BITS-Pilani K.K.

Birla Goa Campus, Physics Department

PHY F344: Advanced Physics Laboratory

Noise filtration with a Lock-in amplifier

Material supplied: 1) Lock-in amplifier, 2) Function generator, 3) Oscilloscope, 4)
Cables

Purpose of the Experiment:

To get familiar with the use of a lock-in amplifier, which is a device used to reduce
the unwanted noise and amplify the desired signal and thus increase the overall
signal to noise ratio (S/N) of the electronic signal being measured. For this purpose,
we shall carry out the following:
1) Determination of the gain-bandwidth characteristics of the preamplifier.
2) Determine the characteristics of various filters (Chebyshev, Butterworth,
Bessel, and narrow band pas filters of Q = 5 and 50.)
3) Study the phase sensitivity of lock-in detection of signal.

Basic Methodology:
For the first part, the gain vs. frequency curves will be determined for two gain values
(G=1 and G=10).
In the second and third part, the gain of the amplifier will be determined as a function
of frequency for the three special filters (Chebyshev, Butterworth, Bessel) and two
narrow band pas filters (Q = 5 and 50.).
In the fourth part, we change the phase of the reference signal and study its effect
on the lock-in amplifier output signal (phase sensitivity) , on the oscilloscope.

Theory of a Lock-in amplifier:

A lock-in amplifier can be thought of as a phase sensitive ac voltmeter. Along with

the input signal, one has to supply a periodic reference signal to the device. The pre-
amplifier then responds only to the portion of the input signal that occurs at the
reference frequency and has a fixed phase relationship with the reference.

To understand the working principle of a lock-in amplifier, consider a sinusoidal input

signal VS(t) =Vo sin(ωt +φ). Let VR(t) = sin(Ωt) be the reference signal. When these
two signals are mixed electronically, the product of these two gives beats at the sum
and difference frequencies : VS(t) VR(t) = 2Vo {cos[(ω –Ω)t + φ] − cos[(ω + Ω)t + φ] }.
So, when the input signal has a frequency (ω) different from the reference frequency
(Ω), the product oscillates in time with an average value of zero. However, if ω = Ω,
we get a sinusoidal output (Vs), offset by a DC (zero frequency) level as:
VOut(t) α VS(t) VR(t) = 2Vo { cosφ − cos[2Ωt + φ ] } .
So, if one is able to adjust the reference frequency (Ω) with the signal frequency (ω)
and make the phase φ=0, one can extract the DC component of this product and get
a direct measure of the signal amplitude Vo.
1
 BITS-Pilani K.K. Birla Goa Campus, Physics Department
PHY F344: Advanced Physics Laboratory

φ = 0o Average VOut =0 φ = 90o

VOut VOut
Average VOut is positive

VR VR

VS VS

Average
VOutV=0
Out is zero
φ = 270o
φ = 180o Average
Average VOut is negative
VOut VOut

VR
VR

VS VS

The above figure shows how the output signal VOut varies with the phase difference
between the signal (VS) and the reference (VR). Please note that although VR is
sinusoidal, internally, it uses a square wave of the same frequency. In the case
where the phase difference φ between VS and VR is zero, the output signal VOut is all
positive. This is because, in the first half, VR = +1. So the signal wave in the first half
is positive. In the second half, both VR (= -1) and VS are negative, and hence the
product is positive [This signal appears as a full wave rectifier output]. In the case
where the phase difference φ is 90o, the output signal VOut is part of VS between -90o
to +90o. The part between 90o to 270o comes out with a change of sign (as VR = -1)
and hence looks identical to the first part. When averaged out, it becomes zero. For
the 180o case, the output is just negative of the 0 o case and for the 270o case, the
output is just negative of the 90o case. These signals will be verified during the
experiment (Part 4) by displaying VS and VOut on the oscilloscope for different phase
differences.

Theory of electronic Filters:

A combination of a resistor with inductor / capacitor or both, forms a frequency filter.

