The Surprising Origins of Digital Audio Sampling Rates

[@markhesse2928]

Always enjoy your educational videos. This one reminded me of a video by Techmoan from about a year ago demonstrating an early digital audio recorder that used VHS tape and it seems a fitting complement to yours since it includes a little history of PCM recording. kzbin.info/www/bejne/jYendKuKqdNnhbM

[@FilmmakerIQ]

Wow!!! that's right in line with what I was talking about. Never saw that before!

[@bradcole8191]

@@FilmmakerIQ The fabled Alesis ADAT.

[@video99couk]

@@bradcole8191 That's a bit different, as it's multi-channel and doesn't have to retain compatibility with a normal video signal. I have ADAT and DTRS machines in my studio.

[@Capturing-Memories]

@@video99couk Indeed, ADAT and Tascam 8mm are multi channel recordings, they use digital signal not analog monochrome video signal.

[@35milesoflead]

@@FilmmakerIQ Techmoan is an awesome channel. It's geek lite.

[@bobbysands6923]

I have a "digital" VCR--Toshiba--made in the late 80s. It has analogue ins and outs, but the audio is PCM. My band recorded an entire CD with this format because we couldn't afford a DAT machine. When we went to master our mixes, we took the VHS tape and the mastering house was able to decode it digitally. It was amazing at the time. The VCR cost about $1100, and I still have it, and it still works.

[@robfriedrich2822]

I saw an early version of the VHS tape based digital recorder from about 1980, pretty advanced. I used VHS HiFi for mastering and later Minidisc

[@flooey580]

I had this same machine: the Toshiba DX900. Known as the "poor man's DAT." If yours still works, consider yourself lucky! The most obvious thing to go wrong is the rubber idler wheel turning into a hockey puck and any of the rubber belts disintegrating - but that's common with any older tape machine. I mixed & mastered many home studio projects on it with good results (the audio specs of this machine probably exceeded what I was putting into it!). I'm pretty sure it used the Sony PCM-F1 format but was 14-bit instead of 16.

[@lucasrem]

bobby sands You needed Sony Digital Beta machines, you need the I/O to make masters!

[@lucasrem]

bobby sands You needed adaptor as the Sony PCM, too get it in PCM out of the VCR cassette. Many VRC recorders supported the PCM audio recording on VCR. You should be able to read the tapes you made, restore them, are they still good now?

[@lucasrem]

Sony sold VHS digital too, the PCM 1 adaptor is able to master them on a computer now Digital Beta was for cameras and cheap broadcast.

[@ian_b]

Just as an anecdotal data point, I've been using 48k since the late 90s when I started using Cubase VST with a SoundBlaster Live! card and found that the card wouldn't sync properly at 44.1k, because it was internally 48k and doing an inaccurate downsampling to get 44.1. It's funny now remembering how I was annoyed at the extra disk space it used, totally irrelevant now. But back then, when I'd spent a lot of cash (for me) on a "massive" 17.4GB hard disk for audio recording storage, it was a significant overhead.

[@Liam3072]

Haha, believe it or not, I am STILL using a Sound Blaster Audigy 2ZS for audio production, which relies on an evolution of the chipset used in the SBLive! and has the same "48khz or inaccurate downsampling" limitation (the original SBLive card relies on the emu10k1 chip, while the Audigy2ZS relies on the emu10k2 chip... there's even an Audigy Rx released a few years ago relying on an emu10k2.1 chip which still caries this limitation). Which leads me to usually work with a 48KHz workflow, and then manually downsample to 44KHz in software if need be. It also seems to me that, as video and audio become more intertwined (e.g. people listen to music on KZbin), 48KHz is increasingly becoming a standard even for audio only.

[@ian_b]

@@Liam3072 Excellent! I've actually considered building a small PCI machine just so I can use my SBLives again. it feels to me that in many ways we've gone backwards in PC audio, I mean the SBLive is over 20 years old and was an absolute powerhouse. And I was actually running two in the same machine using modified drivers! My final production config was one using the Live! drivers and one using the E-Mu APS drivers which had some great effects. Good days.

[@greatvedas]

I totally get a feel of that 17.4 GB HDD. I have a similar memories!

[@escalator9734]

@@ian_b Or you could get a usb audio interface that has no issues with whatever sample rate you use, better dac's, asio drivers, no ? is there a reason to use soundblasters/ non audio prod focused pci cards ? (this is a real question I'm not trying to be antagonistic)

[@noop9k]

@@escalator9734 and, if he insists on Creative, could use 0404 or, say, 1212M. Older recording audio cards can be often found for cheap. Especially the PCI ones. Or Firewire, though getting it to work reliably on Windows is not easy.

[@zedcaster]

I share a bunch of your stuff with my TV students, I hope you haven't minded ;). I've taught the 44.1 vs 48 lecture for ages but without the PCM background. This is another gem that I can add to help make my dull lectures more engaging. Thanks John Hess!

[@FilmmakerIQ]

Share away! That's the whole reason why they're in KZbin!

[@9899074547]

The old filmmaker IQ is back 👍🏽

[@FilmmakerIQ]

Who are you calling old? :P

[@9899074547]

​@@FilmmakerIQ Haha, not you but the nice explainer videos you used to do back then. I remember I binge watched these videos in a single day few years back. Keep it up Benjamin Button 🤪

[@musar03580]

Yes. More like these, please. This type of IQ video is the reason I subscribed in the first place.

[@c2ashman]

YES, exactly. I miss the old explaining tech videos from Filmmaker IQ. Thats why I watched his videos in the first place.

[@arfmf]

@@FilmmakerIQ Old, (definitely) not obsolete.

[@davidwillmore]

As a computer engineer with a background in signal processing, I would like to commend you on a very well presented summary of this issue. I found no issue with any of the facts you present. Sadly, that is very rare. You, sir, are a wonderful resource and I hope people appreciate you for it.

[@PixelPipes]

You're the first person to explain this in a way I could actually comprehend.

[@garyaduke]

Great video! It is very interesting to learn why CDs use 44.1kHz. My job involves processing audio recorded from telephones. Historically, that's usually involved a sample rate of 6 or 8 kHz (due to limitations of the old hardware), although, more recently, there's been an increase in the use of 16, 24, 32, or even 48kHz. From that perspective, 44.1kHz audio is a major pain, because a lot of software used in the telephony sphere (e.g. speech recognition) still tends to expect 8kHz. If someone records some audio externally and then wants to import it into the system, it's a lot easier to deal with 48kHz than 44.1kHz if and when it's necessary to convert it to 8kHz!

[@joshuascholar3220]

I once wrote a resampling library for a phone system. I did that by filtering the signal (so it wouldn't have anything over the nyquist of the target rate) then making sure it was oversampled by at least 4 * the target rate (sinc filters for oversampling), then using splines to interpolate down. That's a method that works for arbitrary rates. It's been used in synthesizers.

[@arxaaron]

Excellent. Just excellent. My 50 year media production career started with broadcast TV in the 1970's golden age of analog and rode through the evolution to digital audio and video production (starting with the introduction of character generator computers and the Ampex Digital Optics devices circa 1980). Always been fascinated with the tech of the various format standards and how the analog legacy dictated a lot of the digital video format decisions, but I didn't know about the direct influence of analog video legacies on digital audio standards until I saw this video. Thanks! Liked and Subscribed!

[@TuckMuffin]

Love videos like these! A surprisingly in-depth review of some foundational material along with historical and contemporary context. Learning disguised as a good time, great job!

