Sep/Oct 2005 41

9957 N River Rd
Mequon, WI 53092
[email protected]
An LMS Impedance Bridge
By Dr George R. Steber, WB9LVI
Come learn about LMS impedance measurement and
build a unique PC sound card impedance bridge.
Q
uite some years ago, I built an
impedance bridge using a rela-
tively new (at that time) DSP
microprocessor, the TMS32010. It was
based on a technical paper
1
that used
the LMS (least mean square) algo-
rithm. Results of that project were
mixed. It was helpful because it veri-
fied that the LMS algorithm could be
used for real-time impedance mea-
surements. But it was disheartening
because it was not very accurate and
could only operate at a maximum
bridge frequency of 50 Hz. So I filed it
away for future reference.
Over the years Ive maintained an
interest in impedance measurement,
digital signal processing (DSP) and
programming. Recently I did a novel
project
2
using a PC with a sound card
and some DSP techniques to imple-
ment a low-cost curve tracer (I versus
V) for devices like Zeners, LEDs and
transistors. During that task I some-
times thought of the old LMS bridge.
Could it be implemented on a PC? Af-
ter a lot of study and some serious
modifications, the answer is a re-
sounding yes. In fact, it turned out
to be much more than I had hoped for,
yielding a wide range, low-cost imped-
ance measuring system. An abbrevi-
ated article describing construction
and operation of the LMS bridge has
been written for QST.
3
Presented here
are the technical details behind this
unusual system, its operation and
some additional practical applications.
In case you dont have the QST article
handy, I will also present some mate-
rial on installation and operation of
the bridge.
Nearly any PC can be used in this
project, as there is no need to modify
in any way. You can use one of the
newer 3 GHz PCs or dust off that old
200 MHz PC thats sitting on the shelf.
No need to open the cabinet either
since access to the sound card stereo
line jacks is all that is required, and
that can usually be accomplished from
a panel on the rear of the computer.
Of course I am not giving out guaran-
tees that this project will work with
your system, but I will say that I have
tested it with a 200 MHz Pentium Pro,
a 500 MHz Pentium III and a 1.1 GHz
AMD Athlon processor running
Windows 98 or XP with a Sound
Blaster (SB) Live! sound card.
So, if you have a Pentium or AMD
PC with a Windows compatible full-
duplex sound card, you may have the
basis for a very good Windows based
impedance measuring system. All you
need to do is build the simple circuit
described, connect it to your computer
sound card and run the program. This
impedance bridge allows you to auto-
matically measure inductors, capaci-
1
Notes appear on page 47.
42 Sep/Oct 2005
tors, resistors, input impedances,
audio transformers, negative resis-
tances and more at a wide range of
audio frequencies. It has outstanding
capabilities and accuracy.
The cost of the project is less than
$1 (yes, one dollar!) not counting the
PC and power supplies. The circuit
uses only two resistors and a dual op
amp. It can be built on a solderless
breadboard like I did or you can de-
sign a printed circuit board for it. In
any case you will need a digital volt-
meter (DVM) for calibration purposes,
although even this is not absolutely
required.
As usual, I am getting ahead of the
story. As a professor, now retired, I am
obliged to present more background
and theory on this subject. I think you
will find it interesting, so please try
to resist the urge to skip to the end of
the article.
Impedance
Impedance is basically the opposi-
tion to current flow. It is a more
general form than resistance alone.
Impedance can have a resistive part
and a reactive part. For resistors, the
reactive part is very small, unless
theyre wire wound. Inductors and ca-
pacitors have both resistive and reac-
tive parts. The resistive part is often
modeled in series with the reactance.
Although the series model is used
here, other models are sometimes used
with parallel resistors. At a given fre-
quency, impedance can be written in
either polar (vector) or rectangular
form as in (Eq 1).
jX R Z Z + = = (Eq 1)
where Z is impedance in ohms, |Z| is
the magnitude of Z, is the angle of
Z, R is the real (or resistive) part of Z,
and jX is the imaginary (or reactive)
part of Z. The two forms in Eq 1 are
related by:
2 2
X R Z + =
and

=

R
X
tan
1

(Eq 2)
As noted above, impedance can also
be written in other forms but in this
article we will always model the un-
known impedance as a series combina-
tion of resistance R and reactance X.
