Gregory Thompson

Theory:

Figure 1. Lock-in Amplifier Block Diagram.

Small signals are often hindered by noise. A lock-in amplifier can solve this problem.
Typically a commercial lock-in amplifier comes with two channels in order to calculate the
amplitude of the signal with small noise.

Figure 2. Product of two sine waves and trigonometric identity relating their frequency. In our
demonstration, the frequencies are the same, unlike what is displayed above.
Gregory Thompson

Figure 3. Derivation of the identity we use to relate the average value of the modulated signal to
our output.

We started off with a voltage divider that has a function generator as a voltage source.
Two resistors are placed in series and we measure the voltage across the smaller subsequent one.
Four our purposes, we will need a very large resistor in series with a following small one.

Figure 4. LTspice compartment of our voltage divider used to generate the small signal buried in
noise and reference signal.

The signal we obtain is saturated in noise. We can barely see the original wave so we
must amplify it. We used a non-inverting amplifier in order to increase the amplitude of our
original wave.

Figure 5. Pre-amplifier that amplifies signals buried in noise before amplitude modulation.

It is not displayed within the figure of the whole lock-in amplifier set-up, however the
LF411C chip labeled “Pre_amplifier” which has two junctions that contribute to balancing.
Gregory Thompson

Figure 6. Potentiometer we used in combination with the balance junctions of the pre-amplifier
to remove the offset.

Using a potentiometer and these junctions allows us to eliminate the offset. The rest of
the amplifier works as a traditional 5-junction amplifier.

Figure 7. Equations related to each of the channels and their phase shift. Our output for each
channel after the final amplifier is a DC numerical value that also equals what is in the cube box.

In order to determine the amplitude of our signal, we need two channels that have the
same two input parameters: a version of the reference wave and the pre-amplified signal with
noise. After pre-amplification, each channel has the almost identical component flow except one
is introduced to an all-pass filter that makes its reference signal out of phase with the original one
by 90°.
Gregory Thompson

Figure 8. The all pass filter that acted as a phase shifter due to the product of the resistances and
capacitances are roughly equal to 1/2πf.

Figure 9. Wikipedia quotes validating our assumption that an all-pass filter can become a phase
shifter.

Figure 10. LTspice simulation of voltage waves. Red and purple are the relevant waves. As one
can see, they are out of phase with each other by approximately 90°.
Gregory Thompson

The signal from the function generator and small amplitude signal with noise have the
same frequency. Therefore, multiplying their waves allows us to exploit trigonometric identities
to determine the amplitude of the wave of the signal with noise because in practice, we wouldn’t
know it.

Figure 11. LTspice simulation of the variable voltage source. B1, that is the product of the
reference wave and amplified signal buried in noise for the first channel to cause amplitude
modulation. This was done since LTspice doesn’t have the simulation component of a AD633
analog multiplier.

Figure 12. Components of the AD633 chip used in hardware as an analog multiplier for
amplitude modulation.

To do this amplitude modulation in hardware mode, we use an analog multiplier. We

place our components in the appropriate junctions as we would do with any ordinary type of
chip. If one were to take the fast fourier transform of the modulated signal created by two waves
with the same frequency, one would obtain two discernible peaks at a the difference and sum of
frequencies of the two of signals. For each of the channels, we then place the output of our
multiplier through a low-pass filter that eliminates the noise and sum frequency in our fast
fourier transform plot.
Gregory Thompson

Figure 13. Low-pass filter attenuation of noise and wave that has a frequency that is within the
amplitude modulated signal. The wave with a frequency that is equal to zero (the difference) is
kept and trigonometric identities are used to solve for the amplitude of our sine wave.

We determine the numerical value for τ by deciding what frequencies to attenuate. The
output needs to be amplified after measuring the voltage across the capacitor. Signals with small
frequencies allow the capacitor to charge, therefore, after attenuation, our signal corresponding to
the frequency of 0 viewed at the FFT will be present. Our final DC output should be the product
of the amplitude of reference signal and signal we are attempting to measure multiplied by either
sin(Δφ) or cos(Δφ), depending on which is shifted.

Figure 14. Algebraic relation used to determine the amplitude of our signal buried in noise.

Modifications to the original project/Evidence that the circuit works:

The only modification that we implemented in hardware was using the balance junctions
on the operational amplifier to eliminate the offset that was seen after the pre-amplification stage.
We didn’t originally intend to create two channels. Therefore, becoming familiar with the all pass
filter was not a priority until we decided that a phase shift would be necessary. Compared to
other projects, this one entailed a lot of the same components, therefore, there wasn’t much of a
need to modify the original design. What was surprising was that the analog multiplier had
Gregory Thompson

decreased the overall gain of the system by a factor of 10. So when initially looking at our
output, it took a moment to realize why it was off.

If we continued the Tier II project until proper functioning in hardware mode, general
intuition tells us that there would have been many modifications. However, we simulated nearly
the same circuit diagram in LTspice to obtain accurate results. Therefore, it is likely there is a
caveat to our analog mixer or all-pass filter that we were not taking into consideration. Or simply
rebuilding it would work fine.

A comprehensive check of each component was done but not recorded. However,
evidence that the full circuit was well on its way toward proper functioning and completion is
available in the Figure below.

Figure 15. Hardware output. Green and yellow DC voltage showed wrong amplitudes one we
calculate the amplitude of our signal and compare it to the true value.

From the image of our hardware, a more neat configuration could have been done.
Excessive long wires increase the likelihood of human error contributing to non-functionality.
Therefore, for the next electronics project, all wires’ length will be decided beforehand.
Gregory Thompson

We knew that the circuit was not functioning properly because the DC green and yellow
lines (Y and X, respectively) implied amplitudes that were off from the true value. After we
would obtain those values of Y and X from our oscilloscope, we would use the following
relation to solve for the amplitude of our signal, which we could compare with the theoretical
amount obtained by using the voltage divider equation.

2 2 1
R= 𝑋 +𝑌 = 2𝑔
· |𝐴𝑠𝑖𝑔𝑛𝑎𝑙| · |𝐴𝑟𝑒𝑓𝑒𝑟𝑒𝑛𝑐𝑒|
|𝐴𝑠𝑖𝑔𝑛𝑎𝑙| is the amplitude of our original signal buried in noise, |𝐴𝑟𝑒𝑓𝑒𝑟𝑒𝑛𝑐𝑒| is the amplitude of
our reference signal and g is our total gain over all amplifiers in the system. We initially thought
our device was working properly due to coincidence because our first calculation with our first
chosen resistor resulted in an measured/calculated |𝐴𝑠𝑖𝑔𝑛𝑎𝑙| that was very close to the true value,
however, when we replaced the large resistor with other values, solving for the amplitude of our
signal didn’t even result in the same order of magnitude as the true value.

We decided to simulate the entire thing in LTspice instead of continuing to troubleshoot

in hardware mode since that was an option for the level of the project. The LF411 amplifier that
was imported was working just fine. However, the third party analog multiplier we imported to
LTspice was giving us a lot of trouble. Therefore, we decided to do it manually by creating a
variable voltage that has an input of two voltage amplitudes and an output of their product which
does the amplitude modulation for us.
Gregory Thompson

Figure 17. LTspice simulation of the Lock-in Amplifier with both channels.

Figure 18. LTspice simulation output.