Be Wary of CT obfuscation

I did my morning reading, and read with interest the highly technical discussion of saturation at the upper limit of useful range of the SCT006. By the time you finish the post, you’re ready to take a hammer to them. Not so fast…

There’s a history to this, and some of you may have been following, but the vast majority get their news from headlines and post titles, and that may be misleading.

The backstory is that the question of whether this CT might be exhibiting distortion in a 5.4A circuit was answered by me with my data from testing it at 60Hz. Subsequently, readers were warned that those tests are not valid at 50Hz. Specifically, the writer asserted that there was a direct linear relationship of 50Hz/60Hz that decreased the useful range to 16.666A (from an initial questionable 20A). I questioned that relationship then and I still do. What happens if you use it at 400Hz - is the capacity 133A?

I have been admonished for questioning the authority of the gurus that inhabit this forum, and I will surely suffer their wrath for this post as well.

But being a bottom line guy, I connected a venerable SCT013-000 and an SCT006 to my IoTaWatt and got out a collection of light bulbs. Using a jig with 1,2,6, and 8 turns, I obtained these results:



Sure, you can analyze these devices and find flaws. The question is what does it mean to someone trying to find out where their electricity is being consumed? These results work for me. There’s a guy who is going to make a fresh argument about extreme power factors and funky current signals. I stipulate that if you use the SCT006 in that environment - at > 17-18 amps, your results may vary.

BTW/ The argument that capacity is linearly proportional to frequency has not been supported, nor has any difference between 50Hz and 60Hz capacity been demonstrated. I don’t know the answer, I’m waiting for equipment to find out, but that’s the point. I’m waiting for equipment before I state facts that aren’t…

Yes, I did kinda take this personally.


A lot of pile-on with this one. More graphs and test results. Doesn’t really shed much light on why my 60Hz results at 20.1A are less than 1% different from the SCT013, but I remain vigilant seeking that 6% below nominal that was predicted.

Have to give them credit. They are putting so much work into this. The motivation is admirable.

Anyway, just to not look like a slacker and keep up, I restested the 18-20 amp on IoTaWatt at 60Hz using two of each - SCT013 and SCT006. I have to admit, the SCT006’s do not agree with each other the way the SCT013’s do. Nevertheless, they appear to be within the 1% of each other as expected by the specs.

So here they are at 18.1A 60Hz, The Fluke meter says 2.264 x 8 turns = 18.1A
Power factor looks good, the average of the two SCT006 is about .68% below the SCT013’s
Here’s what the IoTaWatt sees for signal. Not too bad IMHO.

So now lets kick it up a notch to 19.2A. Fluke says 2.398 x 8 = 19.2A
PFs are still pretty good, but now the SCT006’s are 1.28% below the SCT013’s
Again, here’s how the IoTaWatt sees it:

OK, lets push it up past 20. Fluke says 2.526 x 8 = 20.2A
So now the SCT006’s (average) is about 2.6% low - Clearly input_4 is having a problem.

OOOH! There’s that saturation. Not quite as big as some others, but nevertheless it’s there.
How about Input_3?

Not too bad at all.

So they both run pretty good right up to 19.2A, then one of them starts to falter at 20.2A.

Still looking for that predicted 6%, but maybe it’s just not there in these two that I have, and others are worse.

I got my 50Hz 230V inverter yesterday, and it works great. The output sine wave is better than my 60Hz utility feed. Waiting for some 230V incandescent bulbs and I’ll run all this at 50Hz. I expect these guys are on the right track and the onset of will maybe move below 20A at 50Hz, but if I were a betting man, I would bet against that new 14A claim, and even the original 16.666A.

To me your results are easily in line with the results shown by @dBC. There will always be some variation in the point of saturation across units due to batch differences, ferrite material differences, split core alignment/gaps etc. I think the point is saturation does occur in all CTs, and the use of a 24 Ω burden moves the distribution of saturation points down to the 14-20A range. This is further exacerbated by the use of 50Hz where saturation kicks in earlier.

Yes, I guess it could be sample variation.

Overeasy, you have a scope right? What does your scope show for your two SCT006’s wrapped around a conductor carrying 20.2A, preferably at the same time with one on each channel.

I’ve attached my raw scope capture files to the scope pics. They’re untouched by human hands and generated by some pretty high-end lab equipment (calibrator and scope) so I’m fairly confident they do represent what at least some of your users will see.

What I’m less confident in is my spreadsheet manual labour to turn that into power readings like:

Vrms: 39.87V (0.3% below nominal)
Irms: 19.14A (4.3% below nominal)
RealPower: 749.8W (6.3% below nominal)
PF: 0.9823

I could easily have slipped up with cutting and pasting (and did several times prior to posting the results). I don’t know how easy it would be, but if you could find a way to feed that data into your IotaWatt as if it came from your sensors, we could at least agree or disagree that that data produces those results. If we disagree then I can recheck my spreadsheets and you can recheck your code.

That’s 20.2A? I read that at about 800mV peak, or 566mv RMS. At 20.2A primary you should be seeing 606mV RMS across the shunt. Are you sure you’re using 24R shunts?

In any case, your voltage levels are significantly lower than mine (and significantly lower than nominal) so that scope trace alone might put you well on the trail of why you’re not seeing the distortion I do.

Never mind… I misread the ticks.

Ebay scope. Not that crisp.

Well I can certainly see why you’re not seeing the under-readings with that signal, it’s way better than I see. It remains a mystery. At this stage I think maybe @philb’s theory is the best we have.