[@wado1942]

I knew about 44.1K but never got 48K, especially since 50K was already an established standard. Thank you! Side note, engineers of earlier digital recorders believed 60KHz would be ideal (you need room for the antialiasing filters to work) but it was pushing the tech a little too far, so they settled on 50K.

[@FilmmakerIQ]

Decca used 50khz in a lot of early digital recording experiments of the late 70s in Europe. It might be a stretch to say it was a standard because nothing was standardize yet, but it was common.

[@wado1942]

​@@FilmmakerIQ Soundstream, 3M and Mitsubishi also used 50K. Those were the most common digital multitrack systems in the late 70s. I forget what Denon used off the top of my head but it was somewhat less. BTW, I'm glad to see you posting again.

[@scottlarson1548]

In the late 80's the amount of data you could put on a CD seemed unimaginable to us. The place I worked for put all of our dozen applications onto a single CD with space to spare and that was writing the contents of multiple hard drives onto this seemingly bottomless media. You could buy a 600 megabyte SCSI drive but the cost was horrific so we didn't.

[@FilmmakerIQ]

I remember the Jazz Drive... We'll never fill up a gigabyte!

[@elimalinsky7069]

I remember that hard drives only started to catch up with CD capacity around 1994. It's so weird these days to imagine optical media having much larger capacity than hard disks or SSDs. Even quad-layer Blu-ray tops out at 100GB, while you won't find a hard drive with less than 1TB these days.

[@Rhythmattica]

I worked in a Electronic retailer here in Sydney, Aus (Dick Smith) in the early 80's... (A competitor to what you'd call radio shack, here known as Tandy) and we had a Tandy (RS) TRS80 clone called the System 80... I sold a 5Mb Drive for $15K (Aus), which also required a $1.5k I/o card option to use it! Fun times indeed. Thats 1 min mono/ 30sec Stereo in sample currency @44.1 / 16bit......

[@lucasrem]

Scott Larson unimaginable? the problem is speed, the scsi I/O interface is not fast enough to store it, VCR was.

[@scottlarson1548]

@@lucasrem Store what?

[@drewcollier1138]

We are really lucky how futureproof the 44.1 kHz CD quality standard was. It did, does, and will always sound good. Still higher quality than the audio of almost everything we stream today. It's hard to believe such a good quality standard came in the 1980s while video standards arguably didn't get futureproof until the Blu-Ray/HD DVD came out.

[@robfriedrich2822]

It was optimized by dithering on the one hand. On the other hand, one could use a higher dynamic at home. It's not allowed to disturb neighborhood, the room is noisy, consumer speakers can't handle high sound levels.

[@Stefan-]

Yeah, they really knew what they were doing when they set that standard. On the recording side 24 bits is prefered, 16 bits is just fine for a mastered music track though.. I record, mix and master my band and have the equipment needed for that so i have both experience and knowledge of that, im also a service tech educated in the Audio/video area.

[@Henrik_Holst]

@Leif And they should be lucky that they cannot hear higher frequencies than what we mere mortals do as the so called high frequency recordings contains lots and lots of noise in the upper frequencies: kzbin.info/www/bejne/g4C5hZJqrrRontU

[@kaitlyn__L]

@@Stefan- have you tried floats for recording instead of ints? Some of my music production pals swear by it and some HATE it.

[@Stefan-]

@@kaitlyn__L No i have never tried floating point, it seems to be just like safety net if you dont do your homework and record at proper levels (read: not to high levels) it wont make it sound any better than 24 bits and you will apparently have wav files that are 30% larger or maybe more, with more data it will also take more CPU power of your computer. If you dont clip the audio in the DAW when recording 24 bits is just fine, you dont have to worry very much about the recording level maybe being somewhat low either since 24bits has 256 times more dynamic range than 16 bits so there is plenty of range for audio production.

[@andydelle4509]

Also note that 48khz is still the universal standard in all HDTV formats, UHD as well. Standard HD frames rates are 23.98hz, 24hz, 50hz, 59.94hz and 60hz. Note the 59.94 still hangs on even though it's need is obsolete in ATSC. But because of conversion to and from NTSC, it lives on.

[@fluffycritter]

All of my video cameras record in 59.94, while my phones record in 60. Friggin' annoying.

[@joystickmusic]

The nice thing about video standards, is that there are so many to chose from.

[@BobDiaz123]

Going to digital does require that any frequency above the Nyquest limit must be filtered out or aliasing will occur. However, the best analog Low pass filters we can make are a sixth order filter with-36 dB/Octave. (Roughly 1/64 the voltage per octave) That's just not enough and any ultrasonic sounds will appear as lower frequency sounds due to aliasing. The higher the sampling rate, the greater the reduction in ultrasonic noise. No surprise that some recording studios like to sample at even higher rates of 96,000 samples per second and higher. Digital filters have an impressive reduction of over -100dB per octave, which puts all ultrasonic sounds into the floor. After that the digital audio can be reduced to a lower sample rate.

[@FilmmakerIQ]

Well put!

[@eliteextremophile8895]

To be honest, there's not much reason to go over 48kHz when you're using somewhat modern professional plugins. First and the most obvious is, 96kHz almost as severe aliasing issues as 48kHz. Sure, you have 2 extra octaves of headroom which will get rid of the aliasing issue when using saturation in one stage, but adding stages will introduce just as much artifacts as 48kHz. 192kHz would get rid of most of the aliasing issues for first two stages, but then again, it's not very practical and still not enough. You can add lowpass to every stage and get rid of the aliasing, but having 192kHz audio files being processed in three or more stages with lowpass in every stage is very inefficient. You get much more efficient results using 48kHz and 4 or 8 times oversampling in plugin (some even has as high as 16 and 32 times oversampling without being too hard on CPU) and it's also much more efficient in reducing aliasing. Then there's the 44.1kHz vs 48kHz and when using oversampling, the difference should be very minimal, right? With 44.1 filtering has to be extremely steep when removing stuff over 22.05kHz without affecting signal at 20kHz. With 48kHz you basically have twice the headroom in filtering out the unwanted stuff that creates the aliasing. So basically 48kHz is twice as good as 44.1kHz and as I've stated it's very inefficient to use higher than 48kHz. Then again, not everyone uses plugins with internal oversampling. Also, someone might as well have Threadripper and billion terabytes of space to utilize, so might as well record, mix and master at 192kHz.

[@BobDiaz123]

@@eliteextremophile8895 Clearly 48kHz is better than 44.1kHz sample rate. Even at 48kHz, the Nyquest Frequency is 48kHz / 2 = 24kHz. Once you exceed the Nyquest Frequency, the frequency appears to go down. That is 24kHz is 24kHz, but 25kHz appears as 23kHz. This is due to aliasing. Assume that we start a Sixth Order Filter at 12kHz, it would be -36dB (1/64 the signal) at 24kHz. Just not enough to fully remove the high frequency noise. I've never staggered a bunch of 6th Order low pass fillers at different frequencies, like 12kHz, 15kHz, 20kHz, to see what happens, so I'm not sure if that's the answer. The reason I suggest 96kHz sample rate is that there's so much head room for a 6th Order filter. The input frequency would have to reach 48kHz before the Nyquest limit kicks in. Assuming the Low Pass is set to 19kHz, 48kHz is 2.5 octaves higher in frequency and offering a -90dB drop in signal level. From that point, a digital low pass filter can be over -100dB per octave and remove anything above 20kHz. Now if staggered low pass filters do the job, the 96kHz sample rate is unnecessary.

