Chapter 17: The Fourier Series

Welcome to Last Minute Lecture.

This free chapter overview is designed to help students review and understand key concepts.

These summaries supplement not replaced the original textbook and may not be redistributed or resold.

For complete coverage, always consult the official text.

In 1794, there was this French mathematician sitting in a prison cell, literally waiting to be sent to the guillotine.

Right, Jean -Baptiste Joseph Fourier.

Exactly.

But he somehow survived, eventually went to Egypt with Napoleon Bonaparte and then published this math theory about heat flow.

A theory which completely by accident laid the exact groundwork for how your cell phone processes radio signals today.

It's wild.

It really is.

So welcome to this deep dive, brought to you by the Last Minute Lecture Team.

Today, we are basically serving as your ultimate one -on -one tutoring session for chapter 7 time of fundamentals of electric circuits.

Yeah, so if you are staring down a circuit analysis exam or, you know, a major project, pull up a chair.

By the end of this session, the Fourier series is going to make complete sense.

I feel like this is arguably the most crucial chapter in the text because it bridges this massive gap.

I mean, up until this point, circuit analysis has been very clean.

Oh, absolutely.

You use your phasor toolkit to find steady state responses, basically turning messy differential equations into simple algebra.

The catch, though, is that phasors only work if your input source is a pure, perfectly smooth sinusoidal wave.

Which is great for an idealized textbook problem.

But, you know, if you look at a digital pulse train or a power supply rapidly switching on and off, that clean, smooth wave is just gone.

Completely gone.

You're looking at square waves, sawtooth waves, or just really jagged periodic signals.

And the second you introduce a square wave into a circuit, that powerful phasor toolkit you spent weeks learning just completely breaks down.

It becomes mathematically useless.

Well, useless unless you apply the audaciously brilliant claim made by that French mathematician, Fourier.

The guy dodging the guillotine.

That's the one.

He proposed that absolutely any practical periodic function, no matter how jagged or asymmetrical it is, can be unblended.

Unblended.

Like, separated out.

Exactly.

It can be represented simply as a massive neat stack of basic smooth sines and cosines.

Okay, so Fourier basically acts as a translator.

If our phasor tools only speak the language of smooth sinusoids, Fourier takes the jagged square wave and translates it into this giant list of sinusoids.

That is exactly right.

And then we can just use our existing phasor tools on each of those sinusoids individually.

Oh, nice.

So we don't have to learn a completely new method of circuit analysis.

Nope.

You just use the superposition theorem to add all those individual sinusoidal responses back together at the very end.

That is such a relief.

So that is the overarching strategy of chapter 17.

And the manual for that translation is equation 17 .3, which the text calls the trigonometric Fourier series.

Right.

And it states that your complicated periodic waveform, which we call f of t, is composed of three distinct ingredients.

Three ingredients.

Okay, let's unpack those.

The first ingredient is a subzero, right?

Yes.

The text defines a subzero as the DC component.

Think about a pure standard sine wave.

Okay.

It spends exactly half its time above the zero voltage line and half below it.

Right.

So if you average that out over one full cycle, the average value is just zero.

But a weird jagged wave might spend a lot more time hovering at positive five volts before briefly dipping negative.

Exactly.

And the mathematical consequence of that unevenness is a net positive or net negative offset.

That average value, calculated over one full period, is captured entirely by a subzero.

Got it.

It's basically the baseline that the rest of the wave dances around.

Beautifully put.

So then we get to the dancing parts.

The second and third ingredients are two infinite sums.

Right.

One is an infinite stack of cosine waves, each with an amplitude called a sub n.

And the other is an infinite stack of sine waves, each with an amplitude called d sub n.

So our primary job for these problems is figuring out those three Fourier coefficients, a sub zero, a sub n, and b sub n.

That is the core of it.

But to find those, we first have to ground ourselves in a couple of variables.

First is the fundamental frequency written as omega sub zero.

Which is two pi divided by the period of your original jagged wave, right?

Exactly.

If your square wave repeats every single second, its fundamental frequency is just the baseline rhythm of that wave.

But the formula also introduces this variable n, and n represents the harmonic frequency, correct?

Yes.

And cosine must be an integer.

One, two, three, and so on, up to infinity.

Oh, okay.

So every sine and cosine in Fourier's stack has a frequency that is a perfect integer multiple of the fundamental frequency.

You have a wave oscillating at one omega zero, another at two omega zero, a third at three omega zero.

And those are your first, second, and third harmonics.

Precisely.

I gotta say, I always get a bit suspicious when math texts claim you can do this to literally any function.

