A Digital Media Primer For Geeks by Christopher "Monty" Montgomery and Xiph.org

Подписаться 475

Просмотров 59 тыс.

50% 1

This first video from Xiph.Org presents the technical foundations of modern digital media via a half-hour firehose of information. One community member called it "a Uni lecture I never got but really wanted."
The program offers a brief history of digital media, a quick summary of the sampling theorem, and myriad details of low level audio and video characterization and formatting. It's intended for budding geeks looking to get into video coding, as well as the technically curious who want to know more about the media they wrangle for work or play.
Original source: xiph.org/video/vid1.shtml
Read in text form here: wiki.xiph.org/Videos/A_Digita...

Наука

Опубликовано:

3 окт 2015

Ссылка:

Скачать:

Готовим ссылку...

Добавить в:

Мой плейлист

Посмотреть позже

Комментарии : 114

@js9415 6 лет назад

We need a new video from Monty. We miss you Monty.

@geot4647 7 лет назад

He does a good job of shutting up the vinyl-superiority crowd who don't understand why digital waveforms aren't actual "steps" and why there is no perfect analog in practice. Obviously a lot of vinyl folk are annoyed by evidence and want to cling to nostalgia.

@sadpepe7937 6 лет назад

BUT MUH SMOOTH, WARM, INFINITELY DEEP ANALOG SIGNAL IS SUPERIOR TO THE QUANTIZED, COLD, UNNATURAL DIGITAL BITS AND BYTES

@SirJeff 4 года назад

Sad but true. Personally I do own a turntable and a record collection but I stand by digital audio formats for the most consistent results. Why did I get into records? Because you get a physical copy of the music, you can see the big art, you feel like you own something instead of paying for streaming services etc. There's the more interaction, and that's what I find fun and amazing. If all I wanted was high fidelity, I'd be going full on after CD's. Perhaps some of the people in the vinyl community loves the sound of the crackles and pops you get from a record due to dust and particles in the groove. I'm fine with them too and sometimes it does feel nice, but "warm analog sound which has infinite samples is better than cold digital sampled sound"? Bullshit. Ain't gonna buy that.

@phazonlord0098 4 года назад

Honestly, Vinyl is better than CD or digital nowadays just because how horribly compressed, squashed the dynamic Range is on digital versions of most songs and since you can't compress the audio as much on a vinyl, it sounds better even with the crappy noise floor, pops and cracles.

@phazonlord0098 4 года назад

Mainly the only reason why I found myself hunting down vinyl rips.

@pilotavery 4 года назад

I mean, I would argue that vinyl is better because it's nostalgic and awesome just like a 1960s mustang is better even though my mom's Honda Accord handles better, is more comfortable, gets better gas mileage, has more horsepower, has better traction, and is safer. That 1960s mustang is fucking awesome though so I mean

@ANigerianPrince 8 лет назад

Who is this wizard? I must acquire all his knowledge!

@Ustaleone 4 года назад

have you found more videos of this guy? he explains everything so well, I must tap this knowledge!

@Corloi 3 года назад

@@Ustaleone ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-cIQ9IXSUzuM.html this one is great

@SteeveBjornson 8 лет назад

YES!!! This is the most accessible primer you'll find on the topics. Instead of trying to go absolute basic, he tells you at the beginning that some concepts you might not understand the first time around. And, because he's actually interesting to listen to (clear, calm, and descriptive), and the topics are presented in an interesting way, you probably will even want to watch it again to learn more.

@kylemattimore 3 года назад

One of the best half hours I've ever spent

@incep 2 года назад

Wow this guy is just so knowledgeable, I greatly appreciate it. Wish it was a more complete channel with just his videos.

@laughing5559 Год назад

Worked at the telco for a while. Sampling was done at 8kHz 7 bit. The 8th bit was used as hook on/off indicator. 8kHz * 8 =64 kilo bits per second. This also an ISDN B channel. 24 ISDN B channels combined to make a T-1 Channel. If you ever tried to make a phone call and got "all circuits busy during an event was causing everyone to try to call from a subdivision, it was because 24 B channels were already in use.