[@kelownatechkid]

Great points Bob. 96k is a common choice in the studio, as you mention. For example, I use the DEQ2496 as a DSP and it unsurprisingly operates at 24bit 96kHz. That model has been around since at least 2003 and 96k was in use long before then. There are so many benefits of higher sample rates, especially since common hardware can decode at ridiculous rates like 32bit 384k and disk space is limitless in the studio. Though in my opinion, for (non-hifi) consumer audio 16bit 44k is good enough. Edit: I should note, this does depend on the studio and software/setup in question. Sometimes there are reasons to avoid higher sample rates.

[@johnhmaw]

Thanks John. I have missed these tutorial videos, which seem to have been replaced by live stream content and similar. This is what I subscribed to the channel for, and I am delighted to have watched this one. I hope you will be able to make more of these very educational and interesting videos in future. Thanks.

[@FilmmakerIQ]

The live streams are a way to stay relevant in the KZbin algorithm. And they are also personally fun for me to do. The problem is these videos take a tremendous amount of research to pull off and KZbin favors people that consistently add content.

[@kaitlyn__L]

@@FilmmakerIQ I’ve also heard they can actually tank your Algorithm Recommendation Likelihood (or whatever it’s really called) if the streams have low view counts, allegedly it considers them similar to video view performance. So they can be a real double-edged sword depending on various factors. I’ve seen a lot of other channels make a separate streaming channel and then promo their videos in the streams. Not suggesting you have to do any particular action, but I figured I should share the knowledge!

[@FilmmakerIQ]

I see it more as telling KZbin, hey I'm not dead!

[@petertorda5487]

Despite, that CD Audio standard is 40years old, only few records has been able to utilize it's 96db of dynamic range. And people believes that if they don't have at least 32bits/ 196Khz, sound will be suffering on they precious soundbar. :-D

[@joshua43214]

Takes very good front end equipment for anything past 19bits to not be lost in the noise floor, and very good back end equipment to be able to hear it. 20bits is really the effective minimum. 24bit audio does sound better than 18bit, but very few people have even seen equipment that can render it, let alone heard it.

[@petertorda5487]

@@joshua43214 It makes mostly sense in phase of recording in studio, as amater musician I'm sampling at 44.1khz@24bits from my old synth rack modules. But if you are good in mastering you can make excellent 44.1khz@16bit downmix (not myself :-)), which will blown your mind on good D/As, amplifier, and speakers. I don't forget to hear Vangelis album DIRECT (CD) on Mcintosh mc2120, trebles and basses has been on 0,0, almost showcase dynamic range.

[@fluffycritter]

yeah recording at 24 makes sense because having the extra headroom gives you a lot more leeway with your recording levels while avoiding clipping, but there's not much reason to make your final rendered output more than 16bit. I've never met anyone who's been able to actually tell the difference between a 16bit and 24bit recording on a proper A/B/X test, or even hear the difference waveform between the 16 and 24-bit version without cranking the volume up so high that the original audio would destroy their eardrums.

[@petertorda5487]

@@fluffycritter exactly :-)

[@MatthijsvanDuin]

@@joshua43214 As long as proper dithering is used the only difference an increase in resolution makes is a decrease in quantization noise floor, so it's a bit weird to refer to the difference being "lost in the noise floor" ... though you presumably mean the analog noise floor as opposed to the quantization noise floor.

[@deaniweenie]

Well, this is as good an explanation as you'll ever find - very clear with minimal techno-math. I'm never disappointed when I visit this channel

[@JohnMcCormack]

That's a well done video, Sir. Well researched, very informative and very well presented. Thank you!

[@markbarrett8180]

Your videos are quite simply the best, most informative free online resource for the technical aspects of film making I’ve come across. Always a delight and your presentation is top notch too

[@LanceCampeau]

Loved this one.... answered a lot of those random questions I've always wondered about...

[@maltoNitho]

Fascinating! Tyvm. Good to see you back.

[@RXP91]

Damn, every subject i think I know well, you teach a ton more. I remember marathoning your education series in Feb2020 when on holiday and learnt a ton. Thanks!

[@ternurasan]

Your explanation is so good and on point! Congrats on the video! Hope to stay and learn lots!

[@ZILtoid1991]

I personally choose 48khz for my game engine since I can implement millisecond-based timing quite easily with that sampling frequency, otherwise it can stretch 44.1khz samples to 48 if needed.

[@AnalogueGround]

Really well presented. i already knew the history but i couldn't have got it across as concisely and completely in 12 minutes - fantastic!

[@Uuuuuuurrgggggghhhhh]

Very interesting and very well explained. This was my "serendipity moment" (learning something you weren't looking for) for today.

[@Ianochez]

This content is soo dense and presented efficiently that I checked half way through thinking the video was twice as long ahaha

[@shalomcu]

Super clear explanation and great visual aids... Nice work!

[@nielsott]

Nice video! To add to this: Nowadays with good conversion algorithms you won't notice it if you convert from 48 to 44.1 if you don't do it over and over again. Thus I do all recording in 48khz, even for CD productions. So in the end I always get a video version and a CD version. But seriously, CDs are almost dead so the very most audio I produce goes into video anyways. One can discuss 96khz sampling, and people do. But also there is a difference between 96khz processing which basically is a 2x oversampling on the entire process, while still having recorded at 48khz, and 96khz recording. But anyways, most videos on YT are being watched on smartphones probably, so anything will be good enough.

[@FilmmakerIQ]

That's why I'm thinking everything is going to start to move to 48... Though strangely, KZbin is 44.1 and I don't think anyone notices the difference.

[@nielsott]

@@FilmmakerIQ I think YT goes 44.1 because in the end it saves them data traffic. As simple as that. Most likely re-encoding lossy formats will give you more damage than resampling from 48 from 44.1 once in the entire process. This is why I like to upload videos with 24bit PCM audio @48khz to YT. Most times I upload CineForm so that also no bad video (h264) will be transcoded to other bad video (VP9 or yet another h264). So good video to bad video only. :-D Sad thing is archived video live streams that get messed up by YT quite noticeable, at least the video. The audio less noticeable. But of course you can't stream CineForm, your internet uplink would simply break down. For audio there's platforms like BandCamp that actually do deliver 44.1 FLAC (lossless audio). Luckily.

[@kaitlyn__L]

@@nielsott what’s your average bit rate in your CineForm files? :D

[@nielsott]

@@kaitlyn__L The video bitrate is… don't know exactly, I think it's 1 to 2 gigabytes per minute, depending on the content. The audio bitrate is 24bit 48k PCM stereo, so 2.250kbit/sec if I used my calculator correctly.

[@unityoc]

Great to see you back John and thanks for this very insightful video!

[@Nets-nutsBr]

Hearing you is a nice class about how to use the voice pitch and level to explain physics of audio waves. Simple and it works ... wow!

[@JKVisFX]

I have been recording/rendering my personal electronic music projects at 48k/24bit or 48bit for yours now. With kind of powerful workstations like the one I have, I can pretty much throw whatever I want at it.

[@virginia7125]

I remember some high quality formats that really didn't stick around. SACD & DVD audio. I have Natalie Merchant "Carnival" on DVD audio. Only a few DVD players would play the digital audio. I never purchased any SACD discs. I have "Las Ketchup" & "Axe Bahia" on a video CD. This was the forerunner to DVDs. Again, only a few DVD players played this oddball format or it only played on a computer. XP at the time. I remember the DAT tapes & Sony mini discs. Cassettes had chrome ıı or metal ıv & Dolby B or C. Later came S. To erase the metal required strong magnets in the cassette recorder. TDK & Teac made some studio version cassette tapes with ball bearings. (not for retail sale) That was extremely clean with headphones on. Later came MP3s. Standard was 128. Some car CD players would play those. Then you had the up to 320 & variable bit rate. WMA & AAC was out there too. Real Player had some oddball AT8 format.

