Current Transducer Low Range Accuracy?

0.5% accuracy of 1000A CT = 5A and 1% of 50A CT is 5A it's the same, but this is your objective, first check from calibration of both CT and see how accurate they will measure usually you will see something like 1%+/-span. This will be the determining factor if its okay to do over 50A CT and use 1000 A CT.

Anyhow, don't use correlation on metrics like this. It's a bad practice, ratio and proportion is not advised, valid and recommended as substitute for the real ones. This is if you want to have a real results.

Uhm.... 0.5% accuracy of the 1000A CT is 50A, but the 1% accuracy of the 50A CT is 0.5A, so they are not the same. These are the instrumentation accuracies in the datasheet.

I fully understand the most common answer saying "no you should use the 50CT for such low currents as the 1000A CT will have 5A accuracy". Thats not neccessarily true in my case. See my results (matching perfectly) and my explanation.

Exactly, seem like you know the answers already.

Stick to the standard and importance, your means is yours personal only and can not and will not be treated acceptable.

There is three sides of every story, there is yours, theirs and the truth.

To which side do you think your means will fall?

I am not sure if this is an irony or not... I am not looking forward people just saying I am right, I need my thought porcess be confirmed by experts, or rejected by experts.

I am 100% OK to hear I am wrong if its explained. Please help me understanding what I see here.

Thanks!

Averaging filters out random noise errors, but does not correct systematic errors like sensor unlinearity, power supply related gain variations, aging, temperature drift of hall sensor etc that apparently do get worse as you get near the noise floor. The problem is not that you can't compensate against those factors with DSP or lookup tables, it is that you can't do it easily and long-term reliably. Now If you make those measurements in a controlled environment, you could have some confidence, but this does not just apply all around. S.M.

I dont need long term reliability as I can repeat the software offsetting before each one of my processes. And as none of the processes takes more than 10sec, I guess we can ignore the degradation by time.

The environment I have is controlled in the terms of temperature, but again as I can do the software offset everytime, I dont think even that matters too much. And I can certainly.

I am really not looking forward to get a universal solution how to use any instrument in their low ranges, but looking to understand why do I see what I see. And I see that accross multiple sensors, so claiming that its just a coincidence, or luck is quite unbelievable.

It's not that unheard off to achieve better than expected specs of a device using smart mods or shortcuts like your averaging, it's no magic, it has been statistically explained, averaging=less noise, higher resolution and the main drawback is slower response, so have no remorse, it's common practice. I've done it enough times myself. But you can't blame the manufacturer for not mentioning it now, can you? (LOL) S.M.

No mate... I dont blame anybody. I am just shoked by the fact that I can achive such correlation between two instruments designed to operate in such different ranges, and I am willing to know if I do something bad or it is acceptable.

By the way after I take my 1000 readings under no current condition and calculating the average, I dont average the samples anymore. I just sample the current signals at 2kHz or 5kHz and substract the average. That gives me the correlation I mentioned.

What you have are probably brand new current transformers. If they both work on the same principal (primary turns to secondary turns), then the results can be remarkably comparable.

If there is no magnetic material in the path, (and thus no hysteresis) then the result is not surprising also.

I don't find this a surprise now that you've highlighted it. If there was a concern, the power companies would make us isolate our mains supply metering for high load motor/pumps from the low building loads, but they don't.

The other alternative is that the measuring meter you are using is so coarse that it cannot discriminate a difference if it did exist, but with modern available metering this is unlikely to be the case.

You measured the value and know the accuracy of your CT's. The values of accuracy are the limits that are placed on them. If they are more accurate than the accuracy class, that does not matter. Ct's are usually very acurate and many times are better than they are supposed to be. What you have experienced does not surprise me. The problem with the 50 amp CT is that it will saturate at higher current values and most traducers will do the same, in fact they may totally burn out at higher than maximum current. The maximum current is much higher than the 50, but must be checked on by verifying the additional ratings of the transducer. I would not generally put a 50 on a circuit with that much available current.

The OP refers to Hall Effect current measuring devices, some look like CT but have analog voltage or 4-20 mA output (not 1 or 5 A). I've used some in projects where DC current or TRMS current was required. Despite the integrated electronics they seem reliable but are way more expensive than wound CT. Referring to precision I could obvioulsy only state the one of the manufacturer data even if real life precision when new was of course better (didn't check the long term stability).

In any case, as others mentioned it above, you can't rely on any better specs than those issued by the manufacturer. I'm even somewhat surprised by the excessively good results you got.

I am aware of what the OP has. The transducers when I was still working were mainly used for DC and were very needed to have accuracy from very low values to very high values. Most greatly exceeded the the rated accuracy under precise tests, the same as AC CT's do. I did work on some AC systems where they were used, but as you said, the cost was higher than a conventional CT (at that time). Still, all in all, equipment can usually be used in a manufacturing or utility application with tests to confirm accuracy or suitable service, and most codes allow testing to confirm suitability. However many industrial customers simply ignored codes because they rarely were written for heavy power uses. Around offices etc, they followed them of course, but around really heavy power that I dealt with they simply had no application.

Just to demonstrate the correlation please check the attachments.

The first one shows the signal acquired from the CTs at the same time. You'll then most of the cases its not even visible that we acquire two signals as the overlapping each other that well.

The second shows the histogram of the differences after the signal dropped below 10A.

It is not surprizing since the Hall effect is linear per se. However the electronics you use to collect data is not totally so that your uncertainty nwill dpend on the nonlinearity of the amplifier and on the rounding error of the ADC. The best solution could be touse 2 gains for the 2 ranges and to calibrate the sensor WITH its amplifier for the lower raqange you intend to measure. The gain is useful in order to reduce the ADC step error since increasing resolution.

Thats true, but the resolution / error of my acquisition device is so low that it is negligible, wont matter more than about 10mA and I certainly dont need that accuracy.

You should make sure that you are not exceeding the current ratings of the 50 A current transmitter when you go up to 1000 A. Then use readings from the 50 A for zero to 50 and readings from the 1000 A current transmitter for readings from 50 A to 1000 A. It should work perfectly. Make sure that the two are calibrated so that there is no discontinuity at the transition from full scale on the 50 A to the 50 A reading on the 1000 A transmitter. Fancy software to check this would be useful to avoid jumps in the system if there is a discontinuity in the two readings.