I mean, there has to be a physical limit, right?

You can't just scribble a chaotic infinite line on a page and apply a Fourier series to it.

You are absolutely right.

There are limits.

And they are defined by the Dirichlet conditions.

Dirichlet conditions, okay.

These are basically the rules a function must follow for the infinite series to actually converge, meaning the stacked sinusoids actually add up to the uridial waveform.

Let's figure out why these rules exist.

The first condition is that the function must be

I assume that's just because a physical voltage can't, you know, simultaneously be five volts and negative five volts at the exact same millisecond.

The physical universe simply doesn't allow it.

So the math doesn't either.

Makes sense.

What about the others?

The second and third conditions state that the waveform can only have a finite number of discontinuities, like that instant jump in a square wave, and a finite number of peaks and valleys within one period.

Well, yeah, if a wave oscillates up and down infinitely fast within one single second, it would contain infinite energy.

Exactly.

And we can't use a sum of finite sine waves to build something with infinite energy.

The math would just shatter.

Right.

And that underlying logic probably applies to the final Dirichlet condition too.

The one about the integral of the absolute value of the function over one period having to be finite.

It does.

Yeah.

But honestly, in practical terms, any physically realizable signal you can generate in an engineering lab will naturally meet all these conditions.

So you don't usually have to panic about them breaking down on your workbench.

No, not usually.

Speaking of the workbench, when I look at the textbook graphs showing all these sinusoids stacking up to build a square wave, something physical jumps out at me.

Well, a square wave has a perfectly sharp 90 degree corner, but we are trying to build that corner out of curved, sweeping sine waves.

To me, that sounds like trying to build a perfectly straight wall using round legos.

I like that analogy.

You can use smaller and smaller legos to get close, but the very edge is always going to have a slight curved bulge.

That is actually a brilliant way to visualize a famous mathematical quirk known as the Gibbs phenomenon.

Gibbs phenomenon.

Yeah.

When you sum up all those Fourier terms, they do a spectacular job of approximating the flat top of the square wave.

But right at the points of discontinuity, where the square wave tries to instantly jump from zero to one, the sum of the continuous waves will actually overshoot the flat top.

Because the mathematical round legos we're using physically cannot make an instant 90 degree turn.

Exactly.

They overshoot and then they ripple a bit before settling down into the flat section.

Yes.

And here is a fascinating detail for anyone taking an exam on this.

Definitely take notes here.

No matter how many infinite terms you add, that overshoot never disappears.

Adding a million harmonics just makes the ripple narrower, squeezing it closer to the edge.

But the height of the overshoot is permanently fixed.

Really?

Fixed at what?

At about 9 % of the peak value.

Wow.

A permanent 9 % bulge at the corners.

That is a great visual to lock away.

Now, looking at the math again, keeping track of separate sine and cosine waves for every single harmonic seems really clunky.

It is.

But the text mentions we can combine our sub n and b sub n coefficients so we don't have to carry both around.

Okay.

How do we merge them?

We merge them into what's called the amplitude phase form.

Instead of viewing the cosine and sine terms as separate entities, picture them as the two legs of a right triangle.

Oh, like basic geometry.

Exactly.

If a sub n is the base and b sub n is the height,

standard trigonometry allows us to find the hypotenuse.

That hypotenuse becomes our single combined amplitude called capital A sub n.

And the angle of that hypotenuse gives us the phase shift phi sub n.

So we take two clunky terms and merge them into one single elegant cosine wave that just has a specific height and a specific starting angle.

That's so much cleaner.

It really is.

And transforming the data into this format allows us to create an incredibly useful visual tool,

the amplitude spectrum.

Right.

Because instead of graphing voltage against time like on a normal oscilloscope, you graph the amplitude a sub n on the vertical axis and the harmonic frequencies on the horizontal axis.

And that creates what the text calls a line spectrum.

It looks like a series of vertical spikes planted along the frequency axis.

And usually the spike at the fundamental frequency is the tallest, right?

Specifically, yes.

And as you move to the right into the higher harmonics, the spikes generally get shorter and shorter.

It visually proves that the bulk of a signal's energy lives in the lower frequencies.

It maps the DNA of the signal.

But, and here's the catch, to find those amplitude spikes, you have to calculate a sub zero, a sub n, and b sub n by doing complex integration over a full period.

Yes, you do.

Doing that manually for every harmonic in a homework problem sounds like an absolute nightmare.

I would spend an hour just doing calculus before I even touch the circuit analysis part.

It is tedious, no doubt.

Isn't there a shortcut?