[@loginregional]

Y aserejé ja de je De jebe tu de jebere Sebiunouva Majabi an de bugui An de buididipí Aserejé ja de je De jebe tu de jebere Sebiunouva Majabi an de bugui An de buididipí Aserejé ja de je De jebe tu de jebere Sebiunouva Majabi an de bugui An de buididipí Holey Underwear, I actually have this version. Don't ask me why. And I said a hey, a hah ... dammit I don't recall... but I _do_ know it's not jibberish, it's almost a patois of english pushed into spanish sounds. "My hobby on the boogie and the booty ..." And the DJ that he knows plays the hymn (anthem) at twelve (midnight), for diego it's his favourite song, and he dances and he likes (enjoys) it and he sings it... I said a hey, ahah...

[@v12alpine]

DVD audio is normally compressed in something inferior to modern mp3. Not surprised it didn't take off.

[@meneerjansen00]

What a brilliant video. Sometimes "every-day things" can be so complicated.

[@lucasrem]

Codecs, he is not understanding it, too complicated?

[@squidcaps4308]

48k also makes circuit design just a bit easier, you can use less aggressive anti-aliasing filtering.. and it simplifies the math a bit in digital side. The other number, 24bit is even more important specially in broadcasting as it allows for more headroom. 16bit is what we need but when you are working with live signals that are often unpredictable you will enjoy having those few bits extra. This works well also in other audio production, 24/48k is what vidiots wanted and it happens to also make things better elsewhere. So, thanks.. i guess.. There is no need to go higher, we can prove it quite definitively, also when it comes to processing but with one crucial detail added: 48k requires oversampling and filtering at each processing stage (384k project sample rate is equivalent to 48k if latter uses oversampling scheme, it also uses 8 times the bandwidth and file space and has a higher processing load. 192k is worse and 96k is damn right idiotic, it is considerably worse to a point where it can be audible..). So, turn all of those "high quality" and "oversampling" buttons on in your plugins, they are there for that specific reason. Especially important with all processing that affect the time component of the signal in anyway, this includes compressors.. but not stuff like linear saturation or simple gain, distortion or even delays (repeating is not modifying the time component.. ). Chorus and phasers are especially problematic and oversampling is absolutely mandatory. 24/48k is perfect format for all audio production.

[@philiptong4978]

assuming there is an abundance capacity to process/store that higher density data, wouldn't it be better to ditch oversampling (i.e. interpolation/guess) but use original recorded data at the equivalent sampling rate? (48kx8=384k, work for noise shaping, editing, downsample for various consumption)

[@FilmmakerIQ]

The issue is computational processing power. Audio mixing involves often hundreds of simultaneous tracks with long effects chains. Where as there's some flex and give with sync when it comes to video sync (there's about 100ms wiggle room before you can tell something is out of sync), there is almost zero wiggle room when mixing audio tracks. Add on top of that oversampling a recording to 384khz is just downright pointless as it introduces more sampling errors and virtually no microphone can generate the response that could effectively utilize the frequency space. Now there might be some filters that might benefit by the oversample but they can do that on their own.

[@FilmmakerIQ]

My understanding is limited but basically the underlying concept is that going beyond say 88 or 96khz, all your doing is capturing frequencies we aren't going to hear anyways. An in doing so we introduce intermodulation errors. Digital sampling theory states that we can recreate perfectly any signal that is has half frequencies that are half the sample rate, we need extra headroom to band limit the signal to prevent aliasing. But after that, we're just eating resources that's won't really get us any more benefit and potentially leadimg to degrading the signal.

[@philiptong4978]

it's about collecting (record) as much audio PCM/PDM data as you can, convert/process/edit using as much precision as you can, then downsample (depends on your targeted viewer/audience and expected playback equipment) for the final release, whether it be 16/48 or 44.1 or 24/96 or DSD64 or DSD128 or ...

[@Mxsmanic]

I am happy to see new videos from you.

[@loginregional]

First video, looks ok. You're not the Frugal filmaker, but... you might get me hooked. Oh, and a shoutout for mentioning Umatic, which is the second time in a couple of days to be mentioned, since I had been talking to a friend about some old editing equipment and flying erase heads. No mention of "white-a-key" - the technique we used to get a clipped area out of the picture without a fancy switcher (piece of white paper, careful positioning, clip to white which dropped out and you see what's in the other channel/camera/playback source). Of course, you still needed the TBC. Only criticism: you dropped 'kilo' twice, just before the moire moment. Nobody but me noticed.

[@AdemVessell]

Great explanation! But then why am I so tempted with 96khz?

[@FilmmakerIQ]

There's legit reasons to work with it as it did make it easier to develop filters to cut off the unwanted high end frequencies. But for playback it is really superfluous.

[@AdemVessell]

@@FilmmakerIQ Yes, for Masters I find 96 or even higher is a nice place to be. Personally I use and always have used 48khz for both music and video. There's been notable differences to my ear when using various effects and tools in regards to the audio processing :)

[@AttilaSVK]

I have one of Sony's PCM to video adapters, not from the PCM-1600 line, but the later, consumer oriented PCM-501ES model. It's similar to the Technics PCM deck that Mark Hesse posted, but it lacks a transport. Instead, it has a composite video input and output, designed to be hooked up to a VCR (be it VHS, Betamax or something else) It uses a different video encoding method than the 1600 series, but the three samples per line are still clearly visible. Of course, it does record and play back at 44.1 kHz. I wasn't sure about the origin of 48 kHz. I thought it might had to do something with DAT tapes, since that was the earliest widespread use of the 48 kHz sampling rate, outside of the film and video world.

[@AttilaSVK]

Oh, and Sony's PCM-501ES (601ESD and 701ES) offered 16 bits of resolution, while the Technics deck could only do 14 bits. Here's a video of my PCM-501ES playing back some music from a VHS cassette: kzbin.info/www/bejne/g32ndqqLectorac

[@gregfaris6959]

Excellent contribution!! You will come across a lot of mythology surrounding the choice of 44.1Khz as the sampling frequency for the Compact Disc, but this is the true explanation. Rotating-head video machines were the only recording format we had at the time with the recording density needed to record continuous program at a sampling frequency above the Nyquist/Shannon rate of 40Khz, to deliver 20 Khz audio - and the U-Matic machine chosen for the Sony PCM 1600 - 1610 and 1630 mastering process delivered 44.1Khz using three frames per sample. Other formats did exist; As early as 1980, Sony's F1 EIAJ format allowed recording of 20Khz audio on VHS tapes at a lower cost, and a sample rate of 44.056Khz, but the PCM 1600, later to become the 1610 then 1630 had already established itself as the industry standard for mastering of CDs.

[@dream.machine]

Glad you went over this! Great video 👍

[@Capturing-Memories]

I have the Sony PCM-601ESD PCM adapter, I recorded a lot of albums from my computer to VHS and Video8 tapes, a lot of fun. But there are computer programs that can do that now if you have a video card that can pull the video off of the tape and convert it to PCM inside the computer using special app or even record a PCM audio back into the video tape without needing the PCM adapter.

[@dtsdigitalden5023]

That was fascinating! It's one of those things I just took as a given, and never realised I wanted to know, but then you popped onto my KZbin feed, and I realised I really wanted to know this. Great explanation! Subbed.

[@angelwars3176]

Thankyou Sir. A much needed detailed but still understandable answer to an often asked question.

[@harfenspieler]

Enlightening and so pleasantly presented! Thanks very much for this.

