Anomalies testing voltage transformers 2

[MHEA]

When testing voltage transformers and their associated wiring I normally apply about 1% of the rated primary voltage and get a secondary voltage exactly primary voltage divided by the VT ratio. On some transformers however, I get ratio errors of up to about 10% when doing this. If I then increase the primary voltage to about 20% of rated primary voltage, this ratio error decreases to a negligible amount. It would seem to me that this is tied in to the design of the VT
Can someone possibly give me a technical explaination of this issue.
Many thanks

 

[waross]

Your test voltage may be too low to properly magnetize the transformer core. If the flux is so low that the hysteresis becomes significant then errors are inevitable.

Bill
--------------------
"Why not the best?"
Jimmy Carter

 

[electricpete]

An excerpt from IEEE 62-1995(R2005) - "IEEE Guide for Diagnostic Field Testing of Electric Power Apparatus-
Part 1: Oil Filled Power Transformers, Regulators, and Reactors 6.1.2.3.1 Voltmeter method"

Two ac voltmeters are used, one connected to the high-voltage winding and the other connected to the low
voltage winding. The high-voltage winding is excited to a voltage not exceeding the rating of the voltmeter.
Both voltmeters are read simultaneously. A second set of readings should be taken with the instruments
interchanged. The values indicated should be averaged to calculate the ratio.
A meaningful ratio measurement may be made using only a few volts of excitation. The transformer should
be excited from the highest voltage winding in order to avoid possibly unsafe high voltages. Care should be
taken during the application of voltage and during the measurement. It is important that simultaneous readings
of both voltmeters be made.
The voltmeters used should have accuracies commensurate with the requirements of a 0.5% ratio calculation.

That standard seems to suggest you should be getting a good reading even at 1% of primary voltage.

Magnetic non-linearities may play a factor as mentnioned above... I really don't know.

Another thing that comes to mind is that the signal/noise ratio is decreased as you decrease excitation voltage. Let's say for some reason you have 0.2vac error in your reading. If you are reading 1% of 120volts = 1.2 volts, that could be a 20% error. If you were reading 10% of 120 volts = 12 volts it would be 2% error. What type voltmeter are you using for measurements?

=====================================
(2B)+(2B)' ?

 

[electricpete]

Going back to non-linearity, I thought hysteresis would be more relevant at high excitation than low. (not positive).

I do know that as we increase excitation from very low to very high, the effective permeability tends to start low, increase to a maximum value at around the knee of the excitation curve, and then decrease again as the iron goes into saturation.

So at very low excitation the inductance is lower and resistance plays a larger role in the total impedance. Any resistance present on the primary side can cause some error in the measurement. Since the resistance is highest fraction of impedance of the leakage/magnetizing branch at very low (or very high) excitation the resistance-related error in ratio measurement would be higher at very low excitation. Having said all that I rather doubt it would be enough to have a significant effect.... just thinking out loud.

=====================================
(2B)+(2B)' ?

 

[waross]

What are you using for voltmeters? With old analogue meters (d'Arsonval type) and some early models of digital meters A distorted wave form would cause metering errors.

Bill
--------------------
"Why not the best?"
Jimmy Carter

 

[scottf]

MHEA-

It does have to do with the VT design. For a VT designed to operate at a relatively low flux density (i.e. a more conservative design), energizing at 1% of rated voltage will result in measurable error on the secondary. However, 10% off sounds bit high to me, but it's possible.

We get calls all of the time from customers trying to do a ratio test on 69 kV VTs using 120V on the primary and measuring the secondary voltage. They get ratios off by between 1-3% sometimes. The units are rated with a 0.3% accuracy class, so they call believing there is a problem.

Electricpete-

The clause you cited does not apply to voltage transformers. There is quite a bit of difference in what someone testing a power transformer would deem as "meaningful" and what someone testing a VT would deem as "meaningful".

 

[electricpete]

Thanks Scott. They mentioned 0.5% in the same paragraph so I assume that is the ballpark of what they meant by meaningful.

But the scope is "power transformers" which I can believe would exclude instrument transformers exactly as you said.

Out of curiosity, what's different about design of potential transformer that makes it more sensitive to ratio errors at low excitation than a power transformer?



=====================================
(2B)+(2B)' ?

 

[scottf]

Not a power transformer guy, but my understanding is that PT's are generally designed with much lower flux densities...which varies depending on standard and voltage class. There are a number of factors that drive this with accuracy performance and ferroresonance resistance being the 2 most prominent that I can think of.

Haven't thought it through all the way, but the winding designs are very different between PTs and power xfmrs. The HV windings of PTs are generally much "tighter" physically, in part to limit the leakage reactance.

 

[MHEA]

Many thanks for your comments they are much appreciated.

