@audiohaze The reason 44.1kHz was picked is historical and an artifact of PCM adapters that were developed at the time. At the time the PCM adapters found in video recorders (tape) were somewhat crude but functional. Different manufacturers fought to establish this standard but ultimately PCM adapters that functioned in the following way were used. The PCM adapter worked by taking 3 samples per line for 245 lines @ 60Hz. So 3 x 245 x 60 = 44,100 = 44.1 khZ. This is also handy of course because it would easily allow for the same PCM adapter to be used in an application utilizing 44.1 audio and 60hz video at the time.
So in essence there is no good technical reason -- just an artifact of engineering at the time which was implemented in that way to keep the circuitry small. And of course all engineering is a trade off with complexity. The only requirement is that it is more than 2x 20khz.
@@AfferbeckBeats In 50Hz video there are 37 lines of blanking, leaving 588 active lines per frame, which is 294 per field. So in a similar calculation you get: 50 x 294 x 3 = 44,100 Hz = 44.1 kHz
Higher sample rates are useful for processing. Many plugins upsample the signal before processing because it delivers a better result. This is especially used to combat aliasing that occurs when harmonic distortion is introduced. So it can make sense to record in higher sample rates than your delivery format. Also, higher bitrates does not really give you more headroom in that sense (well 32bit float somewhat does). It only gives you more dynamic levels. But that will allow you to lower the input signal without raising the noise floor, thus giving you more headroom before clipping. If you record at 16 bits it's wise to keep the input level as high as possible without clipping to get the cleanest signal. 24 bits allows you to back it off a bit and still get a clean signal. 32 bit float is a bit special as it goes beyond 0. You can still in theory clip the signal, but in practice you won't. 32bit float converters are actually multiple ADCs in parallell with individual gain circuits. Pretty much the same way as HDR photography works. It samples the audio at different gain levels to deliver a signal with an extremely high dynamic range that is practically impossible to clip.
Higher sample rates like 88.2kHz and 96kHz are also really common in film post-production and sound design as the higher resolution allows for more options as you process the audio, since you have more information in your recorded audio you can really stretch and warp your clips with significantly less artifacting
Random note on sample rates: higher sample rates have less latency at the cost of higher cpu usage. Also, though we can’t hear above 20khz, I believe the myth that higher rates sounds better is a consequence of less aliasing in analog plugin emulations. So it’s not necessarily that higher sample rates make it sound better, it makes your plugins operate better because there’s less aliasing. If you were mixing mostly with analog outboard and linear plugins, like some fabfilter plugins, there would be no difference sound wise between 44.1 and 96khz.
Agree. I sometimes find 96 kHz recordings of the same source less satisfying than 48 in many situations - probably because the computer is just struggling more. Call it computational distortion. And this is unnecessary, because the Nyquist Theorem doesn't suggest that twice the highest frequency you need to reproduce is "good enough", it states that nothing above that can offer any improvement. The above is so because we are counting waves as they come "at us" - head on, such as when we sit in a chair at the beach and count waves per minute. This count is frequency, pure and simple and all that matters is peaks and troughs. So, in terms of counting frequency, very high sample rates don't translate to a smoother picture. This is counter-intuitive, so it takes some wrapping of the head around. Sample rate is not part of trying to reproduce a wave as we picture it from "the side" - such as they appear on a scope. It's weird but, to reproduce a sound, perfectly, all we need to measure is frequency of peaks over time and amplitude over time. BTW, Ricky, IHMO the RE 20 continues to kill on your voice. Still my fave,
@@gr500music6 yeah, no, not what I was saying at all. Functionally this is no difference in sound quality. This is a bit to deep. A computer struggling (which btw doesn’t cause distortion unless it starts to screech and tear which means you are killing your CPU with plugins typically then) doesn’t create any type of distortion, audio or information wise. What I am saying is the aliasing is what people perceive as sounding bad, not the sample rate. You can’t hear above 20khz. period. EQ curves above 20 can be heard but that’s only because part of the curve are audible. Recording in 96khz will literally sound no different than 44.1khz. It’ll capture the audible frequencies and the non-linearities of your analog front end without aliasing because aliasing is produced when a plugin is emulating stuff above nyquist. So. No lol sample rate will not effect tracking and is much more a consideration in mixing and mastering because of analog emulations and maybe some linear eq cramping. You could make a small argument that interactions above 20khz are audible but I’ve never found those to make any significant impact on the sound.