[@shinypb]

Mind blown. Thanks for making this-I appreciate you! ✨

[@flamencoprof]

01:03 to 02:56 That's at least two or three one hour lectures from when I was trained in the Eighties. I congratulate the Filmaker for his clarity and jargon-free explanation. I was especially impressed by "Sound is just compression energy, flowing through the air", but the following description of PCM is just the best.

[@PaulKretz]

*The greatest cine-tech tutorial channel!* Thank you, John!

[@unvergebeneid]

One little thing to add: one might wonder "Why not just use 40 kHz these days, given that we don't record on video tape anymore?" Well, to avoid aliasing artifacts, you need to use a low-pass filter on the recording to remove all frequencies that are higher than what the Nyquist-Shannon sampling theorem allows. The steeper the filter, however, the more artifacts the filter itself introduces. An infinitely steep cutoff would be a disaster, therefore you want to have a bit of give in the sampling frequency. Which also means that in principle, 48 kHz _could_ sound better than 44.1 kHz, even though no human can hear those frequencies.

[@eliteextremophile8895]

Dan Worall, the Attenborough of audio engineering, goes very in depth about the samplerates from audio engineering standpoint. I couldn't recommend it more. Video is called "Samplerates: the higher the better, right?" and it's on FabFilter's KZbin channel. Even if you don't have any prior knowledge about audio engineering, it will demystify the whole topic very efficiently.

[@unvergebeneid]

@@eliteextremophile8895 I will have a look later but it's important to note that I was just talking about sample rates close to the minimum, not about 192 kHz and stuff like that, that's not doing anything for you other than to overwhelm your audio equipment. Unless you're a bat researcher, you don't need that sample rate. More than 16 bit, however, can be quite useful and at least has no detrimental effect.

[@unvergebeneid]

@@eliteextremophile8895 very good video indeed. As I'm not a music producer myself, I definitely learned something about that part of the tool chain! Here is the link BTW kzbin.info/www/bejne/Y5umqHypiZWLbq8

[@paulstrahan4670]

@@unvergebeneid I completed a Diploma of Audio Engineering at SAE back in 1994 and embraced digital audio almost immediately when it was released. While I agree that anything over 20k is not audible to the human ear (most people are lucky if they can hear much above 16k) it is interesting to note that frequencies above our hearing threshold still have an impact on the end result due to harmonics and intermodulation distortion. This is one of the main reasons that MP3 compression never sounds as good as a raw wave file, because the frequencies that it eliminates for the sake of file space may not be in the general hearing range but have an indirect effect on the rest of the audio that is in the range of your ears. Please don't take this as a condescending or rude comment it's just something that I learned a little about in my course and have also researched a little over the years, I run a small Music and Video production studio here in Melbourne Australia and generally use 48k as my standard sample rate and 24bit for that little extra dynamics and lower noise floor even though my audio interface is capable of 192k sample rate.

[@unvergebeneid]

@@paulstrahan4670 isn't intermodulation distortion a result of imperfections in audio equipment and should ideally not occur at all? So if you lowpass those inaudible frequencies away, wouldn't that bring the signal closer to the ideal? As for compression, from my experience, mp3 is most noticeable in percussion, not in anything harmonic.

[@ChrisTuttle]

Simply put, incredible video! Even the slides are awesome. I've been text messaging pics of them to my buddies in real-time as I've been watching. Sure, I could have sent a KZbin link, but hey, what's one more conversion standard? As they say, the good thing about standards, there's so many to choose from!

[@akhlikeys123]

The process of recording sound stored in the form of thousands of indivdual measurement each at a discrete unit of time called.. very thanks for a super video....

[@kshred3043]

It really helps me to understand technology when the history of the development explained so clearly. For example what was an arbitrary, unconstrained choice versus what was guided/dictated by legacy. Thank you.

[@m1ian]

Thank for that John Very Interesting answered many Questions

[@KRAFTWERK2K6]

It's interesting to see digital audio data being saved as picture information on videotape. Just as digital data for computers used to be stored as binary code audio-signal like on Commodore 64 Datasettes or ZX Spectrum tapes. A clever concept. Something similar was later used in the 90s for VHS tapes to save digital files from harddrives to a Videotape. The software was sold for PC and came with a video-output card that had to be installed into the computer. You would then select all the files you wanted to backup on VHS and it would then pack everything into a archive dump and export it as a Videosignal which was a black & white video signal (series of blocks that looked like animated barcodes and QR codes).

[@owendoconnor]

Similar to using dialup modems to transmit data over public phone circuits which were designed to carry only the narrow band of frequencies necessary to convey intelligible human speech. Those ZX Spectrum tapes sounded much the same as a modem.

[@KRAFTWERK2K6]

@@owendoconnor Yup. You could even hear if a turbo was used or not, which removed the many many redundancy blocks in the binary data, that was often checked multiple times to assure data consistency.

[@tonydotnottingham]

Really enjoyable and well presented, thanks!

[@Crlarl]

VHS PCM is gonna be the newest hipster audio format.

[@hobokenhifi7015]

Kinda right now. People are buying up old Alesis ADAT recorders for really cheap. They're pretty good. I've had one for years.

[@allyemeraart]

it's just pointless to me IMO compared to actual analog, as the same bit-depth exceeded "0 to 100" distortion nature of digital and lack of actual artefacting (saturation, compression) to the audio signal itself leads it to not have any kind of sonic aesthetic/ FX use case in audio production/ mastering. All though, perhaps some hardware has some means of A/D/A that may be coloured, and quite complex ie exiting, stereo processing. I'm sure it has other uses that people would dig up though.

[@hobokenhifi7015]

@@allyemeraart in the scheme of things, yeah they're pointless because we have even low tier stuff that outperforms it. But as a standalone recorder or if you need so IO via ADAT in a pinch or in the rare case you get someone looking to transfer tapes. They're good to have around

[@rredline]

Very very nice clarification. The history of digital audio is very interesting!

[@FluxCondenser]

Fantastic overview. Thank you!

[@rayderrich]

That was more complicated and interesting than I expected, thanks!

[@ampsmashed]

Good to see you back in the saddle Hessy

[@bernaridho]

Rare to find explanation. Great job

[@jittisoolt5927]

Ace. I really appreciate your efforts. Thanks.

[@tempermentalw7269]

Wow great stuff thank you. I am blown away. I'm still recording in 41

[@chrispeoples4606]

John, great video! Will definitely use it in my physics classes as a part of my sound and waves unit.

[@wdavem]

I've played a lot with a pcm-601esd with Umatic and VHS. Very interesting device! Now I've HEARD that sony "Umatic Digital" was the master format sent to CD duplication houses. I can't help but wonder how much data is lost to video drop-outs.

[@FilmmakerIQ]

Probably not as much as you think. Professional tape formats have always been much more robust than consumer formats like VHS. There's some surviving Quad tape from the 1950s that still look sharp today. Plus a digital signal is more robust than analog... A little noise won't really affect the quality. Then I believe some of those early models weren't truly 16bit, they were 13bit and used the extra bits as error correction

[@wdavem]

@@FilmmakerIQ Ok I didn't really factor in error correction, my 601 esd might have something. It can do 16 bit and 12 bit. It seems to break up the sound by frequency and volume across the screen on the left and the right side is one large area that is random looking. I fix many formats of vtrs for a tape preservation service actually (up to 1" type C) and have tried feeding the pcm various garbage to see what it will do. I remember large drop-outs do make a high pitch 'sparkling' noise, but I couldn't quite tell what small dropouts did exactly. BUT I was using a normal umatic vtr. Duplication houses might have had tricked out gear that did direct off-tape RF processing or something. I've had plans to connect 2 video pcm tape playback decks to a mixer and run transition effects on the video signal feeding the PCM audio processor... It didn't like the last mixer I tried, lol. I've got more of them to try now though!