@hangugeohaksaeng 2 года назад

This is such an awesome straitforward introduction to a topic I've never been able to get a strait answer about. Thank you very much!

@kayokk- 2 года назад

It's always fascinating, and there's always something new to learn. Thanks, Monty.

@phillieg58 3 года назад

Old analog and digital audiophiles need to watch this. Digital audiophiles need to understand oversampling improves frequency response using less expensive anti-aliasing filter.

@monkeyxx 6 лет назад

This is one of my favorite GD videos evr, just scratching the surface but it smell so good!

@d4t4b4s3f4c3 7 месяцев назад

I have an overewhelming urge to invite this guy over for Christmas dinner at my mom's house

@MrFryfish 8 лет назад

I like your videos a lot. You explained very well (and remind us..) not to distance to much from the basic knowledge ... then approach the rest of it (new "science"..) cautiously... all other (new) technologies, as far/much as everyone of us need of these ...

@theemu6353 2 года назад

Thank you so much for your videos! Greets from Germany!

@maesitos 8 лет назад

Wow this is soooo interesting

@RXP91 6 лет назад

I LOVE this guy. Jesus christ - what a hero

@tylerbostwick1 8 лет назад

Amazing!

@qt9967 3 года назад

Repeat after me, "I am a kiln. I am not an oversized microwave"

@sl96131396 8 лет назад

13:20 i love this guy

@Audio_Simon 5 лет назад

Walt.. Nyquest was researching telegraph when he has his theorem? Did mechanical devices suffer from aliesing problems??

@Audio_Simon 5 лет назад

What is the jazz piano at the start, please? Monty Alexander would be fitting I suppose.

@GeorgeOu 3 месяца назад

NTSC DVD is 720x480, not 704x480 shown at 18:50. DTV used internationally outside of the US is 704x480 pixels.

@Zamicol 6 лет назад

Chris!! You are amazing! Why don't you have your own channel?

@Audio_Simon 5 лет назад

Sorry so many questions. Signed and (linear) unsigned encoding; does the sign use up one of the stated bits? Say 16bit signed is the first bit the sign?

@mina86 2 года назад

Sort of but not exactly. Integers typically use 2’s complement representation so if the most significant bit is set that the number is indeed negative. However, if you naively strip that bit you don’t get number’s absolute value. 14:14 demonstrates this. 0x8000 corresponds to -2¹⁵. This is different in floating point numbers where the most significant bit does represent sign directly. If you zero that bit you get absolute value of the number. And flipping that bit is equivalent to multiplying by -1. (At least so long as we ignore NaNs.)

@medwaystudios 7 лет назад

Great video. At 14:50 isn't the exponent 8 bits?

@mina86 2 года назад

It is. Float has 23-bit mentisa (which yields 24-bit precision mentioned in the video), 8-bit exponent and one bit for sign.

@CharlesHess 3 года назад

Maybe address the non-digital laser discs from 40 years ago. I’m only slightly familiar with them. I’d love to hear your take.

@Zorlin0 8 лет назад

nice nexus one!

@jasonlisonbee 7 лет назад

Why encode the entire waveform rather than the frequency that creates it and time deltas for start and end?

@iopqu 7 лет назад

Because the frequency and the amplitude changes. When you hit a guitar string, the original note might be D+0.04 cents, but as it keeps ringing it will go down to a D if it's tuned. This is because when the string vibrates further, it stretches more, and the higher tension will produce a higher pitched sound. As the string starts to vibrate less, it gets quieter and lower pitched. There is no real way to tell if a certain sound in real life has a certain formula that is easier to compress than the samples it produces.

@hooman_talakian 4 года назад

you are a wild GEEK!

@dannyd1224 6 лет назад

ahhh i always wondered why my camera looked like that on the tv set as a kid lol

@chopsueysensei Год назад

A geeky and totally unimportant correction.. IEEE-754 actually uses 23 bits for the magnitude or mantissa (not 24), and 8 bits for the exponent (not 7)

@meaninthemirror 4 года назад

13:25 that sneaky smile..