There are numerous instruments
available to measure impedance in-
cluding the ubiquitous ohmmeter for
resistors, resistance bridges, ac
bridges for capacitors and inductors,
automatic LCR bridges and vector im-
pedance meters.
Learning From The Old LMS
Bridge
Since many of the ideas for the cur-
rent project were derived after look-
ing at the problems of the original
LMS bridge project, we will look at it
first. It is shown in Fig 1. The signals
are all sampled signals, but we will
not denote that at this time for sake
of clarity. In this bridge, Vr and Vx are
two sinusoidal voltage sources with
the same radian frequency
0
, but with
different amplitudes and phase shifts.
Reference voltage source Vr is of con-
stant amplitude A and zero phase
shift. However, Vx has a variable am-
plitude and phase shift. They can be
written as follows:
Vr = Asin(
0
t)
Vx = Bsin(
0
t + ) (Eq 3)
The parameters B and are con-
trolled to balance the bridge. Vr and
Vx are generated via D/A (digital to
analog) converters from the micropro-
cessor. Voltage e(t) is read into the
microprocessor with an A/D (analog to
digital) converter. Other elements of
the bridge are the unknown imped-
ance Z
X
and the reference resistance
Rm. When the bridge is balanced (volt-
age e = 0) the unknown impedance is
given by
=
A
B
Rm Z
x
at the frequency
0
(Eq 4)
Expressing V
X
in terms of in-phase
and quadrature components yields:
V
X
= B cos sin (
0
t) + B sin cos
(
0
t) = W
1
A sin(
0
t) + W
2
A cos(
0
t)
Fig 2Impedance measuring circuit and interface to sound card. See text for more
information.
Fig 1Old LMS bridge. Zx is the unknown
and Rm is reference resistor.
(Eq 5)
where W
1
= (B/A) cos and W
2
= (B/A)
sin are the weights of the in-phase
and quadrature components, respec-
tively. With B and expressed in terms
Sep/Oct 2005 43
of W
1
and W
2
, Eq 4 can be written as:
balance at jRmW RmW Z
x 2 1
+ =
(Eq 6)
The terms RmW
1
and RmW
2
are the
real and imaginary parts of Z
X.
To bal-
ance the bridge, one starts with ini-
tial values of W
1
and W
2
and itera-
tively modifies these to force e(t) to
zero. One method of doing this, the
LMS algorithm, requires that the er-
ror be found at each new sample and
updated values of W
1
and W
2
be calcu-
lated that hopefully force e(t) to zero
over time. The LMS algorithm does
that well but we wont go into how it
does it right now.
First we observe a few things about
this bridge. On the plus side we see
from Equation 6 that at balance, the
real and imaginary parts of Z
X
depend
only on the weights and Rm (reference
resistor) and there is no requirement
to know the amplitude A. That is a nice
result. On the minus side, since the
LMS algorithm requires that calcula-
tions occur at each new sample of e(t),
we have only one sample period to
find new values of W
1
and W
2
. In
addition, the sampling of e(t) and the
outputs Vr and Vx must be synchro-
nized. Note too that since the sources
Vr and Vx are generated digitally they
must be calculated for each new point
on the sine waves. Finally, we see that
the unknown, Z
X
is floating above
ground, which is not desirable.
Consider now a typical sound card
in a PC. It has very good 16 bit A/D
and D/A converters. We can easily gen-
erate the signals Vr and Vx and out-
put them to the bridge via the sound
card line outputs. Similarly, the error
signal e(t) can be read with a line in-
put. But how do we keep the inputs
and outputs synchronized within one
sample period? If you are familiar with
Windows you probably know that all
sound card I/O is done via buffers. And
Windows decides when to empty and
fill them. Because the input and out-
put buffers may be long and not syn-
chronized this argues against using
the LMS algorithm, which needs to
make sample to sample decisions. I
wrestled with this problem, but could
make no headway until I started look-
ing at different circuit topologies.