Like a way to look at a waveform and know before doing any math that certain coefficients are just going to equal zero and can be ignored?

There is.

This is where recognizing symmetry becomes your most powerful shortcut.

The textbook highlights three main types of symmetry that will literally save you pages of calculus.

Oh, thank goodness.

Okay, the three types are even, odd, and half wave symmetry.

Let's logic our way through these.

First is even symmetry.

The rule is that the waveform is a perfect mirror image across the vertical y -axis.

Right, so if you look at a standard cosine wave, the left side of the y -axis is a perfect reflection of the right side.

By definition, f of t equals f of negative t.

So because the jagged wave acts like a perfect mirror image, it shares the fundamental mathematical property of a cosine wave.

Exactly.

So if it behaves like a cosine, I'm guessing the Fourier series will only need cosine terms to build it.

You nailed it.

That translates to all the sine coefficients.

Every single b sub n term equaling perfectly zero.

You don't even have to write out the integral.

Wow, that's amazing.

It gets better.

To find your sub zero and sub n terms, you don't need to integrate over the entire period from negative to positive.

You just integrate the positive half from zero to t over two and then multiply the result by two.

Okay, that cuts the calculus workload in half instantly.

So what if the wave has odd symmetry?

Well, if even symmetry is a mirror reflection, odd symmetry means a wave is flipped horizontally and then flipped vertically.

Like a standard sine wave.

Mathematically, yes.

f of negative t equals negative f of t.

It is anti -symmetrical.

Following the previous logic, if the jagged wave mimics the anti -symmetry of a sine wave, its Fourier series should only contain sine terms.

The math confirms that.

If you spot odd symmetry, your dc component, a sub zero, is zero.

Your cosine coefficients, a sub n, are all zero.

Beautiful.

You only have to calculate the b sub n terms.

And again, you could just integrate half the period and multiply by two.

Knowing those two rules alone will save you so much time on an exam.

The third one is half wave symmetry.

Yes.

This is when the first half of the cycle looks like a mountain and the second half of the cycle looks like the exact same mountain, just perfectly inverted into a valley.

Right, the bottom half is the negative mirror of the top half.

The textbook defines this as f of t minus t over two equaling negative f of t.

Okay, so what's the shortcut here?

The shortcut here is uniquely powerful.

If you identify half wave symmetry,

the Fourier series will exclusively contain odd harmonics.

Wait, odd harmonics?

Meaning we only calculate for n equals one, three, five, seven?

Yes.

The even harmonics where n is two, four, six completely vanish.

Both a sub n and b sub n will be zero for every even number.

That's a huge time saver.

And in addition, the dc component is always zero.

You only integrate to find the coefficients for the odd numbered spikes on your spectrum.

So the lesson here is check for symmetry before you even touch your calculator.

Okay, so we've unblended the wave.

We've used symmetry to find the Fourier coefficients without pulling our hair out.

But this is a circuits class, not a pure math class.

Right, we need to apply it.

Yeah, if I have a circuit diagram with a resistor, an inductor, and a voltage source spitting out a weird square wave, how do I actually find the output voltage?

The text outlines a systematic four -step procedure that brings all these tools together.

Step one is the math we just completed.

Express the non -sinusoidal excitation as a Fourier series, breaking it into its dc component and its ac harmonics.

Exactly.

Step two requires shifting our mindset.

We had to transform the circuit from the time domain into the frequency domain, just like we do in standard phasor analysis.

So a resistor stays as resistance r, but an inductor becomes the impedance j omega l, and a capacitor becomes one over j omega c.

Yes, but here is the critical detail where students often make a fatal error.

Oh, pay attention to this part.

The frequency omega is no longer a single fixed number.

Because we have an infinite stack of harmonics, the impedance of your inductor and capacitor will be different for every single harmonic.

So you can't just plug in a single number for omega.

No, you must leave the frequency as n times omega zero in your equations.

Because an inductor resists high frequencies much more than low frequencies, the hundredth harmonic is going to face a massive impedance bottleneck compared to the fundamental frequency.

Exactly the case.

So moving to step three, you calculate the circuit's response to each component individually.

Starting with the dc component.

Right, and remember, at zero frequency, an inductor acts as a dead short circuit, and a capacitor acts as an open circuit.

Right, right.

You find the dc output voltage or current, then you calculate the phasor response for the ac harmonics, keeping your equations generalized in terms of n.

So I'm basically creating a master algebraic formula that spits out the response for any harmonic number I want to plug into it.

Building on that, step four is where you execute the superposition principle.

