Podcast on LTI Systems: Frequency Domain Analysis

Kapitoly

Úvod

Kúzlo sínusových vĺn

Skutočné vs. komplexné signály

Frekvenčná odozva

Generalized Frequency Response

A Practical Example

Time vs. Frequency Domain

Key Properties of Frequency Response

Visualizing the Response

Sinusoids and LTI Systems

The Unchanged Frequency

Amplitude and Phase Shift

Final Recap

Přepis

Oliver: Dobre, o tomto som vôbec nevedel – a myslím, že toto si musí vypočuť každý. Systém LTI môže dokonalej sínusoide urobiť *len* dve veci. Iba dve! A to je všetko.

Hannah: Presne tak. Je to jeden z najzákladnejších princípov, ale je neuveriteľne silný.

Oliver: Počúvate Studyfi Podcast. Sme tu, aby sme si prešli systémy s lineárnou časovou invarianciou, teda LTI, a prečo sú sínusoidy ich najlepším priateľom. Hannah, od čoho sa odrazíme?

Hannah: Začnime tým, prečo sú sínusoidy také výnimočné. Sú to prirodzené, základné signály. Predstavte si ich ako čisté hudobné tóny. Každý z nich je definovaný len tromi parametrami.

Oliver: Dobre, tromi parametrami. Čo sú zač?

Hannah: Po prvé, amplitúda, čo je v podstate hlasitosť alebo sila signálu. Po druhé, frekvencia, teda ako rýchlo signál kmitá. A po tretie, počiatočná fáza, ktorá nám hovorí, kde presne vo svojom cykle signál začína.

Oliver: Amplitúda, frekvencia, fáza. Rozumiem. Znie to dosť priamočiaro.

Hannah: A teraz prichádza tá zaujímavá časť. Hoci sínusoidy, s ktorými sa stretávame v reálnom svete, sú skutočné signály, pre analýzu je oveľa praktickejšie použiť ich komplexnú verziu. Nazýva sa to analytický signál.

Oliver: Počkať, komplexnú? Akože s *imaginárnou* časťou? Môj stredoškolský učiteľ matematiky má práve nočnú moru.

Hannah: Zostaň so mnou! Znie to zložito, ale v skutočnosti to veci zjednodušuje. Komplexný signál v sebe spája informácie o reálnej aj imaginárnej časti, ktoré sú znázornené v karteziánskych súradniciach. Alebo, a to je ešte lepšie, môžeme si ho predstaviť ako rotujúci fázor v komplexnej rovine.

Oliver: Rotujúci fázor... Takže ako šípka, ktorá sa točí dokola s konštantnou dĺžkou?

Hannah: Presne tak! A rýchlosť jej otáčania je frekvencia. Tento spôsob uvažovania je pri analýze systémov LTI nesmierne užitočný.

Oliver: Dobre, vráťme sa teda k tomu, čo si hovorila na začiatku. Že systém LTI môže sínusoide urobiť len dve veci. Čo to znamená v kontexte tohto rotujúceho fázora?

Hannah: Znamená to toto: keď pošlete sínusoidu – náš rotujúci fázor – do systému LTI, na výstupe dostanete opäť sínusoidu s *tou istou* frekvenciou. Fázor sa bude stále otáčať rovnakou rýchlosťou.

Oliver: Takže sa frekvencia nikdy nezmení? Nikdy?

Hannah: Nikdy. Systém LTI nemôže vytvoriť nové frekvencie. Môže urobiť len dve veci: môže zmeniť amplitúdu signálu – dĺžku fázora – a môže zmeniť jeho fázu – počiatočný uhol fázora.

Oliver: A táto predvídateľná zmena je to, čo nazývame frekvenčná odozva?

Hannah: Máš to! Frekvenčná odozva je komplexné číslo, ktoré nám presne povie, ako veľmi systém zmení amplitúdu a fázu vstupnej sínusoidy pri danej frekvencii. Je to ako charakteristický odtlačok systému.

Oliver: To je fantastické. Takže ak poznáte frekvenčnú odozvu, môžete predpovedať, čo systém urobí s akýmkoľvek sínusovým signálom. To znie ako superschopnosť pri spracovaní signálov.

Oliver: So that covers our perfect, normalized sinusoids. But what if the input signal is a bit... extra? Like, what if its amplitude isn't just one?

Hannah: Great question! And the answer is actually really simple, thanks to linearity. If our LTI system is linear, it treats all amplitudes fairly.

Oliver: What do you mean by 'fairly'?