Impedance Measuring Circuit
Consider the circuit in Fig 2. In-
stead of using two sound card outputs
and one input, as in the old LMS
bridge, it uses one output and two in-
puts. By intentional design, the two
inputs are synchronized with each
other but not necessarily to the out-
put from the sound card. Resistor R1
is there to provide a ground reference
for the sound card output. Rm is the
reference resistor and Zx is the un-
known, as in Equation 1. We see that
the unknown impedance is now
grounded. Two op-amps U1A and U1B
provide isolation and buffering of the
bridge voltages. They are connected as
unity gain, high input-impedance, low
output-impedance drivers. Vr is the si-
nusoidal voltage applied to the circuit
via a line output of the sound card. It
is fed back to the sound card input
right channel via U1A. The voltage
across the unknown Zx is buffered by
U1B and fed back to the sound card
input left channel via U1B. Shielded
audio cables are suggested for connect-
ing to the sound card.
U1 is a cheap LM358 dual op-amp
or equivalent which can be powered
from bipolar power supplies of 3 volts
to 15 V. It is best to keep the power
supply voltages low to protect the
sound card line inputs in case of prob-
lems. My circuit runs at about plus
and minus 3 V. A bipolar battery sup-
ply could also be made using four AAA
batteries with a center tap connected
to ground.
Since all measurements depend on
Rm, it is critical that we know its re-
sistance precisely. To effectuate differ-
ent ranges of the instrument, we may
also wish to suitably change the value
of Rm. We will discuss more about this
later on.
There are many ways to measure
impedance with the circuit of Fig 2.
All of them require that a sinusoidal
signal Vr be applied to the circuit and
a number of sequential samples of Vr
and Vx be captured in buffers. Dis-
cussed below are several methods of
doing this.
Three Measurement Method
An old method of calculating im-
pedance, sometimes called the three-
voltmeter method, can be used. Here
is how it works. Measure and record
the three voltages Vr, Vx and Vrm as
shown on Fig 2. We see that
= = Z Rm
V
V
Z
rm
x (Eq 7)
To find the magnitude of Z, simply
take the magnitude of Vx divided by
magnitude of Vrm and multiply by Rm
as shown below.
plying the law of cosines results in the
equation for the phase angle as
shown below.
Rm
V
V
Z
rm
x
=
(Eq 8)
where T stands for transpose. So X
k
is
actually a column vector and the sub-
script k is used as a time index. Simi-
larly, we define a weight vector

180
2
180
2 2 2
1
x
x
V Vrm
V Vrm Vr
cos
It is a little more complicated to
show, but considering the three volt-
ages as vectors in a triangle and ap-
in degrees
There are other ways to calculate
that involve multiplying the voltages
and filtering, but they dont provide
more benefits. A drawback to this way
of finding is that it cannot distin-
guish between positive and negative
angles of reactance. However, the cor-
rect can be found by looking at the
zero crossings of Vrm = A sin(t) and
Vx = B sin (t + ) since at t= 0 (and
multiples of the period), Vx = B sin
which is > 0 for inductances and < 0
for capacitances.
This method was implemented on
a PC with good results over much of
the measuring range. If you looked
no further, this would be a an accept-
able method of finding impedance.
When either Vx or Vrm is small,
however, there is a substantial er-
ror in . This was attributed to noise
and other errors from the measure-
ment of these voltages. More sophis-
ticated methods can reduce these
errors.
Least Squares Method
The literature is full of different
error-minimization criteria but the
most widely used one is the least
square approximation originated by
Gauss. Simply stated, the least
squares principle involves selecting
the function that minimizes the sum
of the squared errors. A more com-
plete discussion can be found in
Reference 4. It is applicable to both
continuous and discrete systems.