[@Capturing-Memories]

I have the 601ESD too indeed it’s fun, U-magic is a robust system, it allows a lot of tape real estate to avoid drop outs, Besides the PCM adapter have error correction for drop outs.

[@fyabun]

Thanks for boiling it and making it simple. Great job!

[@garethonthetube]

Excellent explanation. For those of you who can't stick with the whole 12 minutes. Engineer 1. '' We need to build an oscillator that runs at a high enough frequency to sample the audio'' Engineer 2. '' This video deck has a load of stuff already in it running at 44.1kHz'' Engineer 1. '' That'll do nicely. Let's have a beer''.

[@owendoconnor]

Later formats like CD, DAT etc were so robust largely thanks to error-mitigating innovations like EFM and CIRC. Those early digital audio formats on u-matic etc seemed to be very much synchronous bytestreams. Did they use any form of error checking like parity bits or such?

[@kaitlyn__L]

In other video comments on other channels I’ve had discussions with people who used U-matic to record digital audio onto in the late 80s and early 90s. At least at that stage they could detect CRC errors because it was a simple checksum, but they said if the errors started flying up (measured on a separate rack mount unit) you had to rewind and re-record that part until the Master Copy was perfect. So definitely far more rudimentary than CD error correction! Merely error detection :) (I’m not sure what the acceptable error tolerance was, though - given RedBook includes standards for bodging a few PCM frames when they’re missing, just one or two CRCs might’ve been fine? But definitely a large increase necessitated going back and redubbing that segment, all made possible with timecodes of course.) Now I’m not 100% if these early recordings were compatible with a CRC error count rack unit, I suppose they might not have been; but given the machines were used for a long time, and they had their own standards for the video frame, and so on, I would be surprised if that were true. Whatever was used in 88 was probably used in 84.

[@owendoconnor]

@@kaitlyn__L great info. It's all coming back to me now. We still were receiving audio masters on u-matic up until 1995. We had a rack-mounted sample rate converter and a dot matrix printer attached to a Sony PCM 1630. And yes, I vaguely recall examining the printout for excessive errors. The only option was to reload the tape. CIRC is a much more elaborate algorithm that allowed some degree of error recovery which parity or CRC did not. I seem to recall that u-matic also had the PQ points encoded?

[@MiddleMalcolm]

Interesting that they tossed the idea of 60KHz as the standard back then, as several of the more respected designers of digital audio converters suggest that as the ideal range of sample rate, for overall accuracy and quality.

[@FilmmakerIQ]

Sampling frequency quite literally translates to storage space. Instead of a CD being able to contain 70 minutes of music at 44.1 it would only be able to contain 51 minutes at 60khz

[@MiddleMalcolm]

@@FilmmakerIQ My bigger point was, had they figured out a way to make the 60K rate the early standard, they might have inadvertently found a more ideal compromise between quality and file size. Now, we are left with "standards", that straddle a range that has been suggested as ideal. 60KHz-ish. ☺

[@FilmmakerIQ]

But really beyond 40ish khz there is no gain in quality.

[@Henrik_Holst]

@@MiddleMalcolm There exists zero needs for anything better than 44.1khz at the consumer level.

[@sswpp8908]

Many audiophiles only listen to standards beyond CD quality at higher sample rates which are based on 48 kHz. Typical rates for high quality audio are 96 kHz (double) or 192 kHz (quadruple) with some even higher standards as well. Also bit depths are increasing to 24 or 32 bits. I don't believe any of these excessive rates and bit depths add any additional improvements, but since storage isn't an issue anymore it's not really hurting to have more headroom.

[@PavelKnyshov]

The high sampling rate makes some sense in terms of practical implementation. Higher sampling rates make it easier to filter out high-frequency "conversion" noise in the "raw" post-DAC analog signal and, as a result, reduce distortions introduced by the filters and noise in the audible range of signal. Often audiophiles choose high resolution due to the "more scrupulous and thorough mastering" of recordings for high resolution release. This makes sense if the release was ACTUALLY mastered by COMPETENT team for a high resolution release. Usually if we compare CD-quality recording with competent mastering and 384/32 recording with “just average” mastering, the choice is clearly in favor of the former.

[@philiptong4978]

better data format and storage/processing capability does not help if 1. delivered program material does not have the data to begin with (e.g. compare youtube AAC compressed, lossy 128kbps to redbook CD) 2. insufficient resolving power on the playback system (and its viewing/listening env)

[@bryanbarajasBB]

I like your use of zooming in in the beginning. I found that out myself when making our chorus's videos to classical music👌Thanks for this video Bro!

[@milasudril]

48 kHz is nice: 3 periods of 64 samples makes 4 ms. It simplifies timing over USB, leading to lower latency.

[@SlyNine]

Which also syncs up nice with 120 and 240hz displays. Dunno if that matters but hey. I like it lol

[@kaitlyn__L]

I couldn’t discern a difference between 44.1 and 48 in various settings, so besides switching to match my music (half the FLACs I have from Bandcamp are 48 while the other half and my own CDs are 44.1) which is a PITA I just allow for resampling. But I think I left it in 44.1 as default so that CD rips wouldn’t need resampled. But if it reduces audio packet latency, I’d happily put it in 48 since it sounds the same to me anyway.

[@milasudril]

@@kaitlyn__L I would store data in the original format. If all you do is listen to audio, latency does not matter that much, and most media players will upsample in real-time. It is however good to configure your audio server ie JACKD to run at 48 kHz.

[@lucasrem]

HDMi 2.1 needed? 192 24 384!

[@mina86]

Citation needed. USB devices are interrogated by the computer at whatever rate is necessary. Furthermore, audio data is always buffered at the output.

[@javiercastro8466]

A number of agencies like the FBI keep older equipment around in the off chance an old piece of recording format is needed to be read!

[@xcoder1122]

As I already saw the incorrect myth popping up once again in the comments that waves close to half the sampling frequency cannot correctly be reproduced, let me point something out here: *This is nonsense* and based upon a complete misconception of how PCM sampling actually works. Every analog sine wave sampled at at least twice the frequency will be *perfectly accurate* reproduced by the combination of a DAC and a band-path filter following it, which is always required for correct PCM signal reproduction. You can use an analog oscilloscope to prove that. There is no loss in amplitude or shift in phase. A sine wave of 20 kHz sampled with 44.1 kHz and converted back to an analog wave will look exactly the same at the output as it did at the input. If you overlay them in an analog oscilloscope, they match up perfectly and yes, it will always have exactly the same amplitude, too. The misconception here is that when reproducing the wave, the signal will jump directly from PCM sample value to sample value but that's not the case at all. If it would do so, it would produce frequencies that are above half of the sampling frequency and this would violate the Nyquist-Shannon sampling theorem. That's why a band-path filter is mandatory for reproductino. It will enforce that no such violation is possible by ensuring that only analog waves are emitted that pass through all PCM sampling points (*pass through* does not mean "end at those", the wave can go far above that point before it turns around) and at the same time are never above half the sampling frequency. And for those who now say: Okay, this may be true for sine waves, but what about other kind of waves? Every wave can be broken down into sine waves, so whatever applies to sine waves applies to any shape of waves. Of course, saw waves are not reproduced as correct "sharp" saw waves, as sine waves above half the Nyquist frequency are not reproduced (and those would be required for perfect edges) but that doesn't matter as you also cannot hear these either. That means to your ear, saw waves always sounded (= "looked like") as the "imperfect reproduction" does look like. So while you can see a difference for square waves (or saw waves) on an oscilloscope, that difference does not exist for your ears as your ears never captured the waves above 20 kHz and to your ears there never wars a perfect square to begin with. This video explains all the theory about this and even proves that theory using an analog oscilloscope: kzbin.info/www/bejne/aKKpm5eHhZx4ppI

[@FilmmakerIQ]

The issue with your rebuttal is while it's true that every frequency below the Nyquist threshold is reproduced perfectly... Every frequency above the Nyquist threshold creates an alias of itself in the lower frequency. So if there are ultrasonic frequencies being recorded at a low sample rate those ultra high frequencies can appear as lower audible frequencies via their alias. Therefore it is necessary to eliminate all those higher frequencies before we sample the audio waveform. But a low band pass need some headroom to operate. So you can get a diminished response on the upper end of the audible spectrum because that's the low band filter taking effect.