You take your dc response, and you take all your individual ac responses,

transform them back from phasors into time domain cosines, and simply stack them up into a final infinite sum.

It is incredibly satisfying.

You take a massive problem that seems impossible, shatter it into tiny, bite -sized sinusoidal pieces, run each piece through a basic phasor equation, and then just glue the pieces back together to get the final answer.

It's a very elegant process.

You can actually see the math filtering the wave.

Like, if I calculate the output for a low -pass filter circuit, the amplitudes of my higher -end terms will mathematically shrink down to nearly zero.

The math literally shapes the physical wave.

It creates a beautiful harmony between the abstract algebra and the physical reality of the circuit.

But there is one area where this additive process creates a massive headache.

Let me guess.

Calculating average power.

You got it.

I was just thinking about that.

Power is voltage multiplied by current.

If my voltage is an infinite series of sinusoids and my current is an infinite series of sinusoids.

To multiply them together, I'd have to use the distributive property.

Every single harmonic of voltage would have to cross -multiply with every single harmonic of current.

That's millions of terms.

It sounds impossible.

If you had to calculate every cross -term, it would be impossible without a supercomputer.

But we are saved by a fundamental property of harmonically -related sine waves.

Which is?

Orthogonality.

Orthogonality.

Okay, I need a visual for how that saves us.

Imagine multiplying a slow fundamental sine wave by a much faster harmonic sine wave.

Because their frequencies are exact integer multiples, the fast wave will force the resulting product to spend exactly as much time above the zero voltage line as below it.

Oh, during the slow wave cycle.

Exactly.

So when you calculate the average power by taking the integral, which just means calculating the total area under that curve,

the positive area perfectly balances out the negative area.

The total area cancels out to exactly zero.

All those millions of cross -multiplied terms involving different frequencies simply vanish.

That is mathematical magic.

The only time the integral survives is when you multiply a voltage wave and a current wave that share the exact same harmonic frequency.

So the third harmonic of voltage only creates real power when it interacts with the third harmonic of current.

Everything else mathematically evaporates.

This realization forms the basis of Parseval's theorem.

Parseval's theorem, got it.

Parseval formalized that the total average power delivered to a load by a periodic signal is simply the sum of the power delivered by the DC component, plus the power delivered by the first harmonic, plus the second, and so on.

No cross -multiplication required at all.

It is just a simple straight line addition problem.

P total equals Pdc plus P1 plus P2 plus P3.

Exactly.

And if the exam asks for the total RMS value,

the effective root mean square value of the weird jagged wave,

Parseval's theorem says it is just the square root of the sum of the squares of the RMS values of the individual components.

We square them because power is proportional to voltage squared.

Exactly.

Parseval's theorem reduces an infinite chaotic multiplication problem into an elegant string of addition.

Elegance is definitely what the textbook strives for in Xt, I noticed.

Because tracking all these sine and cosine terms, even with the symmetry shortcuts, takes up a lot of space on the page.

It does get messy.

The text introduces a much more compact way to write this called the exponential or complex Fourier series.

Right.

The trigonometric series relies on sines and cosines.

The exponential series relies on Euler's identity.

Ah, Euler.

He's everywhere.

He really is.

Euler proved that trigonometric functions are deeply connected to complex exponentials.

A complex exponential represents a vector spinning in a circle on the complex plane.

So instead of managing a sub -n coefficient for the horizontal cosine and a separate b sub -n coefficient for the vertical sine, we just define a single complex coefficient called c sub -n.

Exactly.

And that c sub -n tracks both the amplitude and the phase at the exact same time.

It condenses the entire infinite sum into a single sleek mathematical expression.

That is so much better.

And when you calculate this c sub -n coefficient, particularly when you're analyzing rectangular pulse trains,

a very distinctive mathematical pattern consistently emerges.

The text refers to it as the sinc function, spelled S -I -N -C.

Sinc function.

I look at the graph of this, and it is fascinating.

Mathematically, it's defined as the sine of x divided by x, right?

That's the one.

If you evaluate it at zero, the math limits out to exactly one.

But as you move along the axis to integer multiples of pi, the sine component becomes zero, forcing the entire function to hit zero.

And if you plot this on an amplitude spectrum,

the sinc function creates a beautifully decaying bouncing envelope.

It looks really cool on paper.

The vertical spikes of your harmonic frequencies fit perfectly underneath this bouncing curve, naturally decaying as the frequency increases.

It basically provides a visual boundary for how the energy of a digital pulse bleeds out into the higher frequencies.

Which is incredibly important for modern communication, because you need to know how much frequency bandwidth your digital pulse is actually going to consume.