Since we will be dealing with discrete
variables for the rest of this discus-
sion, it is appropriate to introduce
them now. Fig 3 shows a adaptive lin-
ear combiner we will use to illustrate
the least squares method. It consists
of unit-time delays, weights and sum-
mation blocks. There are two inputs:
the data samples x
k
and the desired
response d
k
. Sample delays are rep-
resented by z
1
with the samples taken
at points k, k1,, kL, going back
in time through the data samples.
The L-input samples, x
k,
x
k1
,
x
k-L
may be represented as a vector.
X
k
= [x
k
x
k1
. . . x
kL
]
T
(Eq 10)
(Eq 9)
44 Sep/Oct 2005
W
k
= [w
0k
w
1k
. . . w
Lk
]
T
(Eq 11)
From Fig 3, we see that the output,
y
k
,

is a linear combination of the in-
put samples and the weights. The
error signal with time index k is the
difference between the desired re-
sponse d
k
and the output y
k
and is
given by:
e
k
= d
k
y
k
= d
k
W
T
X
k
(dropping the
subscript of W for clarity) (Eq 12)
and
e
2
k
=d
2
k
+ W
T
X
k
X
T
k
W 2d
k
X
T
k
W
(Eq 13)
This is the instantaneous squared
error and is the function we wish to
minimize by adjustment of the
weights, W. This rather imposing task
can be attacked via two main meth-
ods, the non-recursive approach called
the Wiener-Hopf method or the recur-
sive approach of the Widrow-Hopf
(LMS) method. Solutions presented in
Note 4 will be used.
A closed-form solution for the
Wiener-Hopf method can be writ-
ten as:
P R * W
1
= (Eq 14)
where R = E[X
k
X
T
k
], P = E[d
k
X
k
], E
denotes taking the expected value, and
W* is the optimal weight vector. The
procedure is straightforward. We cap-
ture L samples, calculate the expected
values for R and P and simply calcu-
late W*. Surprisingly, in practice, it
works quite well.
The LMS method proceeds by first
calculating the error as in Eq 12 and
then applying a steepest-descent
algorithm to find W
k+1
as:
W
k+1
=W
k
+2
k
X
k
(Eq 15)
where is a constant that controls the
speed and stability of adaptation. This
deceivingly simple recursive equation
obtains the same solution as Weiner-
Hopf, when converged.
Adaptive Impedance Bridge
Refer again to Fig 2 and notice that
ear combiner for this case where Vrm
is the input, Vx is the desired signal
and e
k
is the error. The unknown im-
pedance is given as:
2 1
jRmW RmW Z + = (Eq 18)
Both the Weiner-Hopf and LMS
methods were implemented and simu-
lated for comparison of accuracy and
convergence using real data captured
from a sound card. Many different
cases of unknown impedances, whose
values were precisely measured on a
commercial LCR bridge, were tested.
A sound card sampling frequency of
44,100 samples per second and a si-
nusoidal frequency of 1225 Hz was
used for the test signal. Each channel
was set to a capture length of 11,025
samples, which provided 0.25 seconds
of data. As expected, both methods
tended toward the same, and correct,
solution for Z.
The Weiner-Hopf method provided
the best (lowest) error for this num-
ber of samples in most cases. The LMS
method varied in error depending on
the amplitude of the signals and the
value of . If nothing further were con-
sidered, Weiner-Hopf would be the
method of choice. However, further
experimentation showed that if the
sampled data in the capture buffer
was re-iterated several times, the LMS
method improved dramatically. This,
in effect, increases the number of
samples. Since the LMS algorithm is
so efficient, only a few milliseconds is
added to the calculation. A similar re-
iteration for the Weiner-Hopf method
does not provide additional improve-
ment and would require a larger cap-
ture buffer. Finally, the LMS algorithm
was normalized to improve the speed
of convergence over a wide range of
signal levels.
Both methods are very good and
greatly surpass the three-measure-
ment method described earlier.
Since I wanted to keep the cap-
ture length small (in order to have
about 4 measurements per second)
I chose to implement the LMS
method.