[@xcoder1122]

@@FilmmakerIQ A filter is required at both ends, that's correct. You need one prior to feeding a signal to an ADC and you need one after a DAC. Usually it's not just a low-pass filter as you also don't want very low frequencies to pass through, especially not any DC current. And yes, those filters cannot cut off perfectly at specific frequency, they just increasingly reduce the intensity of the signal starting at their cut off frequency. But this is not an issue to frequencies at the upper end. If the filter does not cut off fast enough, then you get aliases in your sample values, however an alias of, let's say 23 kHz, will not appear near the upper end but at a much lower frequency. To get an issue at the upper end, you need aliases of even higher frequencies and the higher the frequency, the more effective the filter will be. Also keep in mind, that the filter can start to cut off at 20 kHz, yet 44.1 kHz sampling frequency means that frequencies up to 22.05 kHz are reproduced correctly, so you have 2.05 kHz headroom to begin with.

[@FilmmakerIQ]

Well 2khz isn't a whole lot of headroom and as I understand it it is difficult to do a sharp cutoff that smoothly. 48khz sampling would double that headroom and 96khz would give you a full octave plus to work. That's what seems to be the real crux of this discussion. The serious folks are not talking about hearing ultrasonic frequencies or trying dispell Nyquist. It's that it's much simpler to filter out the unwanted high frequencies with more hz (up to a point then it might actually harm the product)

[@JonathanEBoyd]

Very Educational as always really interesting to hear the History behind this brilliant video thanks

[@truefilm6991]

Very comprehensive as always. Of course I won't be able to know all numbers by heart, but it's always great to have a deeper understanding. With that said: I think that the Nyquist-Shannon theorem does not hold true 100%, just like the Fourier analysis. The human ear can distinguish a sine and a square wave at 10 kHz. If a sound wave is nothing more than the sum of sine wave components, then the human ear would be able to hear a 30 kHz sine wave, which it cannot. As we know, a square wave only contains odd numbered overtones (or partials), so the first overtone (or second partial) of a 10 kHz square wave is 30 kHz. I believe that we humans are great at breaking down things (analysis), but very bad at fully understanding systems (synthesis), so we assume that our senses are analysing, which they are apparently don't, but rather recognize shapes and patterns. So I'd say that digital audio is very good nowadays, but raising both sampling fequencies and bit rate would make it better. That's why since the 1990s old analog tapes are recorded onto digital at 96 Hz at 24 bit. I had the pleasure of hearing studio recordings at 96 kHz (ProTools through Apogee converters) and it sounded so much more transparent. Just my morning ramblings, thanks for reading.

[@FilmmakerIQ]

I've recorded choirs at 96khz and at 48khz and I can honestly say I cannot hear any difference between them.

[@truefilm6991]

@@FilmmakerIQ might have been the wrong audio content or circunstances. You can hear the difference in signals such as acoustic piano or nylon string guitar. In more complex signals that involve mixing, you need top notch equipment in order to hear a difference. And if course audio engines and rendering heavily improved in the last 15 years or so and 44.1kHz/16 bit sound very good. Just sharing my experience.

[@FilmmakerIQ]

I've done acoustic piano recordings as well but certainly not with top notch recording equipment. I used to record 96khz but then I tried just switching to 48 just to see if I would miss something... Nope. But I get there are certain filters and processes that benefit from the higher rate... Sure. But I'm also perfectly happy with my MP3s so what do I know

[@truefilm6991]

@@FilmmakerIQ Mp3 at the highest setting (320 kbps, 16 bit) is more than fine for most real life situations. I'm not a purist. Yes, certain plugins for example sound best at a certain sampling rate/bit depth. And of course: if you have the storage capacity and CPU power: when in doubt, choose a higher quality. It certainly won't hurt the sound. Oh: and recording a piano for example: mic placing and levels (gain staging) are much more important than top notch equipment. Now that we have much less problems with noise floor, you can go easy on the recording (tracking) levels. But you know all that. :)

[@FilmmakerIQ]

Someone else linked this video and it covers your question about square waves rather ingeniously: kzbin.info/www/bejne/mXq0anyOiLqtq68

[@video99couk]

Have a look at some of my videos where I get stuck into working with the PCM1630 (a slightly later machine using the same format): kzbin.info/www/bejne/pKa3lX2Xfs5sqLs

[@Veptis]

I wondered about the actual history for a while, so I put this into my watch later a long time ago. It just happened that right before this I watched some lectures on compressed sensing that showed how random compressed sensing can to aliasing to reconstruct a signal at less than half the maximum sampling rate. but shanon Nyquist still applies as the smallest interval of your random samples (and how those are distributed) still applies.

[@douggodsoe]

I wasn’t expecting much, most videos on the topic get the historical details wrong, but this discussion was 100% accurate. Pleasantly surprised. Now, can you explain Shannon’s and Nyquist’s theorem and show why it is not just theoretical, like gravity and evolution? Nyquist stated it, Claude Shannon provided the mathematical proof. There are those who think sampling rates are just guidelines, so we must oversample!

[@FilmmakerIQ]

Thank you! It's ironic that I also get comments saying my historical details are 100% wrong. I'm not going to try to prove Nyquist because there's already a video that did it so perfectly, I can not make anything better. kzbin.info/www/bejne/mXq0anyOiLqtq68

[@TheLastRealJEDI]

While I knew most of the single facts , I never got the bigger picture. Now I did. Thank you for this informative and entertaining video!

[@echopathy]

you're a good teacher, man. cheers.

[@SeverityOne]

This was wonderfully informative.

[@dalemettee1147]

I had a 'portable' Beta recorder which could be used to record sound as PCM. Real ahead of its time. It was only mono though. BTW, I noticed the candle stick phone on you desk. Do you know the back story of the ringer current for telephones? It is interesting. Ninety volts @ 20 Hz. I know crazy.

[@mgbspeedy]

Brilliant video. Great job.

[@mightyfixie4283]

Old computer sound cards had the 44.1 kHz sampling frequency as the native one. Newer ones have 48 kHz instead, as such sampling frequency became widespread amongst computer audio systems since the late 90s to make PCs be able to play DVDs. But despite that, most computer audio is still recorded at 44.1 kHz or multiples (11/22/44/88/etc kHz) even if it isn't meant to be recorded on CDs.

[@FilmmakerIQ]

Be careful with words like "most" because I don't think that's true

[@neilthomas2549]

1980! That clunky Umatic audio recorder box was 1980! Wow. That made me feel soooooo old.