Absolutely critical.

But let's bring all this theory back down to the workbench.

If you're designing a circuit, you aren't going to do these integrals by hand, right?

You're going to use software.

And the text specifically highlights how to use PSPICE.

Yes.

But if you are setting up a PSPICE simulation,

you cannot just drop a square wave into the schematic and press a Fourier button.

The software doesn't intuitively know what the steady state looks like.

Oh, really?

What do you have to do first?

You must first run a transient analysis.

Which means telling the computer to simulate the circuit turning on and running for a certain amount of time with a very fine time step.

Exactly.

So it can capture all those rapid high -frequency harmonic fluctuations.

Once the transient simulation captures that time domain data, PSPICE offers two specific analysis tools.

OK, what's the first one?

The first is the Discrete Fourier Transform, or DFT.

This tool processes the data and outputs a highly precise text file.

By default, it prints out a table listing the exact DC component, followed by the amplitudes and normalized phases of the first nine harmonics.

That is perfect if you need hard numbers for a lab report.

But what if you want to actually see the spectrum?

Then you use the second option, the Fast Fourier Transform, or FFT.

The FFT is a highly optimized algorithm that computes the discrete transform incredibly quickly.

And PSPICE uses it to generate a continuous graphical plot, right?

Exactly.

You get to visually observe the line spectrum, those vertical frequency spikes, directly on your monitor.

It's basically the software equivalent of a spectrum analyzer.

The text actually makes a great point of contrasting these tools.

If you hook a normal oscilloscope up to your circuit, you are looking through a window into the time domain.

Right.

You see voltage plotted against time, so you literally see the focal jagged square wave.

But if you hook up a spectrum analyzer, you are looking through a window into the frequency domain.

It plots amplitude against frequency.

So looking at a spectrum analyzer is literally looking at Fourier's math manifested in the real world.

You see the massive spike of energy at the fundamental frequency, and then the smaller spikes at the third, fifth, and seventh harmonics, just trailing off to the right.

And observing that spectrum is what makes the physical design of filters intuitive.

Imagine looking at that spectrum analyzer, and then wiring a low -pass filter into your circuit.

Okay, I'm visualizing it.

You are physically designing a gate that allows the DC and low frequency spikes to pass through untouched while actively crushing the high frequency spikes down to zero.

Oh, and because you chopped off the high frequency energy that provides those sharp 90 -degree corners, the wave that comes out of the filter won't look like a square wave anymore.

It will look like a soft, rounded, rolling hill.

Precisely.

Now taking that further, imagine designing a highly selective bandpass filter.

You tune the component so it acts like a narrow keyhole, perfectly aligned with just a fundamental frequency spike.

And then you feed a chaotic, jagged square wave into the input of this filter.

Yes.

What happens?

Well, it rejects all the DC offset.

It rejects the third harmonic, the fifth harmonic.

And what comes out of the output is just a pure, perfectly smooth, single sine wave.

You literally use physical hardware to extract a single mathematical concept from the chaos.

That is the true power of circuit analysis.

It allows us to manipulate the invisible environment around us.

And, you know, if we zoom out to a broader perspective, consider the journey of this knowledge.

How so?

When Jean -Baptiste Joseph Fourier was developing this math in the 1820s, his only goal was to figure out how heat slowly diffused through a heavy, solid metal plate.

He had absolutely no concept of electricity, let alone digital logic or radio frequencies.

Yet the mathematical framework he built to describe a hot iron pleat is exactly what allows a modern smartphone to function.

It really is.

The air around us right now is a chaotic, jagged mess of thousands of intersecting electromagnetic radio waves.

And your phone's antenna acts as a bandpass filter.

It relies on Fourier's principles to slice through that messy time domain reality, isolate the specific frequency of your data stream, and reject all the rest of the noise.

It is a profound reminder that fundamental principles transcend their original applications.

When you master these concepts, you are not just memorizing equations for a test.

You are learning the foundational language that governs modern technology.

I think that is the perfect thought to close on.

To our listener, you now have the master key.

You understand the physical reasoning behind the Dirichlet conditions.

You know how to use even, odd, and half -wave symmetry to instantly eliminate massive chunks of calculus.

You have a reliable four -step method for translating and solving any circuit.

And you understand how Parseval's theorem uses orthogonal areas to save you from infinite cross -multiplication.

You are entirely ready to tackle this material.

On behalf of the Last Minute Lecture team, thank you for joining us on this deep dive into Chapter 17.

Good luck with your exam, trust the foundational math, and we will catch you on the next deep dive.