Regarding instruments: At voltages up to 600 volts I used Fluke 175 true RMS multimeters,swopping the HV and LV meters around in case one was faulty. I then managed to get an Omicron CPC100 test set,which is able to go up to 2000 volts on the primary. It showed a 0.2% error at 2000 volts (the VT is rated 11000/110 volts), but similar errors at 100 volt primary to thos using the multimeters. So I do believe the measuring technique. The VTs are installed on a metal-clad switchboard which is still completely dead, do I do not think we have any induced voltages.

Hope this helps a bit.

 

[scottf]

0.2% error on a ratio test sounds like a good result to me.

Why are you questioning the result?

 

[MHEA]

Many thanks for the comments.
Scottf, the unit is great at 2000 volts, using the CPC100, but this is not readily available to me. So I usually use the multimeter test at 100 volts, and this is where the large error occurs - but only on this VT. On most VTs the results with multimeters is spot-on.
Why?

 

[electricpete]

Just to clarify, when you say only on one PT… do you mean
A - you have one PT design that seems susceptible….
Or
B – you have several similar PT’s and only one of them shows this behavior (tests good at 2000volts and bad at 100 volts).

If A I guess we maybe chalk it up to low-flux design as discussed above.
If B I wonder whether residual magnetism might somehow play a role (could be checked by slowly ramping down from higher voltage to demagnetize),


=====================================
(2B)+(2B)' ?

 

[MHEA]

I refer to the last post by electricpete - many thanks for your input.
The issue is "A", one design. So why would a particular design have this behaviour? I would have expected the ratio to be maintained at any flux density. I base this thinking on the transformer equation
E = 4.44 Beta X area X turns X frequency. Also V1N1 = V2N2. This implies that as E is lowered, flux density reduces, but then output E would reduce proportionately.
So I am still baffled.

 

[electricpete]

Here is my take, which is similar to my comments 1 Jan 11 22:12 , with a few new ideas.

1 – We are all familiar with changes in iron behavior that occur at high flux near saturation, but there is also a change at low flux. The relative permeability (Ratio of B to mu0*H = slope of line from origin to any point on BH curve) starts low at low excitation, increases to a maximum near the knee of the curve, and then decreases again beyond the knee (saturation).

This is shown in "Figure MPC" here:

http://info.ee.surrey.ac.uk/Workshop/advice/coils/mu/

(the black curve is mu vs H and tan curve is B vs H.... me is slope from origin to the B vs H curve).

Another view of the same thing: here is data for M22 Silicon steel used for motors

http://home.comcast.net/~electricpete/eng-tips/ToroidSummary.xls

From tab chart1, we see:
The relative permeability at the peak is around 8000.
The relative permeability at the left of the curve is around 2400. From the shape of the cruve we expect that if we had data points further to the left they would have even lower peremability.

2 – This low permeability at low flux tends to decrease the magnetizing reactance at low flux. So let's model the transformer as simple 1:1 transformer (secondary referred to primary)..... assume that the current drawn by the secondary high-impedance measurement device is necessary.... our circuit includes primary resistance R1, primary leakage reactance X1, and magnetizing reactance XM.

By voltage divider:
V2/V1 = j*XM / (j*XM + j*X1 + R1)

If XM goes down due to decrease in effective permeability at low flux and X1 and R1 stay constant, then the ratio goes down. Is it reasonable for XM to go down and X1 to stay relatively constant? Yes. XM is heavily effected by the non-linear magnetic characteristics of iron since the magnetizing flux flows in the core (the flux linked by both windings). However X1 is not much affected because it includes mostly flux paths in air (the flux that links only primary but not secondary.. could include the space between primary and secondary winding as well as the incoming leads.... all of which is air not iron).

3 – It was mentioned by Scott that low-flux design is more susceptible. I guess low flux generally means high Xm, so it reduces the effect of X1 and R1 in the above equation. But it also means that when you reduce the flux by another factor of 100 (from nameplate to testing at 1% voltage), you'll be that much more to the left of the curve where the permeability is decreasing.

There may be other factors at work as well.

=====================================
(2B)+(2B)' ?

 

[electricpete]

Correction in bold:
"But it also means that when you reduce the flux by another factor of 100 (from nameplate to testing at 1% voltage), you'll be that much more to the left of the curve where the permeability is decreasing."
Should've been:
"But it also means that when you reduce the exciting voltage by another factor of 100 (from nameplate to testing at 1% voltage), you'll be that much more to the left of the curve where the permeability is decreasing."

Also I may have been a little off on my discussion of low flux or low flux density PT's (why some designs are low flux density). Maybe someone else will chime in if I said it wrong.

=====================================
(2B)+(2B)' ?

