Power coil inductance measurements with standard desktop RLC meters 1

[homoly]

Hello,

I would like to ask you what is your experience with measuring of power coils ( nominal current few hundred amps ) with ferrite core using standard RLC bridge. My colleague claims that the results will be not usable, but my assumption is that if the measuring device works with the constant current mode and the coil is well designed and not getting to the saturation the inductance computed by increments should be a very good approximation of the real inductance.

If the RLC is not a good way could you advice a manufacturer of the measuring device and type ?

Thanks.

 

[itsmoked]

I would side with your colleague as the minuscule drive a bench top RLC would supply would not actually drive the magnetic core of a large magnetic structure. But that is just a 'gut feel'. Hopefully someone can clarify this.

Keith Cress
kcress -

http://www.flaminsystems.com

 

[electricpete]

I think the permeability curve is relatively non-linear throughout the range, not just at saturation. Effective permeability at very low flux is different than at middle level fluxes. I will see if I can track down an example curve.

Another possible consideration. RLC meter may provide a higher frequency measurment. If frequency is high enough parasitic capacitance will start to alter the computed inductance compared to a power frequency test.

=====================================
Eng-tips forums: The best place on the web for engineering discussions.

 

[electricpete]

I left out the word iron. Of course I was talking about the curve for soft iron core material.

=====================================
Eng-tips forums: The best place on the web for engineering discussions.

 

[electricpete]

Here is a B-H curve for M22 Silicon Steel. Note the vertical axis is linear and horizontal axis is log. It is not very linear anywhere

=====================================
Eng-tips forums: The best place on the web for engineering discussions.

 

[electricpete]

My comments applied to an iron ferrous core. I just noticed have have a "ferrite core" - something I'm not very familiar with.

=====================================
Eng-tips forums: The best place on the web for engineering discussions.

 

[electricpete]

About the only thing I know is that ferrite permanent magnets are equivalent to ceramic permanent magnets. I assume unmagnetized ferrite material is similar to ceramic material.

Attached is a BH curve for "Ceramic 5" (whatever that is... my best guess is that it is similar to ferrite). It is very non-linear.

I am frankly surprised that anything with a ferrite core has such high current (hundreds of amps). But I don't know much about it. If it is a very high frequency application, the role of frequency might be sort of the reverse... parasitic capacitance may play a role at operating frequency but may not show up at lower measurement frequency.

=====================================
Eng-tips forums: The best place on the web for engineering discussions.

 

[electricpete]

Last comment - to the best of my knowledge, magnetic "saturation" is a term usually used in the context of ferromagnetic materials. The ceramic/ferrite curve shown here is non-linear, but quite different than saturation. With the high current, I have to double check one thing - are you sure this device has ferrite core and not a ferromagnetic core?

=====================================
Eng-tips forums: The best place on the web for engineering discussions.

 

[ScottyUK]

ePete,

Look for a ferrite like 3C8 or 3C85. Not great for very high frequency but high saturation flux density compared to other ferrites. I think the opening comments from you and Keith are on the mark.


Thie link is for 3C95



----------------------------------

If we learn from our mistakes I'm getting a great education!

 

[homoly]

Sorry, it was my mistake, it is of course ferromagnetic. I apologize for this wrong translation, I was in hurry .

I will give one example the numbers just to make a picture: core is made from 23ZH100 material and the coil is a part of the inverter PWM to sine output LC filter, where basic of PWM is 7.5 kHz.In this case the current should be around 50A but in general as I can be faced also with coils with nominal In with few hundred amps.

In general I would need to measure coils for PWMs from 2.4 Khz up to 16 kHz.

So as I get from entries, for these frequencies the 1kHz or maybe 10 kHz offered by some RLCs will be not enough advantage to consider the purchase, because the low excitation of the core will be bigger disadvantage in this case.

Do you have some idea for this kind of power coils what should be the proper instrument to measure?

 

[electricpete]

So as I get from entries, for these frequencies the 1kHz or maybe 10 kHz offered by some RLCs will be not enough advantage to consider the purchase, because the low excitation of the core will be bigger disadvantage in this case

That is the way I look at it: inductance depends heavily on excitation (amp-turns) throughout the range (not just at saturation) as a result of the non-linearity of the iron.

When I first heard that concept, it was a little surprising to me since I also thought that iron was linear up until near saturation. Certainly it is non-linear accross the range, but the question could be asked "how much", so I would like to talk about it a little more and if anyone disagrees please feel free to share your experience and insight.

