Current Shunt Vs Current Clamp Meter - 15% difference Which one is right?

I have a variable AC power supply. 0-12V 0-1800A 60hz.

I have always been told to trust the current shunts and not current clamps when measuring current. The current clamps are supposed to accurate to less then 2%. Our current clamps have always been about 5-15% different then the current shunts though. I want to know why, and which one I should believe.

My boss wants me to prove to him that the shunts are more accurate than the clamps but I don't know where to begin.

All items are calibrated annually.

• Fluke 353 (1.5% accurate)
• Ideal 61-746 (1.7% accurate)
• Current shunt 500A 50mV (10:1 ratio) (Empro)
• Current shunt 3000A 50mV (60:1 ratio) (Ram Meters)
• Current shunt 2000A 50mV (40:1 ratio) (Empro) (EDIT: Added)

mV meters for the shunts are fluke 45 and Fluke 289 both read the same

I adjusted the power supply based on the current shunts.

    500A shunt - first Iteration
Shunt         Fluke  Ideal
Target  mV      Current
50      5.5     55       48     49     13%  11%
100     10.5    105      91     92     13%  12%
200     20.1    201      178    182    11%  9%
300     30.1    301      263    269    13%  11%

3000A shunt - first Iteration
Shunt        Fluke   Ideal
Target  mV      Current
100     1.716   102.96  98.3    100.3   5%  3%
200     3.345   200.7   190.7   194.7   5%  3%
300     5.023   301.38  284.8   292.1   6%  3%
400     6.695   401.7   379.4   389.1   6%  3%
500     8.356   501.36  471     484     6%  3%

500A shunt - second Iteration
Shunt        Fluke   Ideal
Target  mV      Current
50      5.125   51.25   45.3    46.5    12% 9%
100     10.304  103.04  91.9    93.5    11% 9%
200     20.3    203     181.2   185.1   11% 9%
300     30.225  302.25  269.3   275.8   11% 9%
400     40.21   402.1   357     366.2   11% 9%

2000A shunt - first Iteration (Fluke 45)
Shunt       Fluke   Ideal
Target  mV      Current
50      1.21    48.4    37.17   38      23% 21%
100     2.44    97.6    84.7    86.4    13% 11%
200     5.02    200.8   181.9   186.3   9%  7%
300     6.31    252.4   230.5   235.8   9%  7%
400     10.09   403.6   360.2   372.2   11% 8%
500     11.31   452.4   405     415     10% 8%

2000A shunt - second Iteration (Fluke 45)
Shunt      Fluke   Ideal
Target  mV      Current
50      1.26    50.4    29.33   29.5    42% 41%
100     2.48    99.2    65.7    65.5    34% 34%
200     5.14    205.6   141.3   144.1   31% 30%
300     7.46    298.4   207     211.5   31% 29%
400     9.89    395.6   284.4   290.6   28% 27%
500     12.5    500     355.3   362.6   29% 27%

2000A shunt - third Iteration (Fluke 289)
Shunt       Fluke   Ideal
Target  mV      Current
50      1.21    48.4    37.2    37.4    23% 23%
100     2.51    100.4   76.6    77.3    24% 23%
200     5.14    205.6   165.5   168.8   20% 18%
300     7.58    303.2   252.6   257.8   17% 15%
400     10.11   404.4   344.6   351.7   15% 13%
500     12.49   499.6   418     426     16% 15%


EDIT: After I calculated the error rates it looks my 500A is off more than the 3kA. I am going to try it with a 3rd shunt tomorrow.

EDIT2:

200A 50mV Shunt Fluke 353 Clamp Meter Fluke 45

• How is "nominal" determined? If it is the power supply reading, then how accurate is that one? What are the exact resistances of your shunts (at the temp they are run at), and how accurate is that value? You can then just work your way up from those values and their respective possible value range and compare. May 18, 2015 at 21:33
• Silly question perhaps, but are the shunts installed correctly? It's suspicious that your shunts are consistently reading 'high' relative to the clamp-ons. May 18, 2015 at 22:27
• @PlasmaHH sorry by nominal I meant my target amperage, the power supply I have is just manually operated. I made my adjustments based on the shunts, then took readings with the clamp meters. May 18, 2015 at 22:47
• @SpehroPefhany Well I would assume so. Everything is tight. I'm not sure how I could screw it up other than that and get and kind of logical reading at all. you stick it in series in the circuit and measure voltage drop across the shunt with the other two terminals. May 18, 2015 at 23:00
• Any idea the manufacturer and brand name of shunts? May 18, 2015 at 23:04