1) Low pass filter: Consider an R-C circuit shown in the

adjacent figure. Here we see that for low frequencies, the
capacitor will offer a high reactance (Xc = 1/jωC). So these
frequencies will produce an output signal. However, the high
frequencies will be shot-circuited (or bypassed) by the
2
 BITS-Pilani K.K. Birla Goa Campus, Physics Department
PHY F344: Advanced Physics Laboratory
capacitor which offers a very low impedance to these frequencies. Thus, this R-C
circuit will act as a low-pass filter. The frequency response for such a low-pass filter
is shown in the adjacent figure.

2) High pass filter: Now if one simply interchanges R and C in the low pass filter, it
will become a high pass filter, as now lower frequencies will be having high
reactance offered by the capacitor and
hence blocked, while the high
frequencies will be having low
impedance offered by the capacitor
and hence pass through.

It may be noted that in the low pass

filter, if one replaces the capacitor by
an inductor, since the reactance of an inductor is jωL, it will bypass lower frequencies
and thus act as a high pass filter. Similarly, if one replaces the capacitor by an
inductor in a high pass filter, it will block higher frequencies and thus act as a low
pass filter. By adding multiple sections, one can make the fall (or rise) of gain with
frequency for a low pass filter (or high pass filter) faster and faster (i.e. steeper).
Similarly, by having two capacitors in parallel and with an inductor in series (in the
shape of “π”), one can have a “low pass π filter”. Similarly, by connecting two
capacitors in series and an inductor in between the two in parallel (in the shape of
“T”), one can make a “high pass T filter”.

3) Band pass filter: If one connects inductor, capacitor and resistor in series and
take the output across the resistor, it will act as a band-pass filter. It will pass a band
of frequencies located at the resonance frequency (ωo = 1/√[LC] ), as at this
frequency, the circuit will offer minimum reactance and all the voltage drop will be
across the resistor. At frequencies away from resonance, the voltage drop across the
resistor will be lesser and lesser and hence signal throughput will be smaller and
smaller. The band-width (Δω) of a filter is the difference between the two cut-off
frequencies on either side of the central frequency (ωo). The “Q” of a band pass filter
is defined as Q= ωo/Δω. A high Q filter is called a narrow band filter and a low Q filter
is called a broad band filter. For a series resonant circuit, Q = (√[L/C])/R. In the
present experiment, we shall find the frequency characteristics of two band-pass
filters, one a broad band filter (Q=5) and one narrow band filter (Q=50).

In the band pass category we have three special filters having special values of Q.

Q=1 is called a “Chebyshev filter”. Q=0.707 (i.e. 1/√2) is called a “Butterworth filter”
and Q=0.577 (i.e. 1/√3) is called a “Bessel filter”.

3
 BITS-Pilani K.K. Birla Goa Campus, Physics Department
PHY F344: Advanced Physics Laboratory
Typical frequency responses of various
band-pass filters are shown in the
adjacent figure. In the experiment, we
shall determine the frequency response
of the three special filters and two band-
pass filters of Q=5 and 50, in the
frequency range 100 Hz to 2 kHz.

There is an important and fundamental

relationship valid for almost all linear
systems. This has application in filters,
amplifiers, detectors etc. This is called
“Kramer–Kronig relation”. It is named
after Ralf Kronig (also known for his Kronig-Penney model and Coster-Kronig
transition) and Hans Kramer, although this
relationship is better known in mathematic
as “Hilbert Transform”. This is a very
general relation applicable to many
complex mathematical and physical
systems. As for the filters or amplifiers,
what it means is that : whenever there is
change of response (or gain) with
frequency, there is usually an
accompanying change in phase. Leaving
aside a few exceptions, this is mostly true.
When the gain or response does not
change in frequency, the phase does not
change. In the case of filters, there is a
sharp change in filter response at certain
frequencies and hence an accompanying
phase change with frequency.

The adjacent figure shows how the phase

change occurs at the peak of the band-
pass filter frequency. It may be noted that
at the resonance frequency of the band
pass filter, there is zero phase difference.
Similar phase changes occur for low pass
and high pass filters in the cut off region,
where there is rapid change of the
response of the filter.