@j7ndominica051 6 лет назад

If a steep anti-aliasing filter is difficult to make, which is a reasonable claim, why is it that old CDs from the 80s usually have a steep "digital" cutoff at around 20.5 kHz with no transition band? How was that realized?

@iopqu 6 лет назад

You can cut it off later if needed digitally, but you need to do an ANALOG filter to limit the frequency before you convert to digital.

@mass1s 6 лет назад

When recording audio, you wouldn't sample directly at 44100 Hz (CD sample rate) with your ADC. Instead, you'd sample at a much higher frequency (like 96 kHz). This relaxes the requirements of the analog anti-aliasing filter: instead of passing everything under 20 kHz and attenuating everything above 22.05 kHz by 100dB, your filter only has to attenuate everything above e.g. 48 kHz. Then, you'd filter out everything above 22.05 kHz in the digital domain (fairly easy, especially compared to an analog filter), and downsample to 44100 Hz.

@dlarge6502 3 года назад

Well back then, with the first CD players they did just that but got it wrong. Eventually later models had a filter that worked well, but wasn't very efficient and was expensive. Years later CD players started oversampling. They read the data off the disc and then resample it to a higher sample rate. That's simple to do. That means they have all the audio data, plus a huge amount of empty samples beyond it thanks to the resampling. Thus they can get away with a much gentler and cheaper filter. Now when you see a CD player in a thrift store that proudly says 2x or 4x oversampling you know what it means. 2x means it resamples to 2x 44,100 and 4x is 4x 44,100. However, 2x and 4x oversampling are much the same, just different filters can be used. Yet you can expect and see that CD player manufacturers were using oversampling as a sale point. You benefit from some form of oversampling but you will see how players competed as to how much even if it didn't make a difference.

@pedrocarvamorim 3 года назад

Damn!

@rotteneffekt4416 Год назад

It's sounds good enough for a normal person. So they don't see or even think about lossless vs lossy. It's just inertia at this point, a format established itself and that's what the masses go with. Flac's cool though.

@vladjovanovic7207 5 месяцев назад

Very nice presentation. However... You state that analog signal gets distorted measurably, irretrievably during transmission and duplication. Agreed! I wonder if you would comment on something that I would like to pose as a question. Since analogue signal is continuous and has true value at any time (let us say in voltage terms), how do you propose to make a perfect copy of that signal by any means of digital sampling? Here is why I am asking this question... Since all digital technology has numerical limitations and cannot solve something simple as a square root of 2 back and forth several times without losing precision, how do you make a/d and d/a converters perfect? A/D and D/A converters are pure mathematics, nothing more... and as such are subject to rounding errors and necessarily have to interpolate and lose data in real-time applications. As such, these mathematical routines are heavily using MMX, SSE and other CPU extensions that do not have uniform mathematical precision capabilities and by default have to approximate data, especially in real time...Therefore, reproduction and sampling of digital signal MUST depend, to a certain degree, on the abilities of the either DAC chip or CPU used to reproduce the signal back into analogue domain. Do you agree? Secondly, analogue is a continuous value at any discrete time chosen at any point and you say that much yourself. Please describe the sampling technique that is prefect in converting complex analogue signal, with second, third, fourth and so on, harmonic signal. Sound is a complex, dynamic and non-linear system, as far as mathematics is concerned. Yet you claim that sampling, which is linear and time discrete, gives a perfect copy of the input signal. I beg to disagree. I could go on with sampling, quantization, jitter, sampler voltage comparators, be it dual slope or single stage... All this components have voltage slew rate and final operational speed. And with all of that, digital works in a final range of 20Hz to 20KHz. With filters. Well... just being pedantic. And.. on the top of everything, with all obvious A/D and D/A errors, we have sound file formats with compression on the top!! Please do not try to justify that! I am perfectly aware of my limitation to hear just about 12 KHz... but that is not the point. First we make errors in sampling, A/D, then we assume some values in D/A and then we overlay errors trough MP3, SACD, MPEG and the rest... and then we claim that digital is somehow - PERFECT. I am at fail to understand your position... Digital sound is a color reproduction of Mona Lisa... but, as of now, it still just a reproduction. Keep up good work, and please do not oversimplify things that are not so simple.

