The intuition behind the Nyquist-Shannon Sampling Theorem
Вставка
- Опубліковано 28 лют 2024
- To try everything Brilliant has to offer-free-for a full 30 days, visit brilliant.org/ZachStar/ . The first 200 of you will get 20% off Brilliant's annual premium subscription.
STEMerch Store: stemerch.com/
►Follow me
Odysee: odysee.com/@ZachStar:0
Instagram: / zachstar
Twitter: / imzachstar
Support the Channel: / zachstar
PayPal(one time donation): www.paypal.me/ZachStarYT
Join this channel to get access to perks:
/ @zachstar
2D Graphing Software: www.desmos.com/calculator
Animations: Arkam Khan (For contact info go to www.arqum333.com/)
Check out my Spanish channel here: / zach star en español
►My Setup:
Camera: amzn.to/2RivYu5
Mic: amzn.to/35bKiri
Tripod: amzn.to/2RgMTNL
►Check out my Amazon Store: www.amazon.com/shop/zachstar
The fact that this guy makes hilarious videos on his other channel and unironically useful videos on this one is impressive
What is the other channel?
@@thenextboundary834 Himself Zach Star
Title should be "Are you able and willing to figure out the original signal?"
That's why we solder the microchips onto the board - can't have them running away when we tell them to do math
I laughed out loud when he said “and this makes sense because of the Fourier transform” cause I thought he was going to dumb it down a bunch. Love how this channel is like a “more mature” math channel where not everything has to be explained at a middle school level. Thanks!
This upload could not have been timed better. I'm busy learning about this stuff in my signals & systems class and seeing the graphs and plots really helps
Nyquist-Shannon theorem is so cool! It lets one connect discrete and continuous signals through their information density, which provides very deep insight. You can also generalize it to signals which do not have compactly supported frequency spectrum like gaussians!
And there is a surprising connection to the study of minimal length in quantum mechanics!
this is one of those things that seems simple but is mindbendingly cool. like the 44.1 kHz thing, basically it's saying if we know the signal at these few isolated points, then we know what it is at all times, _unless_ it contains frequencies higher than half of 44.1 kHz, in which case humans can't hear them anyway
Everytime i click on one of these videos, i feel like ive unlocked something magical or divine
You do this better than profs at my „elite“ university. This makes me sooo mad at our education
Because he's focusing on one subject.
He doesn't need to do a full course.
He can spend a huge amount of time preparing 11 minute video.
Professors cannot do that.
Thank you so much. I am very grateful that I can understand this theory and why it is periodic in frequency domain. ❤❤❤
This will help me pass the final exam for my signal processing course tomorrow. Brilliant explanation!
I am in a class where we apply the Nyquist-Shannon theorem for signal analysis.
Thank you thank you...explained to an amateur with a rabid wish to know from first principles. Ive even bought an oscilloscope with FFT to see what a signal looks like without knowing what to look for
I'm literally learning about this in one of my classes and we have a midterm next week, so thank you for the good timing, Zach. 🙏
It’s nice to have a neat visual depiction of how this theorem works. Thanks!
I wrote a test on this just this afternoon. Great timing and would have loved to have had this before the semester! great video
I've been tackling digital signal processing on my own time and this video really helped solidify my understanding of the Nyquist-Shannon Theorem.
god damn, why couldnt you make this vid one semester sooner xd
You are doing amazing things for the field of EEE. Thank you brilliant!
You made this video exactly while I’m taking an ADC DAC course. Perfect timing!
Nicely explained. As always!
Thank you Zach for such a well presented, detailed and accurate introduction to a difficult concept.
Bro, I love your videos, this is the first time you've posted one while I was covering it in a class though. We didn't cover the transform part of it, so that really helped me understand WHY aliasing is introduced below double sampling rate
christ all mighty i am so happy you're posting on this channel again!
Excellent video. Greetings from Panama 🇵🇦
I've known about this for a while, but now I actually understand it! Thank you so much!
Excellent stuff - Sinc Functions, Fourier Transforms, and Aliasing all in 10 minutes. Wow!
I knew all this ...at one time in the past. Nice to see it again. You are the math teacher we never got.
This is such a wonderful visualization, step by step, and not as abstract as drawing on a whiteboard as most professors do haha. Thank you!
The bit on aliasing is a GREAT visualization =)
This comment is sponsered by brilliant. New course that gets you top comment each time
it works!
