Music Technology 101: Sampling Rate and Bit Depth Explained

[jabellae1]

I love your explanation... really you made me happy!

[manalexxxx]

ok i understand the denser the audio (8 bit---- 16 bit----- 24 bit) the better the sound quality. but the hertz? what is that again? so 16 bit -- 24 bit to 44.1 Khz -- 88.2 khz. how do these two work together? explain a little bit better pls.

[MrGolden94]

Thanks man, this helped me with my college essay alot! Love it :)

[domagoj19zg]

thnx man!!! 

[CallusDnb]

Excellent tutorial mate.

[jpelayo22]

thank you this video is being used in class as i Type this comment

[rapagameover]

Thanks!!! this is really usefull

[childofconvienience]

Really really good you explained it perfectly so I could understand it without any troubles though I am not even a native speaker yet.

Oh and btw=) At the end you say "eye" instead of "ear"=)

[childofconvienience]

Really really good you explained it perfectly so I could understand it without any troubles though I am not even a native speaker yet.

[mikevanstaveren]

Thanks for the video! I have a question I hope you can answer it:

I dont really understand quantization, if you sample a audio track, at 44.1 Khz or higher, you have dots, like you showed in your video, if you connect the dots, you get pretty close to the original soundwave you drew on paper.

Why then is there such a thing a quantization? Why is it needed?

[mikevanstaveren]

@mikevanstaveren (continued)

I see it as a sheet of math paper, this sheet of math paper is in the soundcard, here a school boy plays connect the dots all the time, the result is a perfect analog to digital copy, right?

Are you saying that some soundcards may lack proper math paper to play connect the dots? :( (

Im trying to make it as simple as possible because I find it hard to understand)

[MangoldProject]

@mikevanstaveren Your math paper analogy is good, with one caveat: think of the soundcard as the "grid" on the paper, not the paper itself. You DO play connect the dots, BUT not the original waveform's dots - first, when doing Analong->Digital, you're "warping" the input dots to the closest positions on the grid. When doing Digital->Analog, you're connecting the RECORDED (warped) dots. This "connect-the-dots" does NOT match the original waveform. It may be very close, but not exact.

[MangoldProject]

@MangoldProject As for why quantization is needed: it's a necessity. Ideally, we'd like to do away with it, but your computer has finite memory and processing capabilities, so it must store the voltage of the recorded waveform using a finite set of numbers. It's part of the price you pay for going digital (analog waveforms also have "effective" resolutions because of signal to noise, etc., but that's a different story.)

[mikevanstaveren]

@MangoldProject Thanks for your answer! So the more "Grid" your sound card has to "warp" dots the more accurate your sound is going to be! I just saw a nice analogy on you-tube, of a camera, the sample rate is the times the camera take's a picture, and the bit-depth, is how accurate the colours are i.r.l.

But what is kbps all about then? Is it how much info you stream from your device to the speakers?

And why is an mp3 128 kbps (and never xx bit and xx khz ) How are they connected?

[mikevanstaveren]

@mikevanstaveren More clearly maybe: How do you reach 320 Kbps, how can I tell if it was recorded at 16 or 24 bit? If I want to buy a song on spotify it says (high quality 320 kbps MP3) This tells me nothing :(

[Tripbag]

@mikevanstaveren, if you are ever buying an mp3 online, remember: mp3 is a lossy form of compression.

Generally, the recording doesn't have to be the same bit depth as the distributed recording; especially if it is online or on CD. Generally, online mp3 purchases are 16 bit and 44.1kHz, so that the average consumer may easily play the purchased audio using other CD and mp3 players if the wish to burn or play a copy off of a computer.

If it is 24 bit, it will likely say hi-fi, or at least 24bit

[MangoldProject]

@mikevanstaveren kbps means "kilobits per second". It doesn't tell you anything specific about the sampling rate or bit depth. For example, an uncompressed 8-bit, 16kHz recording is (2^8 bits) * 16000 (per second) = 4096 kbps. Same goes for a 9-bit, 8kHz recording (=2^9*8000=4096 kbps). Note these numbers are much higher than the average 320 kbps MP3s use. That's because MP3s use lossy compression to reduce the number of kbps (without significantly degrading the quality).

[nwsm22]

Very Helpful! I loved the Illustrations!

[nwsm22]

Very Helpful! Thanks!!!!!!!!!!!!!!

[Haukenslush]

very helpful. thanks!