Hannah: Think of it this way. The system's response at a certain frequency, that K of omega, is just a multiplier. If you send in a signal with an amplitude of 'A', the output is just the original response multiplied by 'A'.

Oliver: So if the input is twice as big, the output is twice as big? That's it?

Hannah: Exactly! The system doesn't change its fundamental behavior. It just scales the output to match the input. Simple as that.

Oliver: Okay, let's walk through an example. What if we have a simple system with a bit of feedback?

Hannah: Perfect. Let's imagine a system where the output is a mix of the current input and a slightly delayed version of the previous output. A common setup.

Oliver: And we feed it a complex sinusoid at a specific frequency, say 2 pi over 40.

Hannah: Right. First, we find the system's transfer function, which is basically its mathematical identity. Then we plug in our frequency to get the frequency response, K.

Oliver: And that K is a complex number, right?

Hannah: It is! And that's the magic key. Its magnitude tells us how much the system amplifies or reduces the signal's amplitude. In this case, it's about 0.82, so it gets a little quieter.

Oliver: And the angle of that complex number?

Hannah: That's the phase shift. It tells us how much the output signal is delayed, or shifted in time, compared to the input. Here, it’s about negative 31 degrees.

Oliver: So we can see on the plots how the output is smaller and shifted. But what if we calculated this in the time domain, using a function like lfilter in Python?

Hannah: Ah, you'd see something interesting. For the first little bit, the output would look different. There's what we call a 'transient region'.

Oliver: A transient region? Like the system is... warming up?

Hannah: That’s a perfect way to put it! The frequency response calculation assumes the signal has been playing forever, from negative infinity. The time-domain calculation starts at zero.

Oliver: So it needs a moment to settle into that steady rhythm we see in the frequency response. It has to 'forget' that there was silence before.

Hannah: Precisely. After that initial warm-up, it settles into what we call the AC steady state, and the two results match perfectly.

Oliver: Got it. Now, are there any universal rules or properties for this frequency response function?

Hannah: There are! Two big ones are periodicity and symmetry. The first is easy: the function is periodic with a period of 2 pi.

Oliver: Meaning it just repeats itself over and over again, like going around a circle?

Hannah: Exactly. The response at a frequency omega is identical to the response at omega plus 2 pi, or 4 pi, and so on. We only need to look at one slice of it.

Oliver: And symmetry?

Hannah: This one's cool. For any real system, the response at a negative frequency is just the complex conjugate of the response at the positive frequency. It's a perfect mirror image.

Oliver: Which I assume saves us a lot of calculation time.

Hannah: You bet it does! We only need to compute half of it.

Oliver: So how do we actually see all of this? We plot it, right?

Hannah: We do. We typically create two plots. One for the magnitude, showing the gain, and one for the phase, showing the time shift, both as a function of frequency.

Oliver: And tools like MATLAB or Python have functions like freqz to do this for us?

Hannah: They sure do. You just feed it the system's numerator and denominator coefficients, and it spits out the response.

Oliver: But that gives us the response versus the relative angular frequency, omega. How do we make that useful in the real world?

Hannah: That's the final step. We just scale the frequency axis by our sampling rate, Fs. This converts the abstract 'radians per sample' into concrete Hertz, which is what we care about for things like audio signals.

Oliver: Okay, that makes so much sense. So for our final topic, let's talk about something really cool: what happens when you feed a sine wave into an LTI system?

Hannah: This is my favorite part! The answer is surprisingly elegant. If your input is a sinusoid, your output will also be a sinusoid.

Oliver: The *exact* same sinusoid?

Hannah: Almost! It will have the exact same frequency. That never changes. But its amplitude and phase… those might be different. Think of the system like a funhouse mirror for waves.

Oliver: It stretches and shifts it, but it doesn't change the wave's core identity?

Hannah: Exactly! The system's transfer function, which we write as G of e to the j omega, tells us how.

Oliver: And how does that work?

Hannah: The magnitude of that function, K, scales the amplitude. And its angle, phi, tells you the phase shift. So the output is just the original wave, but scaled and shifted.

Oliver: So we can predict the output perfectly just by knowing the system's transfer function.

Hannah: That’s the magic of it. We use Euler's formula to break the input cosine into complex exponentials, see how the system affects them, and then put it all back together.

Oliver: Amazing. So to recap everything… LTI systems are predictable, and their response to complex signals can be understood by looking at simple inputs like sinusoids.

Hannah: You've got it. It’s a foundational concept in so much of engineering and physics.

Oliver: Hannah, this has been incredible. Thanks for breaking it all down. And to everyone listening to the Studyfi Podcast, thanks for joining us. Keep studying!