In the experiment, we will also have a look at the phase change with frequency for
the two band-pass filters of Q = 5 and 50 in the 100 Hz to 1 kHz frequency range.
4
 BITS-Pilani K.K. Birla Goa Campus, Physics Department
PHY F344: Advanced Physics Laboratory

Theory of Noise: For this, students may please refer to the hand-out of the
experiment on “Determination of Boltzmann constant from noise measurement”

Experimental setup:

The Signal Processor Lock-In Amplifier (SPLIA1-A, supplied by M/s Teachspin, Inc,
USA) is made up of 7 modules, which are briefly described here.
1) Reference Oscillator: It provides the “coding” or modulation signal for the
experimental apparatus. This oscillator is also the source of the reference signal
for the lock-in detector. It can also provide a test signal that can be used to study
a variety of signal processing configurations.
2) Function Generator: It provides a noise signal, a calibrated attenuator, and a
test signal (combining the reference oscillator signal with the noise source) for
various experiments.
3) Phase Shifter: It provides a shift in the relative phase of its input to its output of
0°- 360° for ac signals in the 3 Hz to 3 kHz frequency range.
4) Pre-amplifier: It gives voltage amplification to the experimental signal to a level
where the noise voltage of any subsequent unit will not significantly affect the
signal-to-noise ratio of the final output.
5) Filter: It selects the frequencies that will pass on to the detector (or to other
modules). It can also provide additional voltage gain.
6) Lock-in / Amplitude Detector: It rectifies the ac signal, synchronously, as a
lock-in detector, or asynchronously, as an amplitude detector (sometimes called
a precision rectifier).
7) Low-pass Filter and Output: It averages the rectified signals coming from the
detectors. This can be done with various time constants (averaging time) and
5
 BITS-Pilani K.K. Birla Goa Campus, Physics Department
PHY F344: Advanced Physics Laboratory
with two different roll-off characteristics (6dB/octet and 12dB/octet). As this
module has a dc amplifier as well as a dc offset control, it is possible to study
small changes in large signals. This can be accomplished by subtracting out the
large dc signal, and amplifying the difference signal around zero to search for
small changes. The unit also has a meter drive circuit along with an analog meter
to monitor and measure the dc output of the signal processor.

The above described components of the lock-in amplifier are shown below:

5) Filter

2) Function 4) Pre-amplifier Output

Generator 6) Lock-in Amplifier 7) Low-pass
filter

1) Reference Oscillator
3) Phase Shifter

Experimental procedure:

Part 1: Gain Band-width characterization of the pre-amplifier

1) Make connections as shown in the circuit block diagram shown below.

2) Connect the output of the external signal (function) generator to the input of the
Noise/Attenuator. Set the attenuator setting to 10 -1 at the input. Set the noise
level to “off”.

3) Connect the Noise/atten output to “+” input of the preamplifier, which is set to DC.
Set the “–“ input of the pre-amplifier to ground (GND).

4) Connect the output of the pre-amplifier to the Channel 1 of the oscilloscope.

5) Connect the signal (which is given to the Noise/Atten input) to the Channel 2 of
the oscilloscope.

6) Set the gain to G=1.

6
 BITS-Pilani K.K. Birla Goa Campus, Physics Department
PHY F344: Advanced Physics Laboratory
7) Vary the frequencies from 10 Hz to 1MHz (by one order of magnitude every time)
using the 5 MHz signal generator.

8) For each frequency, measure the input and output signals on the oscilloscope. To
ensure that there is no saturation, keep the input signal at say 500 mV.
(Heavily saturated signals will have flat top and bottom. This will give wrong
results. In that case, lower the input signal till I/O signals are sinusoidal).

9) Tabulate your results as follows: G=

Frequency (f) Input voltage (p-p) on Output voltage (p-p) on

Log10f Gain Gobs=Vout/Vin
Hz Channel 2 : Vin (mV) Channel 1 : Vout (mV)
10 1
100 2
1000 3
… ..
1000,000 6

10) Set the gain G=10 and repeat the procedure (Steps 8-9) to get the gain-
frequency curve for this gain also, with input signal of 100 mV.

11) Plot Gain (Gobs) vs. log f for G=1.

12) Plot Gain (Gobs) vs. log f for G=10.