@jonah1976 15 дней назад

You'll NEVER capture, store, and reproduce any analog signal perfectly. It's like trying to write down an infinite number. At some point, you just have to put down the pen and accept what you've got. But with digital sampling, it's easy to surpass the accuracy of the human senses. Monty has another video that specifically deals with audio sampling. I highly recommended it.

@johnthecreative 2 года назад

doubling your sample rate cuts your latency in half - at the expense of double CPU load. if your CPU can handle it there's one benefit

@peterselie1779 2 года назад

There's no such thing as latency as a result of sampling. Did you actually watch the video? Do you actually understand the implications of the Nyquist-Shannon sampling theorem?

@johnthecreative 2 года назад

@@peterselie1779 you are wrong. increasing sampling rate decreases latency in a DAW. if you don't know this fact then whatever. BTW this has nothing to do with what in this video. it's just a fact like 2+2 = 4. there's a whole world out there so go explore.

@LucaskrillHC 2 года назад

Also you could just half your buffer length

@johnthecreative 2 года назад

@@LucaskrillHC thanks for the tip. Problem is it's usually not my choice. Lately I've had to increase buffer size for better performance. Maybe my next computer can handle smaller buffer size better.

@artysanmobile Год назад

What is it about a knowledgeable scientist enlightening a RU-vid audience on his chosen topic that chases the know-nothings out from under their rocks of ignorance? Every time Monty posts, out they come, utterly ignoring everything he’s just said and quoting terms they’ve heard but never understood, completely out of context? I guess it’s never occurred to them how silly they look in doing this. Speaking directly to you who do this, it’s ok to learn and remain silent. Those whose theories we are discussing here made their discoveries 75 years ago following tireless research and experimentation. Even more important, their discoveries have been verified and re-proven thousands of times through the decades, even as vastly more accurate tools became available to do so. To expound on how they failed is simply ridiculous.

@montefullmer1120 6 лет назад

Yet, a digital scope has latency issues where it can't refresh as quick as a good ol CRT Tetronix O Scope.

@ASJC27 5 лет назад

Bullshit. They can refresh much faster than CRT. I recommend you watch some of eevblog's videos on digital vs analog scopes. If your digital scope has latency issues it's because it is a shit scope, not because it's digital.

@robertsyrett1992 6 лет назад

24:57 Ho ho ho ho

@acche-rc 8 лет назад

Wish there were more AVI vs MKV vs MP4 vs WebM

@iopqu 8 лет назад

+acche2 Those are just containers, they can have the same video inside of them

@acche-rc 8 лет назад

+iopqu ya. just wonder how those containers store the streams differently

@SlinkiestTortoise23 7 лет назад

iopqu Hi there, what do you think of RME digital converters?

@moeboe6293 6 лет назад

Just what on earth is he drinking at 29:45?

@cebruthius 6 лет назад

Ice coffee? (a bucket of sugar really)

@sadpepe7937 6 лет назад

Wizard juice

@elijahjflowers 4 года назад

*in passive aggressive* lmao

@spirridd 3 года назад

I have no idea what it is, how and why I found it.

@johnthecreative 2 года назад

i get it. he's one of those guys that don't understand the difference between converting the sample rate of one audio file and doing the same to 100 combined audio files being recorded, mixed, and mastered. Yes dude you are right you cant hear a difference between 44 khz and 48 khz in solo. But there is a difference when you do this to 16+ tracks as they are recorded, mixed, and mastered. In blind tests some prefer the 44, but they still hear/feel a difference.

@RaveyDavey 2 года назад

You don't get it. You really have zero idea of what he does or does not understand. Maybe google who he is LOL...and then tell me he doesn't understand audio science We already know that inside of DAWs, mixing and applying effects, especially non-linear effects like compression, high sample rates make sense. But once mastered, none of the reasons are longer relevant.

