Author Topic: Inductors and skin effect on the Impedance Analyzer  (Read 921 times)

0 Members and 1 Guest are viewing this topic.

Offline The Electrician

  • Frequent Contributor
  • **
  • Posts: 254
Inductors and skin effect on the Impedance Analyzer
« on: February 17, 2013, 05:00:57 AM »
The same analyzer I used to examine capacitor characteristics can also measure inductive effects.

One very instructive measurement shows the increase in losses in a length of wire due to skin and proximity effect.

The image shows the AC resistance of 3 pieces of 18 gauge magnet wire, about 35 cm long..  Three measurements were made and superimposed on the display.  One piece is measured as a simple unwound piece of wire.   This is the bottom curve in the image.  The increase in AC resistance as the frequency increases is due to skin effect.

One piece of wire was wound into a single layer, close spaced, solenoid with about 1/4 inch inside diameter.  The measured AC resistance of this solenoid is the middle curve.  Notice how the AC resistance increases more than when the wire was not wound into a solenoid.

The third piece of wire was wound into a 3 layer solenoid; its measured AC resistance is the top curve.  The fact that there are 3 layers and not just one as in the second measurement is responsible for the more rapid increase in AC resistance at first, followed by an inflection point and a reduced slope.  This effect is called eddy current screening by Snelling in his classic book on soft ferrites, and is well known to designers of magnetics for switching power supplies.
« Last Edit: February 17, 2013, 05:02:48 AM by The Electrician »

Offline JuiceKing

  • Regular Contributor
  • *
  • Posts: 208
  • Country: us
Re: Inductors and skin effect on the Impedance Analyzer
« Reply #1 on: February 17, 2013, 05:07:19 AM »
Very neat!

What frequency is represented by the first bold line mark, roughly 1/4 of the way from the 100 Hz line?

Offline The Electrician

  • Frequent Contributor
  • **
  • Posts: 254
Re: Inductors and skin effect on the Impedance Analyzer
« Reply #2 on: February 17, 2013, 05:41:53 AM »
The horizontal direction is logarithmic.  At the bottom of the screen you can see that the start frequency is 100 Hz and the end frequency is 1 MHz.

The bold vertical lines are: left edge = 100 Hz; next bold vertical line = 1 kHz; next line = 10 kHz; next line = 100 kHz; right edge = 1 MHz

Look at the upper right of the display.  Both the A and B cursor have their frequency and the magnitude of the measurement indicated.

The A cursor is at 215.44 Hz, B cursor at 100 kHz.
« Last Edit: February 17, 2013, 05:46:45 AM by The Electrician »

Offline The Electrician

  • Frequent Contributor
  • **
  • Posts: 254
Re: Inductors and skin effect on the Impedance Analyzer
« Reply #3 on: February 17, 2013, 09:30:09 AM »
A common source of confusion for hobbyists is determining the impedance of the windings of an audio transformer.  It is often believed that the resistance measured with a DMM's ohmmeter function (the DC resistance) is the impedance of a transformer winding.  This is not true.  Let's make some measurements.

I'm going to be discussing some measurements on a transformer with various resistive loads connected.  I will draw some conclusions that might not be always true if, for example, unusual loads were connected such as a negative impedance or a capacitor.  I'm not considering such loads; only resistances.

I have a small audio transformer with two windings; it's rated for 100k ohms on one winding and 1k ohms on the other.  I'll call the 100k ohm winding the primary.  First of all, the DC resistance of the primary is 2012 ohms, and the DC resistance of the secondary is 71 ohms.  These resistances are far from 100k and 1k ohms, so apparently the rated impedance is not the DC resistance.