There is a gentleman that performed some single phase tests on a stationary three-phase motor as described here:

http://maintenanceforums.com/eve/forums/a/tpc/f/7161085912/m/5731057063?r=941105701#941105701

In this case, the voltage was held constant (10vac) and the frequency was varied in the range 0 – 1000hz. This results in change in flux density (volts/hz), and results in excitation (other than dc) was far below satuartion because the equipment is designed for 220volt, 60hz.

You can't see the attachments unless you register at the site. However, I have included attached an extract from the last two attachments where you can see the computed inductance during this test went up from 0hz to 250hz, then started decreasing from 250hz to 1000hz (and presumably beyond). The relevant graphs are in slide 6 and 7.

I would say the same behavior is shown in the iron B-H curve. The curve I posted before was a log-log plot . You can see a linear plot at the link below where the non-linearity accross the range is again apparent and they give a very brief discussion of the reason for the shape of the curve:

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

1 -Close to the origin a slow rise due to 'reversible growth'.
2 - A longer, fairly straight, stretch representing 'irreversible growth'.
3 - A slower rise representing 'rotation'.
4 - An almost flat region corresponding to paramagnetic behaviour and then ?0 - the core can't handle any more flux growth and has saturated.

Now of course you have to do some mental gymnastics to convert a B-H curve into a V vs I curve.
v(t) = d/dt (N*A*B) = N*A*d/dt(B) = N*A*dB/dNi * dNi/dt
dNi/dt = N*di/dt = v(t) / [ N*A*dB/dNi]

We can integrate the right side to determine current for a given sinusoidal v(t) using dB/dNi which is the slope of that B vs H curve shown (neglecting effects of hysteresis). Of course the resulting I is nonsinusoidal but we have to look at the fundamental component. That gives a mental roadmap to convert the curve B vs H into I vs V (whose slope is 1/[w*L]) . Clear as mud I imagine.

If I get a chance (maybe this weekend), I will use FEMM to do the calculation and plot a curve. It is a subject I'm interested in anyway for other reasons. I don't know what 23ZH100 but I assume it is similar to M22 Silicon steel (?).

By the way parasitic capacitance should not be a factor in the fundamental component response at 16khz... shouldn't be a factor until the Mhz range.

So – my bottom line again is that I do think your effective inductance will depend significantly on fllux density due to nonlinearity of the iron. In that case, there is not much choice except to test it close to the range that you expect to be operating it. (same flux density... could use power frequency test at proportionally lower current to achieve the same flux density... test in the actual application is even better). I have given reasons for my conclusion above. I am interested to hear if anyone else has experience or insight on this aspect.


=====================================
Eng-tips forums: The best place on the web for engineering discussions.

 

[homoly]

Electricpete, thank you for your valuable entry. In principle the your mental gymnastics is quite clear.

The 23ZH100 material datasheet can be found at

http://www.nicore.com.cn/en/upload/download/pic-0-13-1.pdf.

In this case it is quite linear characteristic.

I went through some 50Hz transformer core material characteristic which are more likely similar with the sample characteristic you have provided and I would say the desired "delta" measurement area with constant current method should begin at least at 0.4T inductance ( the transformers we are using are designed typically to work at 1.2T ).

As a mater of fact it seems there is no universal recipe but always the first task is to get the information on working inductance at nominal current, then check the characteristic of the core material and select the less inductance from which the error of the measurement is acceptable.

I will check the RLC bridge measuring methods, but it seems the only good way will to find a bridge capable of work with external power supply. Then the voltage vs current characteristic(dI/dV ) should be converted backwards to dB/dH information.

I think to change the 1/wL slope with higher frequency to simulate higher excitation is also not the best way because if the H ( current ) and B characteristic is not linear I think you cannot determine what is a proportional frequency raise to decreased current which will simulate the same inductance.

I will look forward for your simulation result.

Gabriel.

 

[electricpete]

A gapped core can be used to stabilize the value of the inductance against nonlinear characteristics of the iron. It provides a large airgap reluctance in series with the iron reluctance so that variations of iron reluctance has less effect. Do you know if you have a gapped inductor?