Higher sample DO sound better with DSD (direct stream digital) recording for sure. It is a different method than the "CD quality" PCM method. It uses a sampling rate of about 2.8 million (2.8224 MHz) samples per second vs. 44.1k (44,100 sample per second) or 48k (48,000 spc). but it uses a 1 bit depth...not 16 bit or 24 bit. The sound is incredibly detailed. Listen to a well mastered SACD and compare it to a standard CD of the same music...or watch RUclips videos with a live recording that switches back and forth...even through compressed RUclips audio, the difference is incredible. Who cares about Nyquist and "people can't hear above 20k anyway"? It sounds better WITHIN the normal range of human hearing! (DSD naysayers have got to be functionally tone deaf...check it out for yourself!) "CD quality" is fine for the car and earbuds...but in a good listening environment...it's DSD for the win!
44.1kHz was chosen because it fitted well with scan rates etc used with ntsc television. They knew it had to be over 44kHz to get the filter design to work well and cheaply enough with existing methods. The extra 0.1 was to do with making it play nicely with television standards in the US.
Something with phase: If you have inverted phase in a stereo recording (One channel canceling the other channel) you will end up with a very strange sound (Basically it'll sound like the noise is coming from inside your head and it will sound completely different between speakers and headphones. Also singular speakers that mix both channels down into one, like smart speakers, won't handle it well)
I'm an electrical engineer so this was cool to listen to. Good job. Nyquist Theorem states that, by sampling a signal at a rate that is at least two times the maximum frequency or bandwidth B, the original analog signal can be re-created from the samples, however, to achieve this lower limit, an ideal, brick-wall lowpass filter would be required, which is not realizable, therefore, by sampling at a rate a little higher than 2B, we relax the lowpass re-construction filter requirement into something that is practically realizable. I once applied this theory to relationships and came up with "The Sampling Theorem for Chicks". Good times.
Good video. I think you should be looking at aliasing when you talk about sample rate. Really important with digital synths, digital saturation, digital clipping and digital compression. I guess the other thing is large pitch changes. 96khz will do all 4 of these things better than a low sample rate. Internal oversampling isn’t always an option. We can’t afford to always work at 96khz but in most situations it will sound better. Bit depth, 32 but is always desirable. Accidentally clip a performance, well not really in 32 bit, you can turn it down. Noise floor is lower again. Now playback on the other hand is a different story. For that 44.1 and 24bit is significantly greater quality than everything on Spotify, and more than adequate. But I think Phase and DB points were interesting and adequate. Good points for the majority of listeners.
@audiohaze Not boring at all, but I wanted to add that the reason for increasing the bit depth, and thereby the dynamic range is not really as you mentioned to increase the high end, but to reduce distortion of low level sounds.... since human perception of volume is a logarithmic scale (the reason for using dB of course), very quiet sounds are necessarily very low level, so while you have 16 bits at the top of the volume scale, it could be that you only have 10 or 12 bits for quiet sounds, which will increase distortion greatly... thus using 24 or 32 bit samples leaves more resolution on the table for quiet sounds. Also, increased sample rate, I believe does offer advantages for digital processing, leading to less distortion being created when pure digital effects and mixing are used.
When you recreate your signal from digital back to analog you get multiple specters of the original signal. And their mid freqs are spaced one from another by the whole the spectrum range of the original signal. So if you use only a double freq to sample your signal you will get these copies going back to back. Then you must filter out all copies and leave the first one, your recreated signal. BUT you cant filter perfectly. You will get aliasing in high freqs. That why you need some extra space and more higher sample rate.
love the anti-clickbait title... I did make it through to the end, although I was hoping to find out some ways in which misunderstandings of these concepts relates to mixes sounding bad
Making the imponderable ponderable! Really well done. Over the years (I'm 57) I've picked much of this up via trial and so much error. This will give people a head start if they want to delve deeper. I had to learn phase without the digital benefit of being able to move the waves with my bro's 2 track. Ugh!
I’m a musician, but my job is as a sound designer. Most film and TV shows are recorded at 24 or 23.976 frames per second, in the US. This pairs very nicely with 48kHz. The standard spec I am asked to provide sound for film and TV is 48kHz/24-Bit. I actually record most of my SFX at 96kHz/32-bit. This is because when using a lot of processing, the tools we use can more effectively manipulate the higher sample rate/bit depth then convert later. Basically, an explosion recorded at 96kHz/32-bit FP will sound better when processed and played back at 48kHz/32-bit FP than a signal originally recorded at 48kHz/32-bit FP. Human voices usually sound horrible at 96kHz. I would argue most musical instruments sound worse at 96kHz as well.