The kink in your Input_4 image above seems much worse than either of those on the scope, or is it just an optical illusion because there’s a V to compare against?

Trystan put me on to this

I’m not following sorry?

I cleaned the mating core surfaces and clamped them shut.

I was on the other line with Trystan and he was asking me to check the mechanical aspects of the CT. Meanwhile you wanted a picture on the scope, so that’s what I took. Here are the IoTaWatt traces:

So you are suggesting the extra clamping force holds the split ferrite together better and reduces saturation? That would likely be enough to explain the differences in your results vs @dBC.

Have all your results been using this setup, or is it a new thing? Not the most practical of setups in a switchboard, kind of negates (an then some!) the benefits of the smaller physical size of these CT’s.

The plot thickens. I just went over my numbers and found an error. Over the last week or two I’ve gone through the IoTaWatt phase correction algorithms to increase accuracy. I had spent a day recalibrating my collection of representative CTs, but didn’t do the SCT006. So I just checked it for reasonability against the VT (I’ll check the full spectrum against the AC line tomorrow). The VT has a 1.8deg net lead, and I measured the SCT006 at about 1.08 relative to that at about 4 amps. It’s a little higher as the current increases so I set it to 3 as a happy medium. Now the net between the CT and VT id about 1.2, and that’s what is corrected.

Results are much better. Up to about 15 Amps, they pretty much match the SCT013 watt for watt. Here’s the 20.1 amp case:


So with my CTs, properly phase corrected, I’m getting about 0.7% low. This doesn’t explain any of the saturation point questions, but it does explain why with perfectly good waveforms IoTaWatt wasn’t getting better results.

I think the solution is going to be for one of us to test the other’s samples. Different results it’s the test protocol, same results it’s the samples.

You think there might be some variation in these things? I received mine from Gwyn, I think about a year ago. They are two of the samples that were sent to him by YHDC. I wonder if he can identify the rest of that batch, and see if they are all different. If the stock doesn’t match the samples, maybe there’s recourse.

That 30A version is being actively advertised on alibaba. Like to try one of those out.


I’m a little confused about where you’re at. Can we go back to your post #2 for a minute please? As I understand it you’ve got 4 inputs measuring the one signal which the Fluke says is carrying 20.2A. Input 4 is distorted and its RealPower is 3.9% lower than Input 1. Do you still have that CT? Can you still replicate that situation?

Yes, I still have both CTs. I can replicate the situation, and the result should be the same plots, but the power inducated by IoTaWatt will only be around 1% less than the sct013s.

As the saturation sets in, there is more shift, and in post 2 the phase correction was set very high, resulting in the larger effect on the real power. With lower values, the ct now matches the sct013 nearly exactly until saturation starts, where it starts to fall off but still less than 1% at 20A.

Understand that the plots are still of the raw data. The shift correction happens after when the samples are processed.

There are four CTs connected and they are all clamped to the same 8 wire shrink tubed bundle of primary wire. The Fluke meter is in series with the incandescent light bulb load on the primary.

The IoTaWatt doesn’t directly report current. What you see in the shots as PrimaryA is an output I defined that simply divides power by voltage. As the power factor falls at saturation and the real power becomes less than apparent power, that number might fall by .1 or so. The true current is the Fluke x 8. I did double check by doing an rms calculation on the data points in the plot. The rms amps match the meter x 8.

The clamping of the CTs was only during that later set of pictures and explained what was going on then. The latest numbers with calibrated phase correction are without the clamps, although there does still seem to be a little residual improvement, probably from cleaning the mating surfaces of the CTs.

I’m trying to work out whether the difference is we’re just seeing different amounts of distortion or whether we’re processing it differently. It’s hard to tell just eyeballing your distorted Input_3 input_4 in post 2, to my 60Hz 20A distortion plot which you can see in this post.

Could you attach the raw sample pairs for both the Input_3 and Input_4 plots of your post 2 above please? That will give me a distorted one and a not distorted one as a reference. I should then be able to compare your CT distortion levels with mine.

Yeah, Id have to recreate that. I was pasting into the same spreadsheet destructively. Look, the distortion in your scope trace is pretty self-evident. I have no reservations about your competence to produce a 20A primary and measure it on the scope and I’m not questioning that. I hope you have the same feeling about my data, and the picture from my scope. I know in the past you have questioned my ability to collect a viable set of samples. That’s not a discussion I want to have.

Given three traces, one with visible distortion and two with marginal, I think at this point the CT’s themselves are in question. My point above was that YHDC seems to have made changes to this line over time and it’s not out of the question these CTs are just different. The practical solution, to my mind, is to get to the bottom of that.

Beyond that is the question of what effect 50Hz has. It’s important to understand that, and I’ve been remiss to ignore the issue until now. I’ve taken action and have a 230V 50Hz platform and I’ll get on that posthaste.

If you could please, I think it would be really helpful.

SCT006.ods (100.5 KB)

Load is like 20A now. What can I say, I’m using light bulbs and the voltage is down today. Probably better to match your precision 20A tester anyway.

Thanks. Yes, unfortunately that one doesn’t show any significant distortion at all. What I’m after are two captures one from your Input_3 CT and one from your Input_4 CT like you had in post 2 above. I’m trying to compare your distorting CT with your non distorting CT… even if you have to bump up the current a bit to get there. I appreciate it’s not so easy when you’re at the mercy of the grid voltage.