Part 2: Frequency response of the three special filters

1) Connect the output of the Reference oscillator to the input of the filter and
Channel 2 of the oscilloscope.
2) Connect the output of the filter unit to the Channel 1 of the oscilloscope.
3) Set the filter band pass frequency to 500 Hz.
4) Set the input peak to peak voltage to 2V (on Channel 2)
5) Set the filter to “Bessel Filter”.

7
 BITS-Pilani K.K. Birla Goa Campus, Physics Department
PHY F344: Advanced Physics Laboratory
6) Scan the frequency from 100 Hz to 2 kHz by steps of 100 Hz up to 800 Hz and by
steps of 200 Hz thereafter.
7) Tabulate your results as follows: Filter = Bessel (Q=0.577)
Frequency (f) Input voltage (p-p) on Output voltage (p-p) on Gain
Log10 f
Hz Channel 2 : Vin (V) Channel 1 : Vout (V) Gobs=Vout/Vin
100
200
300
…
2000

8) Repeat Steps 6 and 7 for the Butterworth filter (Q=0.707) and Chebyshev filter
(Q=1) in the frequency range 100 Hz to 2000 Hz.
9) Plot Gain (Gobs) vs. Log10 f , for all the three filters on the same graph.

Part 3: Frequency response and Kramer-Kronig relation for Q= 5 and 50

1) Keep the same set-up as in Part 2

2) Set the filter to Q=5 and set Vin to 500 mV

3) Scan the frequency from 100 Hz to 1 kHz by steps of 100 Hz.

4) For each frequency, note down the time lag and corresponding phase shift
between the input and output signals (on Channel 2 and Channel 1).

Note that if Δt is the time lag between starting of the sine curves (zero cross over
with positive slope) on the two channels of the oscilloscope, then the phase lag is
given by φ (o) = 360 f (Hz) Δt (s). Note that φ < 360 o.

5) Tabulate your results as follows: Q= 5

Input voltage Output voltage Time Phase as Calculated

Frequency Gain
on Channel 2 on Channel 1 : lag on CRO Phase shift
(f) Hz Gobs=Vout/Vin o
: Vin (V) Vout (V) Δt(s) (φo) φC = 360 f Δt ( )
100
200
300
…
1000

[Theoretically, the phase should be between +90o to -90o. So the sign is

important. If it displays say 270o, it may be recorded as -90o. For the calculated
phase, if the calculated shift is more or less than this, 360o (one cycle) may have
to be added or subtracted to get the correct phase].
8
 BITS-Pilani K.K. Birla Goa Campus, Physics Department
PHY F344: Advanced Physics Laboratory
6) Set the filter to Q=50 and set Vin to 50 mV
7) Repeat Steps 3 to 5, for Q = 50 in the 100Hz to 1000 Hz range.
8) Plot Gain (Gobs) vs. Frequency (linear) for
both the filters on the same graph.
9) Plot Phase Shift (φo) vs. Frequency (linear)
for both the filters on the same graph.
10) Plot Phase Shift (φC) vs. Frequency (linear)
for both the filters on the same graph.
[The phase plots should be like the one shown in the above figure].

Part 4: Lock-in detection

1) Connect one output of the Reference Oscillator to the Phase shifter.

2) Connect the output of the phase shifter to one input of the Detector.
3) Connect the output of the Detector to the Channel 1 of the oscilloscope.
4) Connect the other output of the Ref. Oscillator to the second input of the Detector
and also to Channel 2 of the oscillator.
φ = 0o
5) If the output signal (Channel 1) looks clamped
VOut
(flat) at the top and bottom, lower the input
signal till the output becomes sinusoidal. VR
6) Keep the phase shift knobs to zero position.
7) Set the Reference Oscillator frequency to about VS

160 Hz, such that the signal and reference are

exactly matching (i.e. zero delay) with each other and you get an output
waveform like a full wave rectified sine wave type), like the one shown in the