LOVE IT! But still, I was waiting for more mixing tips :P (I studied Audio Engineering in Vancouver 3 years ago but still fighting with my mixes... :( ) Great Channel btw :)
I think it's good to keep in mind that if you experiment with phase on purpose e.g. with bass using some stereo plugins which does such things, when you mono the signal like the music is played e.g. in clubs, it might cancel what you have done similarly how you showed, so bass suddenly disappears when played in mono.
Coming from film sound design perspective, we sometimes record sound (for effects use) in 96kHz. I've found that material recoded that way responds better to time stretching and pitch shifting than 48kHz (the standard). This is because of better resolution, like you mentioned. And I'm not an expert in this area, but I've heard that classical music tends to be recorded in 96kHz.
The phase thing I knew, it’s happened to me way too many times. I actually prefer using a single microphone when I can. It’s easier to get a full sound with that, than 5 microphones and you’re not sure about phase cancellation. And this wasn’t boring! The most boring music exercise Ive done is listen to white noise in frequency ranges, to train my ear to hear different zones of pitch. Very useful though, much like this video.
This presentation points out why I recommend that everyone learning recording start out with ONE microphone, LEARN how to record an entire group that way, then graduate to two mics. Most people these days start out with 50+ channels 'hot' on a DAW and wonder where the problems are. That's how I started learning in 1961, one mic, one mono tape machine, one basement and friends that could actually PLAY music. Best regards, Bill P. Studio 'A' nonlinear
Meh, I do understand phase. I’m at the phase where I like watching boring videos. Hah. He thinks he could punk me. The part I didn’t get were the sign waves, yield? Stop? Speed limit? I’ve never seen them wave at me. Oh, I’m also at the phase where I just tell bad jokes to waste time between meetings.
Hello, Nice presentation, let me clear up the dB haze for you. A decibel is the smallest CHANGE is sound amplitude that can be detected by the average human being in either direction (louder or softer). Example --- A signal raised to 78 dB from 77 dB or lowered to 76 dB from 77 dB is the minimum detectable change in loudness. Which explains why I crack up when these youtube cowboys talk about increasing or decreasing a filter response curve by 1/10 dB. By definition, the change is INAUDIBLE ! Best regards, Bill P.
I've been told that although you can't hear 96kHz, you can feel it. If you have the proper equipment, and all the proper systems, etc. But at the moment that's still quite an investment. I think monitors that can capture 96kHz is still in the multiple thousands. I personally don't think 96kHz will even catch on unless the hardware for it gets cheaper. (to my understanding, that also includes wires, monitors, systems, etc.)
Really good video. However I found bit-depth explanation a bit confusing. One way to look at bit-depth as you say is dynamic range - on that I agree. But I prefer to view bit-depth more as precision - with higher bit-depth you can measure smaller differences in signal amplitude. So, for example, when you have extremely low volume recording with 16-bit depth, after amplification (normalization) you will get a lot of loud digital noise and with 24-bit there will be much less of such noise. TLDR - higher bit-depth good for processing audio and bad for bigger size files.
To expand on the topic of kilohertz: if you are writing music for CD or the movie, it is better to use 48 kHz because when you need to convert to 44.1 kHz, the loss will be small (divide 48 by 44.1 and you will see the result), but if you are writing music in 44.1 kHz, it is better to stay at this frequency and not change it.
higher bit depth actually has no real impact on dynamic range, it just determines how loud your noise floor is, and 16 bits or more will get you a noise floor that's entirely inaudible unless you're playing it RIDICULOUSLY loud or are dithering multiple times (at 24 bit even that doesn't matter): technically higher bit depths do increase your headroom relative to the noise floor but that metric is basically useless
10 decibels (or 1 bel, actually) means a ratio of 10:1 in a value. When you double a value, that's the same as increasing it by 3.0103 dB. Doubling the amplitude of a sine wave *quadruples* its power, so its *power* goes up by 6.0206dB (it's doubled twice). So the dynamic range (in power) of 16-bit audio is 6.0206*16, or 96.3296dB.
I would be interested in seeing a video with this phase cancellation on a real track. I mean, I'm not sure I'm having those issues with Garageband, but it would be cool to see if I'm messing something up, and of couerse, an example of how to fix it. I'm a noob when it comes to mixing, and most of the time, I don't even know where to start.
What I've heard is that 88.2/96kHz and above benefits from moving any quantization noise above the hearing threshold. What my practical take is, higher frequencies (think, percussion/hi-hat) sound worse at 44.1 vs 96kHz. Not sure about 48kHz though (maybe it's good enough?).