Setting the impedance analyzer to sweep from 100 Hz to 5 MHz, and plotting impedance versus frequency, we obtain the plot shown in the first attached image when measuring the 100 k ohm rated winding.  There are two curves shown; the top one is taken with the secondary open circuited and the bottom one with the secondary short circuited.  Now with a little thought we see that the impedance measured at the primary as shown can't be less than the bottom curve no matter what resistance we connect to the secondary (we can't apply less than zero ohms).  Likewise, the impedance at the primary can't be more than the top curve no matter what load is connected to the secondary (we can't apply more than infinite ohms).  So we see that the impedance measured at the primary must fall between the two curves shown no matter what resistance we connect to the secondary.

The second image shows a family of impedance curves (I've limited the high end of the frequency sweep) with various resistances connected to the secondary.  The bottom curve is taken with zero ohms applied to the secondary, and then successively increasing resistances are applied, namely 100, 300, 1000, 3000, 10000 ohms and finally an open circuit (infinite ohms?).

The rated impedance of the transformer primary is 100k ohms, but the measured impedance varies from about 800 ohms to over a megohm at 5 kHz.  Why don't we measure an impedance of 100k ohms, the rated impedance of the primary winding?  Because a transformer doesn't have a fixed, intrinsic impedance.  The impedance measured at a winding depends strongly on what is connected to the other winding.

Notice, however, that the middle curve is just about 100k ohms over a fairly wide frequency range.  That is the curve we got with 1k ohms connected to the secondary.  The rated impedance of the secondary is 1k ohms.  It looks like we get the rated impedance at the primary is we apply the rated load resistance to the secondary.  Further notice that the middle curve is about halfway between the top curve (secondary open) and the bottom curve (secondary shorted) in the midband region around 5 kHz.

This suggests a method to determine the rated impedances of an unknown transformer.  Plot two curves for one winding; one with the other winding open and one with the other winding shorted.  Draw an additional curve half way between those two curves and that will be very close to the rated impedance of the winding.

To find the rated impedance of the other winding, apply various resistances and find the value that gives a swept impedance halfway between the open and short impedances of the first winding.  Or, we could simply reverse the windings and plot two curves for the open and short condition of the primary while measuring the secondary.  The rated impedance will be halfway between those two curves.

These curves are plotted with a logarithmic scaling of the vertical.  A property of log plots is that at a particular frequency we will have a value for the open circuit impedance and another value for the short circuit impedance.  A point on the plot halfway between those two values will have a value equal to the square root of the product of the open and short circuit values.  We also know that that halfway point has a value essentially equal to the rated impedance of the winding.  This gives us another way to determine the rated impedance.

A transformer with two windings can be treated as a two-port.  A property of a two port known since the early days of Bell Labs, is that if you measure the impedances at one port with the other port successively open circuited and short circuited, and take the square root of the product of those two values, you will have a value known as the image impedance.

At a given frequency, measure the impedance at a winding with the other winding open and then short circuited.  Multiply those values and take the square root; that is the image impedance and it is the rated impedance at that frequency.  Because real transformers have parasitics (non-zero winding DC resistances, distributed capacitances), the image impedance is not constant with frequency.  We see that the transformer I've used has only a finite band over which its behavior is useful.  The impedance of the primary with a 1k ohm load should be 100k ohms; it satisfies this requirement only over a limited bandwidth, from a few hundred Hz to a few kHz.

Plotting the impedances over a band of frequencies allows us to not only see what the image impedance (halfway between the open and short circuit impedances) is, but to see over what frequency range it is nearly constant.

The third image is the same as the second, but with only the rated load applied to obtain the middle curve.  That curve should follow the 100k ohm line at all frequencies if the transformer were perfect.

