U1A is 1.6mV, U1A output is 360mV which is about right on.

NOW, with I1 at -40A, I2 at +40A, I3 at +16A, R1 is at +16A, U1A input is +1.6mV. Now the output at U1A is 920mV. ????

I shut off I3, reduced the inverter load to maintain R1 16A charge on the batts. I2 is +32A, U1A 1.6mV input and 510mV output. ???

I2 off, I3 on at +31A, R1 still at 16A, U1A 1.6mV input and 840mV output.

I2 and I3 are obviously pushing U1A output higher, I3 much more than I2.

I'm now all eyes and ears.

Thank you all!
 

Thread Starter

ralphg

Joined Sep 4, 2011
10
For those interested, the problem was solved by replacing the op amps with TI INA103 Instrument Amplifiers. The errors were caused by common mode voltages on the ground plane which the INA103 cancels out. Now the accuracy is repeatable to within .5% throughout the limits of operation. What little errors there are now appear to be temperature related that may be solved sometime in the next decade using LM50's.
Ralph
 
Top