[Sytrusze]

Wow this is the best explaination so far!! Really clear with al the pictures and comparison great job man thanks alot!!

ps. is it possible to (let's say) in logic pro convert the Sample rate of an audio file from 24bit to 4bit? Thanks

[pmc57]

Good presentation, beautifully paced delivery. Just one issue. I get the impression from your explanation that the more samples taken the better the waveform reconstruction. We need to sample only twice per cycle to regenerate a sinewave perfectly (Shannon: (sine x) / x function) and because the highest freq. of interest is 20Khz we are compelled to sample at 40kHz (ignoring aliasing). The abundant samples at lower frequencies adds nothing.

[bvprulz]

outstanding... the best beginner material i hav seen in youtube regarding sampling and quantisation!!! G8 demonstration..

[philinmotion]

great tutorial ! 

[acedaface954]

this was awesome.

[Pianopod]

Brilliant tutorial. You make a complex subject easy to understand. You are a brilliant teacher!

[Fusionicon]

I ment the sampling and quatising.

[Fusionicon]

What are the equations involved ?

[MangoldProject]

@Fusionicon: I'm not sure what equations you're referring to. If you're talking about a particular statement made during the video, let me know at what time so I can review it.

[axxe21]

Agreed- Extremely clear....Great tut...

[guitarbasslover]

So if I set everything really high as far as bit and sample rate I'm simply going to get a better sound?

[MangoldProject]

@guitarbasslover: generally, the answer is "yes". However, there are a few provisions.

1. Beyond certain thresholds our ears cannot tell the difference. There was a paper in the Journal of the Audio Engineering Society in which a 16-bit, 44.1 kHz AD-DA converter was inserted into the path of an audio track; several audio professionals were blind-tested and it turned out they couldn't hear a difference between the audio with and without the converter. (continued in next comment)

[MangoldProject]

@guitarbasslover (continued):

2. Additionally, there is a "law" which states that, to reproduce frequencies up to X, you need to sample at a rate 2X (the so-called Nyquist's criterion). Since we're interested in audio only up to 20kHz, sampling at 40kHz should be enough. For practical reasons having to do with converter design, this is often extended above 40kHz. However, if you're asking me whether a human being could tell the difference between, say, 192kHz and 96kHz, my answer would be: NO.

[wheelerj92]

thank you!

so much info for my music tech final exam tomorrow

[ilycsfm]

that helped so much thanks for the great visuals! it really makes things clear and easy to understand!

[hurricanerock246]

Cheers mate! That was exactly what I needed to know

[consoleskript]

very well explained, thanks

[seskandari2003]

WHY "Quantize" for? What does it do?

[MangoldProject]

Hey Seskandari,

I'm not sure I understood your question. Could you be more specific? If you're asking what we need quantization for, then quantization is a necessary step when digitizing audio (or any signal, really).

[DonJuanDeMarco2]

Asaf.. im not sure if your aware of just how a good a teacher you are !

Even though you oviously have alot of knoledge about the subjects in your videos, you are still able to 'humble' the information and completely relate to the mindset of a beginner in the most perfectly straightforward and simple manner.

Even for someone like myself who has some knoledge, this is an excellent way to re-cap on things.

I can only wish you much success in your career as gratitude for these videos.

[DonJuanDeMarco2]

The video explains it perfectly from 4:51

The quantization of a sample, in A/D conversion for example, simply rounds off the sample to a set degree.

So if you have a wavelength, you have high peaks and low peaks. With quantization, the peaks are rounded off to a set height/depth.

1 bit means all high peaks are rounded to the same level and all low peaks are rounded to the same low level. The higher the bit depth the less rounding off occurs, meaning the sample becomes more accurate in reprod

[kidspranqe]

its another way to do the sample rate

[ilycsfm]

@seskandari2003

If you are asking why it needs to quantize its because the computer cant recognize random points on a graph. they need to be on a specific line in order for the computer to recognize where it is, and by quantizing you move each dot to the closest line, in order for it to be recognizable by the computer. And as demonstrated to more lines, the less distorted the signal

plase correct me if I am wrong.

[seskandari2003]

WHY "Quantize" for? What does it do?

[mangasprai]

brilliant!

[smiuw]

That was great. Very good explanation.

Thank you.

[Lohmoth]

bravo

[hellmuth26]

very interesting, and very well explained.

[semantixthasorcera]

Peace Asaf

I've been recording music for almost 15 years now and I've never heard anyone explain Sampling Rate and Bit Depth so well as yourself!

Very useful videos for beginners who wish to record their music, as are your Piano lessons. You have taught some gr8 techniques that i find really useful in grasping a better understanding of piano playing as i have only been playing for about a year now & your lessons are invaluable!

Looking forward to the next video as always!

Peace and Blessings

[MangoldProject]

This is my first "non-player" video - I hope you enjoy it. Criticism is welcome.

By the way, despite encoding the video using MPEG-4 @ 640x480, Youtube still degrades the quality considerably. Any experts on the subject with advice on improving this are welcome to make suggestions!