At first glance, we might think a motor would resemble a gapped inductor and so we wonder why the motor data I posted above was so nonlinear. There are some important differences between a motor and a gapped inductor:
1 - The airgap is long in the circumferential direction which may tend to reduce it's reluctance slightly.
2 - More importantly - the motor has a short circuited secondary loop. In this case (the one that was tested as linked above) a cast rotor. The rotor conductors are not visible from outside of the rotor but are just a few mm below the outer surface of the iron. The flux flowing circumferentially along the rotor has two parallel paths - one through the thin bridge between the bar and the gap, the other thru the thicker iron deeper in the rotor. In normal operation at rated flux density the thin bridge of iron is saturated and all the flux flows through the deeper path. During the low flux density testing, the thin bridge is not saturated and a large fraction of the flux may flow through the bridge. Small changes in the reluctance of the bridge due to non-linearity have a big effect on this flux division. Division of flux between these two paths determines whether the rotor conductors are linked by the flux and has a big impact on the resulting rotor current and "load component" of stator current, which affects the computed "effective inductance".

To make a long story short, induction motor was a poor example for me to use since it does not resemble a simple inductor (either gapped or not gapped). I will try to simulate a gapped and non-gapped toroidal inductor when I get a chance.

Do you know if your inductor is gapped?


=====================================
Eng-tips forums: The best place on the web for engineering discussions.

 

[electricpete]

Actually referring to the stationary motor test, the resistive component of the current of course does not show up in the computed effective inductance so the effect of saturation of the bridge upon inductance is more complicated than what I suggested above. ( by the way single phase stationary test is used to test for broken rotor bars based on variability of current as we rotate the rotor... GE has published info suggesting that variation of the depth of that bridge above the bar can give a false alarm when testing at lower excitation levels which goes away at higher excitation levels).

The motor stationary single phase test is a much more complicated system and I probably shouldn't try to analyse the complicated system to make conclusiosns about our simpler system. Sorry, I will stop talking about the motor and focus on the inductor.

=====================================
Eng-tips forums: The best place on the web for engineering discussions.

 

[electricpete]

I did not read your most recent post closely. Now going through it I have some comments questions.

The link you provided looks likes watts/lbm vs H. I don't think there is any way to extract B vs H from this.

I went through some 50Hz transformer core material characteristic which are more likely similar with the sample characteristic you have provided and I would say the desired "delta" measurement area with constant current method should begin at least at 0.4T inductance ( the transformers we are using are designed typically to work at 1.2T ).

I didn't understand what you meant by this. What did you mean by delta measurement area?

I agree with your assessment there is no obvious way to infer performance at other than rated conditions except to apply a correction to the rated condition performance based on the iron characteristics (I will assume from here on out that it is a non-gapped toroidal core unless you tell me otherwise).


An interesting link here:

http://www.scielo.br/scielo.php?pid=S1516-14392005000400007&script=sci_arttext#fig8

You see they derive the excitation curve using a current transformer (the core material may not be similar to yours, but qualitatively I think the same behavior applies).

Figure 4 is the B vs H curve.

Figure 5 is the "permeability" vs H which appears to be taken as the slope from the origin to any point on the B vs H curve. The maximum of the permeability curve occurs at the knee of the B vs H curve (around 5A/M). Inductance vs H would look similar to that permeability vs H curve.

Now one more thing to mention, the curve inductance vs excitation H should resemble the slope of the excitation curve for a CT. On a log-log scale, the CT excitation curve looks like a linear increase below the knee. From that piece of info we might mistakenly conclude the inductance (slope of the B vs H curve) is constant below the knee. The tricky part is that it is a linear increase when plotted on a log log scale but the slope on the log-log scale is not 1, so it is not a linear behavior (constant slope) when we plot a ct excitation curve on a linear scale.




=====================================
Eng-tips forums: The best place on the web for engineering discussions.

 

[electricpete]

http://www.sayedsaad.com/Protection/files/VT_CT/2_VT_CT.htm

Figure 8a at the link above is an example CT curve on a linear scale.

Figure 8b is on a log scale. Start at the very lower left hand corner (I,V)=(0.001,1). Now follow the line up to V=10. The corresponding I is less then 0.005 (it would be 0.010 if the curve were linear). This region of the curve is something like V~I^2, not V~I, even though it "looks" linear on this log-log scale.

=====================================
Eng-tips forums: The best place on the web for engineering discussions.

 

[electricpete]

I think that this type of log-log scale plot is why most people (myself included not too long ago) mistakenly thought that iron is linear below saturation.

=====================================
Eng-tips forums: The best place on the web for engineering discussions.

 

[electricpete]

A quick review of terminology.

The curve that I provided for M-22 is the "normal" magnetization curve or "normal" B vs H curve. For any given point on that curve, the slope from 0 to the point (B,H) is called the "normal permeability"

If you have sinusoidal excitation at some level, the operating point will trace a hysteretic loop (minor hysteresis loop) whose endpoints fall on that normal magnetization curve. If you push it well into saturation or beyond then the minor loop becomes the saturation loop, most often shown.