Fig 3An adaptive linear combiner.
Fig 4An adaptive combiner for the LMS bridge.
Z
Rm
Vrm
Z
Rm
Vx Vr
Vx =

=
(Eq 16)
Sound Card Considerations
A low-distortion, low-noise full-du-
plex sound card is desirable. The
Sound Blaster Live fills the bill nicely
where Vrm = VrVx. Now, let Vrm =
A sin(
0
t) and Vx = B sin(
0
t + ). Fol-
lowing the same procedure as for the
old LMS bridge, Vx can be written as:
( ) ( ) t cos A W t sin A W Vx
0 2 0 1
+ =
(Eq 17)
To find the weights, we apply the
methods described in the previous sec-
tion. Fig 4 illustrates the adaptive lin-
Sep/Oct 2005 45
and probably many others will too.
Since I cannot test them all, I will re-
strict my attention to this one. Refer-
ring to Fig 2, we see that A1 is the line
output amplifier, and A2 and A3 are
the right-and left-channel line inputs
of the sound card. These amplifiers can
be a source of distortion if the proper
levels are not maintained.
There are two culprits here: One is
excessive drive and the other is satu-
ration. If A1 sources too much current,
it will distort. I viewed Vr on my
Tektronix TDS 360 (Digital Real Time
Oscilloscope with FFT) and saw a lot
of second- and third-harmonic distor-
tion when Vr exceeded 820 mV with
Rm =10 and Z= 0 (a short circuit).
Since the full output level is 1.62 V, it
needs to be attenuated. This can be
done either with the sound card mixer
or with the signal generator level con-
trol in my program. I chose to set the
Play level to maximum in the mixer
and set the level to 0.5 in the program,
as it is easier to remember. One of the
nice things about the LMS bridge is
that you dont need to know the am-
plitude of Vr.
The other consideration is that the
line input amps A1 and A2 will satu-
rate if the input voltage is too high.
This is so regardless of the Record set-
ting in the mixer. On the SB this oc-
curs at 820 mV. Since we are using
gains of one in the circuit, we can
prevent this by adjusting Vr as noted
above. So, for my SB Live, I just set
the output level to 0.82 V and both
conditions are satisfied. Just in case,
I have provided a real-time oscillo-
scope function in the bridge display,
so the sine waves can be monitored for
possible flat topping. If a digital volt-
meter is handy, Vx can be measured
and used to calibrate the scope. This
is provided in the program, but it is
not required and does not affect op-
eration of the bridge.
One other consideration is the bal-
ance of the two input channels. Since
we need to calculate VrVx precisely,
the two channels must be balanced.
This function is also provided in the
calibration section of the program.
A note is in order about earlier
Sound Blaster sound cards such as
the SB16 and AWE 32, since there
are so many of these still in ser-
vice. Unfortunately, they do not
provide true, full-duplex opera-
tion. The same may be said of SB
compatible cards, so be wary. For
example, (with the latest drivers)
the SB AWE32 can simultaneously
pl ay onl y unsi gned 8 bi ts and
record signed16 bits. It also has a
built-in amplifier that may over-
drive the LMS circuit. After some
extensive tweaking, I managed to
get one working with this program,
but the results were not as good. I
advise you not to use any of these
cards.
Impedance Bridge Installation
and Operation
The LMS bridge software is avail-
able on the ARRLWeb site and is
zipped for fast downloading. Unzip it
to a new folder and you are ready to
go. Just run the EXE program. It was
tested with Win98 and XP. Fig 5 shows
a screen shot of the bridge; there is a
lot of information on the screen. The
monitoring scope with its controls is
on the left side. The most important
part is in the lower-right corner la-
beled UNKNOWN, where all of the
relevant data about the measured
impedance is displayed.
The bridge is easy to use but some
considerations are in order. The value
of Rm must be known exactly, as all
results depend on it. Strive for 1% ac-
curacy here (or better) and do not use
an inductive resistor; carbon or film
types are fine. Although the bridge has
a wide range, it is best to keep the lev-
els of Vx and Vrm reasonable. The soft-
ware scope helps monitor these levels.
