Working on current measurements

[Rover]

After swapping out my 4 zap 25 amp amploc hall sensors with 4 shunts 50 Amp/50 mv drop and a high side current sense chip (outputs voltage 100v/v) .. that feeds into same ADC the zaps did, I have the following observations.

1. The variation in reads has toned down quite a bit (still need more time with temp/humid fluctuations but I expect the shunts won't change much)
   
2. Unfortunately, sensitivity has gone down (not something I expected, almost like there is a threshhold of 200 -300  mA ). The hall effects produced measurable voltage to the ADC at these levels.
   
3. Looks like precision has gone up, possibly accuracy, I need some different loads to test.
   
4. There is definetly variance amongst the shunts / sensor chip combos (could be my soldering and or the components), I mave to come with a factor to bring the reading accuracy up.. variance is around 1.5%.  I had to this with the halls as well.
   
   

Problem is with lower sensitivity, I don't see the drain from my typically quiesent loads that run most of the time at 200-300 mA (led lighting etc) ,the halls did . I do see their effect when engaged with higher loads (additive).

So I have a zero problem, anybody have any ideas? or should I forget about trying to measure below 200 mA using a 50 Amp shunt?

[scottsAI]

Rover,

Why do you have 4 shunts?

I expect the accuracy of the shunts to be 2%?

Gain of the amplifier within 1%, plus offset error in microvolts. If near 300 that's your problem.

Kelvin connection can play a role in accuracy. (shunt)

Mix these up, you will get different readings. (actual errors will be different on each device, thus the different readings).

R = E/I = 50 / 0.05v = 0.001 ohm shunt.

E = IR = 0.001 * 0.3a = 0.0003v with 100 gain = 0.03v Or 30mv.

Assuming ADC with 5v range: 5/.03 need 8bits min to see 1 bit, need more like 10bits.

10 bits will show you 30mv with 5/6 bits (not much, yet maybe usable)

Next section is accuracy of amplifier and ADC errors on whether you can see the bits mentioned above.

Many Microprocessor ADC have trouble reading all the way down to ground.

Lets not forget Kelvin connection; done the same for all? Properly?

Have fun,

Scott Beversdorf.

[Ungrounded Lightning Rod]

You have to build shunts right or the voltage drop in the joints swamps the voltage drop across the shunt itself.

[Flux]

I have little idea what you are doing but from my experience in industry ( long ago and probably outdated) using multiple shunts in datalogging schemes presents many problems.

As pointed out by others, shunts are 4 terminal devices ( Kelvin Connection) and you need totally isolated inputs for the logger channels. It is virtually impossible to deal with this with a non isolated set up . Perhaps modern inputs can cope with the common mode voltages but they will probably be larger than the signals.

You may be able to get sufficient accuracy for wind turbine monitoring but for accurate machine performance measurements I believe my old company are still using the original isolated data logging equipment and several IT people have tried to sell their kit to them and it just isn't up to the job.

Hall sensors seem so much easier for home use but you have to pay for them (precision shunts are not cheap either). Devices by Honeywell and LEM of the voltage output type or current balance type should get you better than 1%. Using cheap sensors with cut ferrite rings won't be very clever.

Flux

[Rover]

Here is my schematic for the system

Available here http://www.rovr1.com/wind/PDF/currentsense.pdf

the reason for the 4 shunts is that I'm measuring 4 different currents. These are commercial shuts. The 4 shunts are from Allelectronics and are 0.5% accurate 50A/50mV drop

http://www.allelectronics.com/make-a-store/item/SNT-50/50-AMP-SHUNT-50MV-50A-/1.html

The high side current sense IC is a MAX4173H

http://www.maxim-ic.com/quick_view2.cfm/qv_pk/1971

The ADC is a MAX186 12bit 8 channel

http://www.maxim-ic.com/quick_view2.cfm/qv_pk/1070

I'm still going to play around with and see if I can't get the sensitivity up

[Ungrounded Lightning Rod]

Your voltage divider is wired wrong.  Battery should go to the end and output from the tap between the resistors.

[Rover]

no its wired right I drew it up wrong.. crud.. got go fix the drawing

thanks for spotting it.

I know I wired it right since its working

http://www.rovr1.com/wind/datagraph.html

[Rover]

done

[scottsAI]

Rover,

With the additional data I can tell you exactly your problem.

Using the devices you selected your results are within their specifications.

Particularly the MAX4173H, input offset voltage is 3mv max to 5mv over temperature.

Add the gain back 3mv becomes 300mv, swamping out your zero reading.

Typical offset error is 0.3mv,

Shunt current of 300ma reads 0.3mv, same as the offset error.

Yep, this system is incapable of reading 300ma in shunts.

The error signals must be below the signal of interest.

Keep input offset voltage less than 30uV, to read 300vV signal.

The MAX4173H integrates several components, a discrete version can be built with more precise parts. Its gain accuracy is 0.5% typ, 2.5% max at 25C and 4% over temperature.

Let me know if you need help to do so. (or intend to)

Have fun,

Scott Beversdorf.

[Rover]

Thanks for the read on that...I may re-evluate the whole thing.

[scottsAI]

Rover,

I should add the offset error is plus or minus.

In your case the error is neg. hiding your low level inputs completely.

When other loads drive it above this level, then you can see both.

Have fun,

Scott.

[Rover]

Yep, I did figure that out..grin

[scottsAI]

Good.

{Answer everything, even if you think they understand it:-}

[Rover]

Ok... I might be a dumb ..s, It appears I'm also introducing error to the length of the wires from the shunt to the chip. (18g 1') .

As a test , I extended the wires (easier to lengthen then to shorten) , error got worse.

I'dont know why I used 1 ft wires, just dumb. Need to redo the circuitry and have the chis as close as possible to the shunts.

Won't solve all the problems, but some.

[commanda]

Do you have filtering at the input to the chip?

Amanda

[Rover]

Not at rs+ rs-

[scottsAI]

Rover,

Started this yesterday 2pm, storm came in and power failed.

Came back 11pm, Internet/cable not working until this morning.

Wire:

One foot or 10 foot should not matter.

Input of MAX4173H has fair amount of resistance. A little wire resistance should not matter. Therefore something else is the issue.

Was the wire twisted pair? Any radiated electrical noise will effect readings.

Like commanda asked: have any filtering on the inputs?

No, then add 100 ohms to each line, then at input cap to ground (2 cap).

Resistors and Cap must be identical units.

Other choice would be to shield cable if twisted does not work first.

Inverter may be generating the noise, filtering only solution.

Have fun,

Scott.