The free 2-D electromagnetic software FEMM estimates a sinusoidal response to the excitation by taking the normal BH curve and connecting an ellipse between the endpoints. The width of the ellipse is defined by a lag angle (B lags H) which is supposed to be from 0 to 20 degrees for soft steel – I used 10 degrees.

To my way of thinking, this should provide a good simulation below saturation where the hysteresis loop somewhat resembles an ellipse with pinched corners. Far above saturation this approach doesn't seem very good since the hysteresis loop does not resemble an ellipse at all. Also as we get far above saturation, the current waveform of course becomes less sinusoidal and it is difficult to model it in a single-frequency sinusoidal simulation.

The program was solved in FEMM, which was controlled by MATLAB, and results sent to excel.

Here is an excel file:

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

tab BvsHandNormaulMu has numerical data for the normal magnetiziation curve and the normal Mu_Effective (B/H for any value on the B vs H curve)

tab SimulationOutput has the numerical simulation output. For each value of current we compute an H based on geometry and turns. We also find B from simulation results. We compute mu_effective two different ways from the simulation results (very close)

tab chart1 has 4 curves.
We have two B vs H curves (whose axis is given on the right). One is the average tangential B within the core based on the simulation. The other is the input data B vs H

We have two mu_effective vs H curves (whose axis is given on the left). One is computed as B/H based on the FEMM-computed B. The other is computed as the ratio B/H for a given point on the BH curve ("normal permeability").

The peak of the simulation mu vs H curve is lower and shifted toward higher excitation compared to the peak of the "normal" mu vs H curve. I'm not sure what gives rise to this difference (interested in any comments)

Here is a powerpoint file:

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

slide 1 shows overview of the toroid geometry which was solved. Mangetic toroidal core has ID = 0.2m, OD=0.21M depth intp page =0.005m... cross section would be square. 32 1mm diameter conductor turns are wrapped inside and outside of this 2-d simulation. (neglecting leakage doesn't seem like a big problem until very high flux densities which saturate the core).


slide 2 shows the FE mesh
Slide 3 shows the same summary plot as excel,... here we label 3 points: lowH, medium H, high H. The flux profile for these three cases are shown in slides 5-7.

Here is a word file:

http://home.comcast.net/~electricpete/eng-tips/toroid8dotm.doc

This contains the matlab program which produced these results. If you didn't have matlab and wanted to run this program or create a simulation of your own, you could use free programs Octave or lua script language. The code including FEMM commans are very very similar except perhaps for the file handling commands.

What did we learn from this? Not a helluva lot. The results of the simulation are somewhat similar to results which could be extracted directly from the normal B-H curve. I learned how FEMM models hysteresis (the ellipse). I suspect the results up to saturation are pretty good since the system response (current and flux) are fairly sinusoidal up to that point. If we wanted to predict results far above saturation, I think we would need a time domain simulation for example ODE45 in the manner that I suggested 28 Jan 09 7:30. It would be a challenge to model hysteresis this way since the derivative is supposed to be computed from the values of state variables, not from their history. Maybe taking advantage of the fact that we know the applied voltage as a known function of time, we could use the time variable to help figure out what part of the curve we're on.


=====================================
Eng-tips forums: The best place on the web for engineering discussions.

 

[electricpete]

In fairness to FEMM, I have oversimplified my discussion of how they model hysteresis. The lag angle (10 degrees in my case) applies to the H where we have peak normal permeability. For any other point, the lag angle is reduced by ratio:
normal-permeability/peak-normal-permeability

More details and references on that approach are available within the FEMM literature:

http://femm.foster-miller.net/wiki/HomePage

=====================================
Eng-tips forums: The best place on the web for engineering discussions.

 

[homoly]

electricpete, thanks for the comprehensive entry.It will take some time to go through this but I will try out the FEMM software.
According the material I will try to search the B/H curve, my intention was to show what kind of material is this.

By delta I have meant the area from where the slope seems to be similar for BH curve like in working point. Delta because of slope ( derivation ) definition which is defined by X and Y axis differences ratio - so it is only my private, not the precise technical term.

It seems for the transformer irons if I get by current to 0.4T I can get very similar result ( slope ) as 1.2T with nominal current so this was my observation according materials used in our co. for mains 50 Hz transformer. So for this case the 0.4T should give then minimum required current needed for the measurement.