To reap the maximum benefit of the
bridge, Rm should be selected for the
approximate range of impedance. For
example, if Rm is10 , that is the ap-
proximate impedance to measure. This
value can be chosen by trial and error
or by making educated guesses, just
as with most bridges.
In general, you should be able to
measure over a 0.01 to 100 range
based on Rm. The chart in Table 1
illustrates the range you can expect
for a given Rm at 1225 Hz. ( The bridge
frequency will affect this range.) From
the chart, if Rm =10 , you can mea-
sure L between 12 H and 129 mH,
and C between 1299 F and 0.129 F
at 1225 Hz. The software has provi-
sions to store several values of Rm.
Just make sure that is the value actu-
ally in the circuit.
At the high and low ends of the
bridges range, stray capacitance and
inductance start to play a role. These
values can be compensated out by us-
Table 1Range of Bridge for various Rm with a Bridge Frequency of
1225 Hz. (See text)
Rm () L C
10 12.99 H to 12.99 mH 0.1299 F to 1299 F
100 129.9 H to 129.9 mH 0.01299 F to 129.9 F
1 k 1.299 mH to 1299 mH 1.29 nF to 12.9 F
10 k 12.99 mH to 12.99 H 1299 pF to 1.29 F
100 k 129.9 mH to 129.9 H 12.99 pF to 0.129 F
Fig 5Main window of the LMS impedance bridge program.
46 Sep/Oct 2005
ing the bridge, itself, to measure them
with the unknown impedance being
an open circuit and short circuit, re-
spectively. These values, called tare,
are then automatically used to com-
pensate the result. For example, in my
circuit there is about 14.1 pF of stray
capacitance with Z open and Rm =
100 k. So, that value is what I enter
for C tare in the program. Obviously,
this only makes a difference when
measuring small capacitors. Similar
compensation may be made for the
inductive wiring, when Z is near zero.
Once you have balanced the stereo
channels and (optionally) calibrated
the scope, you are ready to start mea-
suring. By the way, all calibrations are
saved so that you only need to do it
once. Next, the mixer that came with
your software or the one that came
with Windows needs to be checked, as
you may have changed its settings.
When you start the program, a little
notice comes on the screen to remind
you about this.
Basically, you want to set the out-
put level, input gain and stereo bal-
ance. The details of how to do this vary
from system to system. Here is how
its done with the SB: In the mixer Play
section, enable Wave and Spkr, set
the sliders to their maximums and
mute all others (including Line to
avoid audio feedback). In the Record
section, enable Line, set it to maxi-
mum and mute all others. Set the ste-
reo balance to center for all controls.
Using The Impedance Bridge
Measuring impedances is very easy
with this bridge, but its helpful to
think about what you are doing, so you
dont misinterpret results. Dont try to
measure a 100 pF capacitor with an
Rm of 10 . Assuming that you have a
ballpark Rm, connect the unknown
to the Z points of the circuit shown on
Fig 2 and click the Start button on
the screen. The bridge will automati-
cally determine whether the reactance
is capacitive or inductive at the mea-
suring frequency. Several items are
calculated and displayed, including
the real and reactive parts of Z, the
magnitude and angle of Z, the L or C
value of the component and its Q or D
factor. The scope is handy for looking
at the relative magnitude and phase
of Vx and Vrm and to see if you have
reasonable levels.
If you arent familiar with measur-
ing impedance (and even if you are)
you may run into some situations that
are unusual or seem to give inconsis-
tent values. As a guide, remember that
the value of an impedance is usually
defined only for a given frequency and
may be different at other frequencies
or signal levels. If you are measuring
a resistor, you will find it has some re-
actance and it will show up in the box
on the screen as either a capacitor or
inductor. Since you know its a resis-
tor, just look at the magnitude of Z or
real part of Z.