[@ClaytonMacleod]

Except there is no resolution. Nyquist-Shannon describes why. Sample rate simply dictates the upper frequency reproduction limit, and bit depth does relate to dynamic range, but it might be better to think of it as how much noise is introduced alongside the desired signal you are trying to reproduce. More bits, less noise. Anyway… A 1000 Hz test tone isn’t reproduced “better” or with “more resolution” when using 48 KHz versus 44.1 KHz. The output would be exactly the same in both cases. Exactly. Nyquist-Shannon sampling theorem dictates this. It perfectly reproduces all frequencies less than half the sampling rate. A higher sampling rate doesn’t give a more accurate result. It doesn’t give better resolution because there is no resolution. You either can reproduce the input signal because the frequencies present are all less than half the sample rate, or you can’t reproduce some of it because the input contains frequencies higher than half the sample rate. As long as it is below half the sample rate you can reproduce it. There is no better or worse. Either you can do it or you can’t. Period. This is a very good demonstration and explanation of how it actually works. kzbin.info/www/bejne/mXq0anyOiLqtq68

[@FilmmakerIQ]

Who are you addressing? I'm not making any of the claims other than calling the audio "resolution" which it sort of is. Sample rate governs the frequency resolution via Nyquist, bittate governs dynamic range resolution.

[@ClaytonMacleod]

@@FilmmakerIQ There is no resolution. As I just said, a 1000 Hz sound will not be represented better by a 48 KHz sample rate versus a 44.1 KHz sample rate. You seem to be saying it would be better since the resolution of the 48 KHz version would be higher. But it is not higher. It doesn’t work that way. The output of both would be EXACTLY the same. Again, the sample rate only dictates the highest frequency you can reproduce. It doesn’t change the quality of the reproduction. It doesn’t change the resolution of the recording. There is no resolution. You can either reproduce the signal 100% or not at all. There is no in between. There is no quality gradation. There is no resolution. Either it works or it doesn’t. Seriously, if you really want to understand it then do yourself a favour and spend the time to watch the video I linked to and you will get a very easy to follow demonstration with examples and a graphic display to illustrate what is actually happening. When someone thinks there is a resolution element to digital audio they do not understand digital audio. There is no resolution element to it. It is all or nothing. 48 KHz doesn’t do a better job that 44.1 KHz. The only difference is 48 KHz can reproduce everything below 24 KHz and 44.1 KHz can reproduce everything below 22.05 KHz. In both cases if your input signal remains below the lower threshold of 22.05 KHz the output of both sample rates will be identical. The 48 KHz one will not be smoother or more accurate. They’ll both be exactly the same. If there were signal content at 23 KHz then the 44.1 KHz sample rate would fail to reproduce it because it is beyond Nyquist, but the 48 KHz sample rate could still reproduce it. But if the input signal remained below the Nyquist point for the lower sample rate then both sample rates would be able to reproduce it and their output would both be the same. Exactly the same. Sample rate is not a quality indicator, there is no resolution. Sample rate only tells you the ceiling for the frequencies you can reproduce. And when you stay below that ceiling the quality does not change because it doesn’t work that way. Really, just go watch the video. You will learn something. It tells it in a much easier to understand fashion since it is a video and can go through the examples visually.

[@FilmmakerIQ]

I've already seen that video which is absolutely fantastic. I'm not disagreeing with anything you're saying and if YOU ACTUALLY WATCHED MY VIDEO instead of just indulging your preconceived notion of what I'm saying, then you wouldn't have needed to write that wall of text. Your problem is you have some qualitative notion of what "resolution" means. Yes, audio sampling has RESOLUTION - the waveform IS THE GRAPH of that resolution. You can't deny that a 48khz has more resolution than 44.1khz - it's by DEFINITION - sampled about 10% more times per second - which according to Nyquist means it can handle about 2khz more frequency on the high end. Everything else - from 0-22khz - yes exactly the same. Do we hear the difference, absolutely not. But then you get into filtering and anti-aliasing and that's a larger story. It's the same thing with image resolution. The human eye can only take in so much detail, you can increase image resolution further and further and it's not going to look any better - no one can tell a 4K image from an 8K image by looking at it on their phone...

[@ClaytonMacleod]

@@FilmmakerIQ The point is the word resolution implies that a 1 KHz output signal is “better” with the 48 KHz sample rate. Do you think the 48 KHz gives a better, higher resolution output signal of that 1 KHz signal? The notion of “better resolution” would mean just that. It would mean that the 48 KHz sampled version of the 1 KHz signal would contain more information, would have better resolution, than the 44.1 KHz sample rate version. Is this the view you hold? Because that’s what better resolution would mean. If that’s not the view that you hold then the phrase “better resolution” is not actually what you are trying to say. You saying that it is sampled about 10% more times per second leads me to believe that this is indeed the view you hold, that the 48 KHz sample rate would give a “better” 1 KHz, as that’s what better resolution would mean. Unfortunately this is not what’s happening, as I’ve said several times now. Sampling it by about 10% more times per second doesn’t give you a better, more accurate output, with higher resolution. You could sample a 1 KHz input signal at 3 KHz or a million Hz and the resulting output would be exactly the same in both cases. The million Hz sample rate version would again not be better. Why? Again, because there is no resolution. You either get a perfect representation or you get an error. There is no better or worse quality. It works, or fails. And if you wish to argue against that I am sorry to say it is simply because you do not yet understand how it all works, even if you think you do. Have a good one.

[@FilmmakerIQ]

How dense are you??? I just got done explaining your problem is you equate resolution to quality like a grade schooler. MORE RESOLUTION DOES NOT EQUAL MORE QUALITY. It doesn't imply it either unless you take a sophomoric look at the world. Are you denying that 48khz has 10% more samples per second? Are you denying that 48khz gives us about 2khz more headroom than 44.1? Did you not read and hear me say no one can tell the difference between 44.1 and 48?? Seriously... Maybe YOU need to rewatch Monty's video and lose your bias.

[@mattryan6886]

This was very interesting to watch and very informative. Definitely going to subscribe and check out the rest of your videos. I’m interested in learning how to edit video, and learning basic film making technique/technology. Just got a Mac mini and a friend of mine is giving me a version of final cut software for me to use and start on

[@neville_bengali]

Beautifully explained

[@niczoom]

Really interesting!!! Thanks for the info.

[@ElectronicazMusic]

Great vid. Fascinating! Thanks!

[@KeefJudge]

Thank you for the explanation.

[@davidcottrell1308]

Frickin' Brilliant! Thanks that is VERY INTERESTING!!!!

[@abrahamwondafrash7549]

priceless audio lesson....thanks man!

[@bingbong7316]

f=2n babaaay! That was a good demystifying presentation, even though you sometimes said Hz when you meant to say kHz, but that's quite forgivable. Liked.

[@shaihulud4515]

Just a little addon: babies are supposed to have a hearing range up to 20kHz - most people lose a lot of range in the first six to ten years (where the range drops to about 18 kHz). Most people from their thirties onwards barely reach 16kHz - most hearing tests for adults only cover a range up to 8kHz. So you're good with 44,1kHz sampling rate - but for reasons beyond the scope of my comment, you should aim at 24bit resolution then. I do my recordings in my studio at 48kHz/24 bit, which is quiet common these days, though many may argue you could use much higher sample rates. But that's another story.

[@FilmmakerIQ]

I agree whole heatedly with 24 but for recording at least

[@shaihulud4515]

@@FilmmakerIQ Absolutely - but the explanation would be worth a video on it's own :) Anyway: I really enjoyed this video, and the way it was presented - I felt free to subscribe :-D

[@JagadishTalluri]

Wonderful. Precise. Power Packed.

[@deemrcee]

Wow as sound tech I wonder why this video was not made earlier in time. So important.

[@timespace.productions7513]

Be safe & healthy, Mr. Hess.