9
 BITS-Pilani K.K. Birla Goa Campus, Physics Department
PHY F344: Advanced Physics Laboratory
figure above. [Note that although input is sinusoidal, internally there is a square
wave generated as VR].
8) Make sure that there are no sharp vertical cuts in the waveform. Here the phase
difference between the two is zero. Take a snap of the oscilloscope screen.
[Phase shifter markings are relative, not absolute]
9) Now advance the signal phase by 90o. In the case, the output signal is part of VS
between -90o to +90o. The part between 90o to 270o comes out with a change of
sign (as VR = -1) and hence looks identical to the first part. Note that this signal,
when averaged, will become zero. Take a snap of the oscilloscope screen.
10) Advance the signal phase further by 90o to 180o phase shift. The output signal will
now be upside down of the signal for zero phase lag. Take a snap of the
oscilloscope screen.
11) Advance the signal phase further by 90o to 270o phase shift. The output signal will
now be upside down of the signal for 90o phase lag. Note that this signal also,
when averaged, will become zero. Take a snap of the oscilloscope screen.
12) Advance the phase further by 90o to 360o. The output signal will look exactly like
the one for 0o phase lag.
Note: This exercise demonstrates that the integrated output of the lock-in amplifier is
highest when the phase difference between reference and signal is zero and the two
frequencies are matched. Noise at all other frequencies gets suppressed.

Results:
1) Plot of Gain (Gobs) vs. log f for G=1. {Part 1}
2) Plot of Gain (Gobs) vs. log f for G=10. {Part 1}
3) Plot of Gain (Gobs) vs. log f for all the three special filters {Part 2}
4) Plot of Gain (Gobs) vs. Frequency for Q=5 and Q=50 {Part 3}
5) Plot of Phase Shift (φo) vs. Frequency for Q=5 and Q=50 {Part 3}
6) Plot of Phase Shift (φc) vs. Frequency for Q=5 and Q=50 {Part 3}
7) Comments on all the above plots.
8) The four screen-shots for φ = 0o, 90o, 180o and 270o. {Part 4}
9) Comments on the above four screen-shots.

10
 BITS-Pilani K.K. Birla Goa Campus, Physics Department
PHY F344: Advanced Physics Laboratory

General Information: (Source : World Wide Web)

1) Inventor of Lock-in amplifier: The inventor of lock-in amplifier is commonly

believed to be the physicist Robert Dicke. In fact, he founded the Princeton
Applied Research company to market this product. However, Dicke himself
mentioned that even though he is credited with the invention of the device, he
had read about it in a review of scientific equipment written by Michels and
Curtis. This review in turn cites a 1934 article by C. R. Cosens !

2) Band stop filter : Like band pass filters, one also has “band stop” filters. For
example, L and C in series with a resistor (across which o/p is taken) will form a
band pass filter which will have high throughput at the resonance frequency
1/√(LC). Now, instead of having L and C in series, if we have them in parallel, the
circuit will “block” or “stop” the frequencies around the resonance frequency.
Such a filter is called “band stop” filter.

3) Chebyshev filter : This filter is named after the mathematician Pafnuty

Chebyshev, as it is based on his Chebyshev polynomials. There are two types of
Chebyshev filters. In Type 1 filter, there are ripples in the pass band just before
fall and in Type 2 filter, there are ripples in the stop band, just after the fall. This
filter has a steeper roll off than other filters.

4) Butterworth filter : This filter is named after British engineer cum physicist
Stephen Butterworth who proposed it in 1930, based on his Butterworth
polynomials. It is known for flattest response in the pass band among all filters.

5) Bessel filter : This filter is named after the German mathematician Friedrich
Bessel as it is based on his Bessel polynomials. In 1949, W.E. Thomson
designed this filter using the Bessel polynomials. It has the speciality that it has a
maximally flat group/phase delay (maximally linear phase response), which
preserves the wave shape of filtered signals in the pass band.

6) Bode Plot : It is an asymptotic approximation of the frequency response, using

straight line segments. It was originally conceived by the Bell Labs
telecommunication engineer Hendrik Bode in the 1930s. Bode Magnitude plot is
a plot of the magnitude (of say Gain, usually in dB) vs. log frequency, whereas
Bode Phase plot is a plot of the phase shift vs. log frequency.

7) Optical Filters: It must be kept in mind that like electronic filters, we have optical
filters also. In optics, long pass and short pass refere to the optical wavelength,
unlike frequency in electronics. One also has band pass and band stop (notch or
“V”) filters. Neutral density filters have a constant attenuation across the visible
wavelengths range and are used to prevent detectors from going into saturation.

11 END