 

[electricpete]

Hmm. Weird typo corrected below:
"assume that the current drawn by the secondary high-impedance measurement device is necessary"
should've been:
"assume that the current drawn by the secondary high-impedance measurement device is negligible"

=====================================
(2B)+(2B)' ?

 

[scottf]

Electricpete is correct.

We run into this all of the time. We have a customer call and say "when I test the GE designs at 200V, they always ratio test fine and when I test your design at 200V, I get larger ratio errors". This is because our design uses a lower nominal flux density, which means that when operated at 100-200 times lower than rated voltage, the ratio error is off. Of course, it also makes our design much less prone to ferroresonance, runs with less losses/heat from core, has higher over-voltage factors, etc....

The good thing is that most field techs have access to insulation power factor test equipment (Doble test) and can take ratio measurements at 1, 2, 5, or 10 kV. Some insulation resistance test sets (Megger test) can operate at these voltages too.

 

[waross]

Would it be correct to say the;
"The MMF required to overcome hysteresis is insignificant. If the flux level is reduced to the point that the hysteresis MMF is significant, then the MMF to overcome hysteresis will show up as a measurement error."?

Bill
--------------------
"Why not the best?"
Jimmy Carter

 

[electricpete]

I don’t know the answer to what extent hysteresis plays a role but fwiw I have an opinion based on one datapoint (FEMM software) that it doesn’t play a big role.

I do know hysteresis is a very complicated phenomenon and difficult to model for simulations. The typical curve published shows results of pushing the iron fully into saturation in alternating directions and they label the width of the loop as 2*Hc (Hc = coercive excitation). Often it is explained that Hc is the amount of excitation required to overcome hysteresis. But it is not the amount required to overcome hystesis for any level of excitation… just for the particular high levels of excitation excitation that push the iron far into excitation. If you reduce the the excitation, then whole hysteresis loops gets smaller (proportionately smaller?) including the width of the loop. Does hysteresis play a larger or smaller role now?

As I said, it’s too complicated for me, so I look for a simpler description/model of this phenomenon. Finite Element Method for Magnetics (FEMM) software uses a model which simplifies the hystesis curve so it can be represented in sinusoidal analysis and still is stated/alleged to give almost correct results for hysteresis losses (and their contribution to changes in power factor).

The model they used is described here:

http://www.femm.info/Archives/doc/manual42.pdf

Also in the nonlinear parameters box is a parameter, PHIh_max. For nonlinear problems, the hysteresis lag is assumed to be proportional to the effective permeability. At the highest effective permeability, the hysteresis angle is assumed to reach its maximal value of PHIhmax. This idea can be represented by the formula:
PHIh(B) = (muEffective(B) / muEffectiveMax) * PHIh_max (2.7)

So here PHI stands for angle (not flux). And PHIh is the hysteresis lag angle i.e. angle by which B lags H in this approximation for hystesis. The approximation uses an oval to represent the hysteresis loop. If the angle approaches 0 it is a line (rather than oval) and there is no hystesis. If the angle is large there is hystesis.

Also we could convert PHIh to more familiar terms as follows:
ThetaPF = (thetaV-thetaI) = (thetaV-thetaB) + (thetaB-thetaI)
We recognize the first term as Pi/2 (since V~dB/dt) and the second term as -PHIh(B)
So ThetaPF = Pi/2 - PHIh(B)

PHIh(B) is proportional to effective permeability. We already know effective permeability is low at low excitation levels and highest at peak of the curve.

Putting it all together:

At low excitation levels we have low permeability and therefore low PHIh(B) corresponding to loops that look like lines (thin) and power factor angle (ThetaPF = Pi/2 – PHIh) near 90 degrees…. i.e. low hysteresis losses.

At the knee of the curve we have highest permeability and PHIh(B) attains it’s max value corresponding to loops that are wider and power factor angle (ThetaPF = Pi/2 – PHIh) dipping farther below 90 degrees…. i.e. more resistive hysteresis losses.

So, if we believe this simplified model, hysteresis plays less of a role at low excitation than it does at high excitation. Also the software in general is pretty widely used and cited in many peer-reviewed papers. I’m not saying the model is perfect, but it’s the only datapoint I have to judge this particular question.


=====================================
(2B)+(2B)' ?

 

[waross]

I think that you and I were reading different pages of different chapters of different books when we studied hysteresis, Pete.
We are talking about lowering the flux density on a transformer that is designed to use a low flux density under normal operating conditions.
We are not concerned with the possible effects of saturation.
Let's reduce this to the absurd;
We have a transformer designed to work at a low flux density and are testing at a few percent of design values. As we lower the flux density we observe errors.
If we reduce the exciting MMF to less than the hysteresis loop, we may have no change in flux density and very large errors, possibly approaching infinity.


Bill
--------------------
"Why not the best?"
Jimmy Carter