Edit:  After all this, hobbyists aren't going to have an impedance analyzer, but we all know how much we want an LCR meter.  Here's another use for an LCR meter.  In the audio band, 1 kHz is nearly midband.  If you want to get a fairly good estimate of the rated impedances of an unknown transformer, just determine the image impedance of each winding (square root of the product of the open and short circuit impedances) at the single frequency of 1 kHz, and consider those impedances to be the rated impedances.  Also doing this at 10 kHz will give you an idea of the bandwidth of the transformer; the image impedances at 10 kHz ideally should be the same as at 1 kHz.
« Last Edit: February 17, 2013, 09:38:10 AM by The Electrician »

Offline amspire

  • Super Contributor
  • ***
  • Posts: 2333
  • Country: au
Re: Inductors and skin effect on the Impedance Analyzer
« Reply #4 on: February 17, 2013, 09:40:08 AM »
The tests on components have been fascinating, but I want to find out more about the analyzer. A photo and some specs.

Perhaps some info on how the analyzer actually works.

Offline The Electrician

  • Frequent Contributor
  • **
  • Posts: 254
Re: Inductors and skin effect on the Impedance Analyzer
« Reply #5 on: February 17, 2013, 09:51:03 AM »
A piece of transmission line can be treated as a two port (one end is the input port and the other end is the output port).

The image impedance concept applies equally to any two port, and that includes transmission lines.  If you want to know the characteristic impedance of a transmission line, measure the impedance at one end with the other end successively open and shorted.  Take the square root of the product of those two values and you have the characteristic impedance (at the frequency of measurement).

I have a 100 foot roll of old style telephone wire for in-house wiring.  It consists of 4 conducters; two twisted pairs.  I connected one end of one of the pairs to the analyzer and plotted impedances.

The first image shows the impedance at one end with the other end open circuited.  The second image shows the impedance at one end with the other end short circuited.

The third image shows the first two images superimposed.  Remember what I said in the previous post about the image impedance on a logarithmically scaled plot being halfway between the open circuit curve and the short circuit curve.  You can see the halfway points on the third image occur where the curves cross.  The impedance there is very close to the dotted 100 ohm line in the middle of the screen.  This verifies what is often said, that the characteristic impedance of a twisted pair phone line is approximately 100 ohms.

Offline The Electrician

  • Frequent Contributor
  • **
  • Posts: 254
Re: Inductors and skin effect on the Impedance Analyzer
« Reply #6 on: February 17, 2013, 10:00:19 AM »
The manufacturer's web page is here: http://www.hioki.com/newproduct/im3570/im3570_e.html

As far as how it works, it applies a sinewave current or voltage at a particular frequency and measures the voltage or current that results.

This Agilent document explains it quite well: http://cp.literature.agilent.com/litweb/pdf/5950-3000.pdf

Offline amspire

  • Super Contributor
  • ***
  • Posts: 2333
  • Country: au
Re: Inductors and skin effect on the Impedance Analyzer
« Reply #7 on: February 17, 2013, 11:01:40 AM »
The manufacturer's web page is here: http://www.hioki.com/newproduct/im3570/im3570_e.html

As far as how it works, it applies a sinewave current or voltage at a particular frequency and measures the voltage or current that results.

This Agilent document explains it quite well: http://cp.literature.agilent.com/litweb/pdf/5950-3000.pdf

The Agilent document is great. I knew how the 100KHz LCR meters work, but I wasn't sure how they did the I to V conversion up in the MHz region.  Looks like it is done with a null detector on the 0V lead of the component, and by applying a signal to exactly balance out the current through the device. Simple! ???

Offline saturation

  • Super Contributor
  • ***
  • Posts: 4205
  • Country: us
    • NIST
Re: Inductors and skin effect on the Impedance Analyzer
« Reply #8 on: February 18, 2013, 01:41:14 AM »
Very nice bit of kit, Electrician.  For 1-5% accuracy a user can make a poor man's impedance analyzer with just a scope and a function generator then do math for the conversions.  Naturally, one needs to be careful of various parasitics caused by the cabling and connectors.

http://www.eevblog.com/forum/beginners/commercial-capacitor-tester-gizmo-or-my-scope-and-signal-gen-combined/msg87586/#msg87586
Best Wishes,


Saturation


Share me

Digg  Facebook  SlashDot  Delicious  Technorati  Twitter  Google  Yahoo
Smf