If you do the math, the Fluke 353 and Ideal 61-746 are within 2.2% error (STD 0.4%). This is well within the level of accuracy of the machines given (1.5% and 1.7%) with the Ideal always larger than the Fluke. For me, this correlation means they are the most accurate.

The current shunt is a manganin resistor. 100µΩ, 25W. Online references state accuracies of ±0.25% (which is down from your 2%). This should be the most accurate according to workplace lore.

If you look at the errors for both the 500A and 3000A, they start high (10%, 3%, 2.5%) and go low (<0.5%). This does make sense because it is a resistor and even though temperature coefficient is 0.00001, it will be most accurate at rated values.

Difference between 500A current shunt and Fluke/Ideal ranges from 9-13% lower (3000A from 3-6%). This tells me there is a systemic error.

The Fluke 45 and Fluke 289 give the same answer. I'd look at how there are connected to the current shunt. We are just dealing with 50mV. Thicker, shorter wires possibly? (Not even sure how it is connected.)

Your problem is you have three answers and you don't know which is correct. Two agree, but the most accurate (in principle) is always higher.

You need another reference which can be verified some way. I always try to go back to basics. I'd go borrow a calorimeter and boil some water.

Edit...

The five error sources inherent to current shunts are:

1) Connection

2) Temperature

3) Frequency

4) Drift

5) Thermal emf

And...

Most modern metrology-grade shunt manufacturers are aware of these problems and have attempted to design them out of their products.

Figure 5 shows a metering type shunt highly susceptible to current and potential connection errors

I substitute your image because his was similar.

The resistance of shunt Manganin rises about 20 ppm (0.002 %) per degree C around lab ambient. Applying current causes self-heating, which changes resistance. This change is not linear. Some shunts rise to a maximum resistance at a certain current / temperature level, then fall as temperature continues to rise.

How fast do you make your measurements?

A shunt must stabilize at each temperature / current level. The thermal mass of a shunt includes its resistance element, its end blocks, the current cable lugs and connection hardware, and the cables themselves. At higher current levels, a shunt may require more than an hour to reach thermal equilibrium. This is when measurement should begin.

Not that they say it's not significant for 50/60 Hz. But you could try your measurements with shielded twisted pair.

Figure 8 shows ac coupling between current and potential circuits. Coupling can be reduced by connecting current leads in line with the shunt and routing potential leads together into a shielded, twisted pair extending at right angles from the shunt (green, not red).

Would this account for 9-13% error always in the same direction? The jury is out, but I'd say yes. It does give you things you can try.

I'd pour over that report and you should be able to easily prove that the clamp meters are the most accurate.

• Even if the resistors value is within 0.25% the voltage drop reading over the shunt is subject to the accuracy of the used meter there. Also thicker shorter wires will not change the voltage measurement here, as the meters have high enough impedance May 19, 2015 at 8:22
• I think I have found the problem it seems to be ac coupling on to the measurement leads. May 19, 2015 at 21:22
• Does anyone have any recommendations on what kinds of shielded twisted pair wire I should try? May 19, 2015 at 21:23

I'm having a bit of difficulty with your numbers. The Fluke and the Ideal readings differ by (typically) 2%, not 20%.

Both sets of readings are lower than your nominal current levels, but you haven't explained where those numbers come from. In any event, typical deviation of your Fluke numbers from nominal is about 5-6%, with the worst being about 15%.

So where does 20% come from?

Since the accuracy of the Fluke is 1.5%, and the Ideal is 1.7%, any difference less than 3.2% is entirely appropriate to the stated unit accuracies, and the two sets of measurements appear to agree to within specified accuracies.

I would be careful that you are reading true RMS in all equipment test cases and not averaging/peak or? Is your current wave form load a sign wave (resistive)? or something else (diode, triac/SCR)?