You are not limited to just L and C
components; the bridge can measure
all kinds of impedances including in-
put impedance, audio transformer
impedance, speaker impedance, sole-
noid impedance and even negative
resistance. Some of these topics are
covered in the following sections.
Measuring Input Impedance
The input impedance of an ampli-
fier or other circuit may be found by
connecting it to the unknown Z termi-
nals as shown in Fig 6. Take care not
to overdrive the amplifier input being
measured. This can be controlled by
adjusting the output level in the mixer
and choosing a suitable value for Rm.
Note, too, that impedance may vary
greatly with frequency, so try several
bridge frequencies.
Heres an interesting special case:
If you want to measure the input im-
pedance of the left channel of your
sound card line input, do the follow-
ing. Bypass U1B and connect the lower
end of Rm directly to that channel in-
put. This makes the input impedance
the unknown Z. On my SB, the input
impedance Zin measured 28.2 k
6.41 at 1225 Hz (with Rm = 1 k).
Measuring Transformer or
Speaker Impedance
The impedances of audio trans-
formers or speakers can be found by
simply connecting them to the bridge
as shown in Fig 7. For speakers, use
Rm = 10 to get started.
As an example, I measured the
reflected impedance of a small audio
transformer (Mouser 42MC003, 1.2 k:
8 ). I connected the primary to the
unknown Z terminals with an 8.2
resistor connected to the secondary,
as shown in Fig 7. The bridge read
1.24 k 13.78 at 1225 Hz (with Rm =
1 k). The reflected impedance de-
pends on the load connected to the
secondary, so you can experiment with
different loads to see the effect.
Measuring Large Electrolytic
Capacitors
While I was working on this project,
the power supply capacitor in my old
oscilloscope went out. The replace-
ment, a 1000 F unit, was too great
for my C meter. So, I put Rm=10 in
the bridge and measured it easily. Be
careful with these kinds of capacitors
and make sure they are discharged
before measuring them.
Large capacitors can often have
leakage and internal resistance. Its
interesting to see how the capacitance
of an electrolytic capacitor changes
with frequency. A junk box capacitor
measured 15.62 F at 120 Hz, and it
read 14.35 F at 1225 Hz. By the way,
it was marked 22 F and thus outside
its minus 20% tolerance.
Measuring Iron-Core Inductors
Iron-core inductors can give some
strange results. First measure a small
air core inductor (150 H, or so) and
vary the output level of the line
by varying the volume out in the
Windows mixer. Note that the read-
ing for inductance barely changes;
only due to noise and harmonic dis-
tortion errors. Now measure a small
iron-core inductor and do the same
thing. You will notice that the induc-
tance decreases with signal level. This
unexpected result (at least for me) re-
quired some research. It turns out that
Fig 6Setup for measuring circuit input
impedance.
Fig 7 Setup for measuring transformer
or speaker impedance.
Sep/Oct 2005 47
the permeability of the iron starts out
low for small currents and becomes
higher and more constant (linear
range) with increasing current. Since
L is proportional to the permeability,
there is a reduction in L at low cur-
rents. If you continue to increase the
current beyond the linear range, the
iron will saturate (I knew that!) and
L will decrease. This last part is of con-
cern to those who design switching
regulators. Reaching such large cur-
rents is not within the capabilities of
the bridge as it stands.
Another interesting thing about
inductors is that they can have a re-
sistive component that may vary with
frequency. Using the transformer pri-
mary as shown above, with the sec-
ondary open, illustrates this. As the
bridge frequency was varied from
525 Hz to 2205 Hz, the resistive part
varied from 1.14 k to 3.93 k. Prop-
erly terminated with 8.2 , it only var-
ied from 1.06 k to 1.26 k over the
same frequency range. Solenoids ex-
hibit similar behavior and their im-
pedance varies with plunger position.