Signals & Systems my favourite course in EE
Didn’t like DSP?
I am currently studying this in my course.and just your video
What a satisfying refresher to Signals and Systems! These topics are really starting to fade away after my graduation
As an electrical engineer that should really be doing my signal processing homework rn, thanks for the video
Mathematics shedding light into logic, reasoning, assumptions, etc. Well done! 🙏
Omg Zach PLEASE make a convolution video ❤
Actually this repeating of frequency domain can help you to process higher-frequency signals using your regular PC sound card's ADC: sampling essentially acts as a frequency mixer in a heterodyne receiver with a lot more of "collateral" bands. Though I don't really know whether motherboards have a low pass filters on mic inputs or not
thanks ❤
I have my digital communications exam tomorrow and you posted this video at the right time lol
Love it
my favorite theorem of all time
This math is so dense my head Hertz.
You should make more videos like this
Hey @zach! This is awesome! May I ask which tools you use to build your graphs and animated visuals?
This gave me ptsd from my Control System course from last semester 💀
Goat
When my knowledge of music makes me familiar with much of the terminology in this video
Damn where were you when I had to learn this shit 7 years ago? Amazing video and really good explanation
Please, make a video about convolution! That would be super helpful!
This brings flashbacks to 3. semester in electrical engineering. Pretty easy stuff as soon as you understand it
Goatt
Finally after 69 years we get to see another nerdy video.
I want to know why an 8-bit bit sampling depth won't show me real distortion levels on a 16 or 24 bit signal, even if the sampling rate is much higher than the frequency of the signal.
Just in time
I don’t understand this, but I certainly hope too soon.
Great
I'm always confused when I come to this channel and get rational content.
Please make a video on convolution math I'm 2 months into signal processing and I still don't understand why I'm doing it.
Shannon the GOAT
how do you prevent sinusoidal dipleneration?
I am willing, but not able to figure out the original signal.
DSP is goated. I'm and undergrad and really interested in the subject and I'm wondering where I could end up working in DSP in the industry. Do you have any tips where a career in DSP could lead?
Hey Zach can you please make a video on Engineering Physics degree
How can you make these Videos? 😊
Shannon-Nyquist can actually be beaten with compressed sensing!
Hi, How can I contact with you?
I love your videos, but this is the very first time I understood almost 0% of this because I’ve never been exposed to this kind of content
I want to follow.
I love you
I've always been a little confused about whether 2f is enough, or if strictly greater than 2f is required.
At 3:50 you say "faster" and use a greater-than symbol, but at 9:58 you say "at least" while still showing a greater-than.
I get that in the real world the sampling frequency is never gonna be exact anyway so you need a decent margin (and you showed CD audio being 44100 not 44k as an example of that), but in theory, can I get away with 2f or do I need 2f+epsilon?
In theory sampling at 2f is enough, in practice before ADC we need analog low pass anti-aliasing filter to get rid of frequencies >f. If you don't do that, noises from bats, etc. will "alias" to lower frequencies that can be audible by humans, which is bad. Analog filters are not perfect, 20kHz low pass filter still passes higher frequencies but with lower amplitude. So, we have to sample much higher to combat aliasing from imperfection of analog filters.
Sampling at exactly 2f is sometimes enough, but in most cases isn't. It depends on the phase difference between the samples and the frequency at f.
If the samples just happen to fall on peaks of "f" - the signal will be recovered exactly. However, if the samples fall in any other place - you get the frequency at "f" with a reduced amplitude. And if they fall exactly on zeros - the frequency at "f" will be lost.
Of course, that's assuming we have a perfect low-pass filter to recover the signal.
Next step is to sample non uniformly
And this is why high resolution audio is a scam
😲😮
It is nice to spend all my free time learning this stuff, but I'm forgetting it faster than I'm learning it. I'm looking at my math worksheets from 10 years ago and I have to figure it all out again. I think I'll leave this to those autistic people that have a freaky ability to absorb it all. I'll never be as good as they are. I'm going back to playing video games.
nifty
Great video, however I feel you missed an important point, the shannon theorem is a sufficient but not nessesary condition for reconstruction is only true for sinusoidal interpolation. In different bases things get very different, this is what compressed sensing works with.
the top comment is not sponsored by brilliant
Mathematics shedding light into logic, reasoning, assumptions, etc. Well done! 🙏