@johnthecreative 2 года назад

@@RaveyDavey read what I said again. Music production is bigger than mastering - are you aware of that? that's the final step and at that point then you are right much of this is irrelevant as it's too late but that's not what i was talking about. I was talking about the 100 steps before that stage.

@montefullmer1120 6 лет назад

I thought a digital stream was of simply zeroes and ones. Thus, the evidence of stairsteps in any sine wave.

@ASJC27 5 лет назад

Yes it is a data stream, no there is no stair step. His other video explains it nicely: ru-vid.com/video/%D0%B2%D0%B8%D0%B4%D0%B5%D0%BE-cIQ9IXSUzuM.html

@dlarge6502 3 года назад

The stair steps only exist on paper to help view the waveform. They don't actually exist. Well, in some DAC chips they do, briefly but they go away once the DAC has finished conversion. Thing is, zeros and ones as you put it can not represent stair steps. They are too small and only represent a tiny point on the waveform. A stairstep, if it were real, would span an amount of time, thus needing many digital samples to represent it. In th analogue domain the stair step represents a higher bandwidth. To change from one value to the next in an instant requires a massive amount of bandwidth and would be part of a waveform with a frequency well beyond human hearing. Thus when we digitise and when we convert back to analogue we band-limit the signals with filters. The band we are interested in is human hearing, which does not have nearly enough bandwidth for a stair step. Any steps that did exist are filtered out simply by the conversion process each way. The output of a DAC is exactly the same waveform as was captured by the ADC with all the frequencies in the band limit intact. No stair steps, no loss. Totally lossless. This ignores any problems introduced after the conversion, as the signal goes through amps and cables and gets attacked by emi and perhaps ends up in a crap speaker.

@GoatTheGoat 8 лет назад

Full of straw-man arguments and misleading analogies.

@GoatTheGoat 8 лет назад

+Scott Petrovits Quote: The entire opening premise is a straw-man, "Digital media -- compression especially -- is perceived to be super elite. . ." After fifteen years as a computer/software engineer I have never encountered that perception. On the contrary in my experience digital media and compression especially is perceived to be a problem with an academic solution. ie. look up the solution in a textbook. New solutions don't appear in this space, not because development is difficult, but because adoption by the market is cost prohibitive. It is more economical to continue broadcasting HD TV with mpeg-2 because HDTV's have mpeg-2 decoders built in. If HD TV broadcast switched to a more modern codec everybody would complain that their TV didn't work anymore. Telegraph is a very misleading example of an early analogy of a modern digital system. Telegraph is very strictly married to Morse code which is a quinary (base-5) system. Where modern systems are universally binary. Furthermore Telegraph/Morse utilizes a universally analog axis to encode and reconstruct its data (time axis -- short vs long dashes interrupted by gaps). Quote: "The sampling therom states that not only can we go back and forth between analog and digital, but also lays down a set of conditions for which the conversion if loss less. . ." Only true if those magic "conditions" include bandlimiting the original signal. But bandlimiting an arbitrary analog signal is by definition NOT a loss less process. That is as far as I can watch without straining a muscle from rolling my eyes every two seconds.

@FernieCanto 8 лет назад

+Ryan Patterson CHECKMATE ATHEISTS.

@FernieCanto 7 лет назад

Çerastes But he said he's an engineer! He *must* be right about everything!... even though he said the telegraph is not an example of digital signal because it's not binary, which is utter bullshit, but he said it in a RU-vid comment! It *MUST* be correct!

@karlfife 7 лет назад

RP: Pedantic. 'Digital' connotes being quantized, not being binary. As for bandlimiting... I'm rolling my eyes at you. Are you pining for the dog whistles removed from your Red Book recordings?

@enneff 3 года назад

If you watch past his tongue-in-cheek intro you'll find that it's actually full of clearly-explained concepts about digital media encoding. I feel sorry for you that you can't even watch a couple of minutes of something without thinking of all the ways you can disagree with it.