Measuring Negative Impedance
Its possible to build a negative-
resistance circuit. An article in EDN
Magazine (Reference 5) about using
negative resistance prompted me to do
this experiment. Connect the negative
impedance converter shown in Fig 8
as the unknown impedance Z. The op
amp can be an LM358. Make sure R
is less than Rm or else you will cancel
it and cause oscillations. I used Rm =
1000 and R = 470 . The bridge in-
dicated minus 470 . Be cautious with
circuits of this type as they are likely
to oscillate.
Final Thoughts
This program runs fine on Win98
and Win XP. It was written in Visual
Basic 6.0. When you run the software
you may get a message like Required
DLL file MSVBVM60.DLL was not
found. This is a Visual Basic run time
file and is already on many systems.
If it is not found, you will need to ob-
tain it and install it on your system.
It is freely available from Microsoft
and other sites on the Web. It is usu-
ally available as Visual Basic 6.0 SP5:
Run-Time Redistribution Pack
(VBRun60sp5.exe) and is a self-
extracting file. The download takes
about six minutes at 28.8 kbps.
If you just want to experiment with
the program, dont worry as it does not
modify the registry or install any other
material on your computer. You can
remove it by just deleting the entire
George R. Steber PhD, is emeritus Pro-
fessor of Electrical Engineering and
Computer Science at the University
of Wisconsin-Milwaukee. George,
WB9LVI, is a life member of ARRL and
was awarded the QST cover plaque in
May 1975 for a ground breaking ar-
ticle on digital slow scan TV. Dr Steber
has considerable industrial experience
as a corporate officer, consultant and
product designer, with 18 patents is-
sued. In his spare time he enjoys rac-
quetball, reading, playing his Bach
trumpet, editing video and astronomy.
He recently restored a previously lost,
badly damaged NBC Tonight Show
program for the Kate Smith Com-
memorative Society. You may reach
him at [email protected] with
LMS in subject line.
Notes
1
M. Dutta, et al, An Application Of The LMS
Adaptive Algorithm For A Digital AC
Bridge, IEEE Transactions on Instru-
ments and Measurements, Vol IM-36,
pp 894-897, (Dec 1987).
2
G. Steber, Tracing Current and Voltage,
Circuit Cellar Magazine, Vol #162, pp 56-
61, Jan 2004.
3
G. Steber, Low Cost Automatic Impedance
Bridge, QST (Oct 2005).
4
B. Widrow et al, Adaptive Signal Process-
Fig 8A negative resistance converter.
Choose R < Rm to avoid oscillations.
folder where it is located.
Numerous components were mea-
sured on a commercial LCR bridge and
compared to the LMS bridge. Said
LCR bridge was rated between 1% and
5% depending on range, component
type and frequency. Good agreement
was achieved between the two, better
than 1% in many cases, a notable ex-
ception being small iron-core induc-
tors (see comments above), which var-
ied more. With any luck you will find
similar results. In any case, dont be
impeded in your quest for a fine LMS
bridge. Perhaps this is one you cant
resist.
ing (Prentice-Hall ISBN 0-13-004029-0,
1985).
5
E. Simons, Negative Resistance Load
Canceller Helps Drive Heavy Loads,
Electronic Design Magazine, March 19,
2001.

New Theories with Interpretations. Read about The Death of Modern
Gravity Theory, Electricity, Flow of Electrons or Magnetism?,
Electromagnetic Pulses or Waves?, Will an Object Launched into
Space Ever Stop?, Distance and Time Are They the Same?,
Electromagnetic Pulse Speeds, and much more.
N
e
w



S
c
i
e
n
c
e


B
o
o
k
H
a
r
d
c
o
v
e
r
P
h
y
s
ic
s

-

A
s
t
r
o
n
o
m
y

-

S
c
ie
n
c
e
s
To order: Fax 972.874-0687 or send order to:
Walter H. Volkman W5OMJ
P.O. Box 271797
Flower Mound, TX 75027-1797
$16.00 Postpaid USA
$24.00 Postpaid Foreign Airmail
30 Day Money Back Guarantee
You must be satisfied with book or return postpaid for full refund of
purchase price. No questions asked.