One note about bit depth: It’s true that bit depth impacts dynamic range and that dynamic range is a product of bit depth, but the “24 bit” in “24 bit, 48 khz” isn’t talking about dynamic range as such. What bit depth actually means is how much data we have available to describe each of those 48,000 samples. I.e. each and every one of those samples contain 24 bits of information. Then to further confuse things, “8 bit music” doesn’t really have anything to do with bit depth. It’s just an aesthetic inspired by the sounds from 8 bit game consoles and computers like the Nintendo Entertainment System or Commodore 64. And the “8 bit” here refer to the capabilities of the CPU, and so we’re getting into computer science rather than audio engineering 😅 Those machines couldn’t really do sampled or recorded audio at all, the sounds they made were generated by synthesis in real time.
Thanks for the distinction! And yes regarding 8bit I was referring to the true 8bit compositions from retro consoles rather than the modern digital recreations :) absolutely 8bit music these days is just a style, but back in the day it wasn't!
Why 44.1kHz sampling rather than 40kHz? Because proper recording of digital audio requires using a low-pass filter to prevent *aliasing*. Without the filter, a (supersonic) 43.1kHz tone would end up sounding (audibly!) like a 1kHz tone. The supersonic content needs to be filtered out of the audio before it is sampled. Filters cause phase shift. The higher the frequency of the low-pass filter, the less it will affect the phase of the audible frequencies. So 44.1 is better than 40.
Not boring, and also, I didn't make it through because I already knew it. 😀 And no, that's not why my mixes suck at all. I think I don't really understand how to use compressors on a mix, and also, my room sucks so anything acoustic sounds like shit. Or perhaps it's my mics.
I would challenge that this may be in your head. That said it seems as though sample rate may have an effect on latency? Maybe this is what you're experiencing?
I think depth doesn't matter outside of mixing/mastering and SR is just for oversampling or whatever it is. Going from 320 to 1,411 does seem like quite a jump though.
there is a free plug; span by voxengo that u can check if its in phase or not. its on the bottom right. if u do smidge a lil, it does sound like you're "double tracking" (proper term is duplicating track) it, but it will be an issue when it is summed into mono. also it'll be +3db louder
I like to record voice over at 96khz so if I need to use a pitch shifter for voice effects I dont degrade the source too much when I need to slow it down but thats juste for voice effect post production.
I watched until the end hoping to find out why my mixes suck. I am an electronic engineer and I know all these concepts really well, but still, my mixes suck.
I've been using 88.2kHz sample rate lately for the higher resolution and because it is mathematically double 44.1. I hear very subtle differences between 88.2 and 44.1 but I have no idea if I am actually hearing it or if I am just telling myself there is a difference lol.
Thinking about the question in reverse. Could you detect 22k? 11k? What about the choices made for MP3 of a quick recording? When would 320m be used and when might 128m be just fine? Factors for considering the use of higher sampling: Quality of the source Quality of gear.. mic, cables, pres, etc Hearing ability (which changes over time and frequency) Disk space And with a bit of humor, it may depend on whether you wish to also “taste” the ice cream slurp… gotta say the new Raspberry 192k interface sounds delicious ;)
what happens if I record a vocal take, copy and paste it onto a new track, so there's two of the same take. pan one left and one right and then slightly nudge one take, it wont phase right? and saves me doing two takes?
Zero dBSPL is NOT no sound! Similarly, Zero degrees Fahrenheit ( or Celsius ) is not no temperature. Zero dBSPL is very quiet, near the threshold of human hearing (Your Mileage May Vary). There is a an Anechoic room with a background noise level of -20.6 dBA. Actual Silence would be minus Infinity dBSPL or dBA
Nice explanation, next time someone ask me to explain sample rate to them i'll just send this XD better to explain it thant me XD. That DB stuff is still confusing for me even after all this time though 😵💫🤣
nice video and thanks for that explanation! In case of which sample rate to choose I learned the following: Most video productions use a sample rate of 48khz and most audio productions like on a CD use 44,1khz. so it's not a question of quality but a question of for what format to produce. Does anyone know if this statement is still true or relevant?
Thanks for the video. Really interesting, wasn't boring at all.😊 P.S.: I use a Rode VideoMicro hooked into my Zoom H1n to record to Podcast episodes. The audio sounds 'loud' enough when I play it on my H1n but not loud enough when I open it in Audacity. Not sure why this happens. Sorry for being a bit off topic, but I would be grateful if you could suggest a solution. P.P.S: I do 'Amplify' the sound but still not very satisfied with the final result.
It may be a product of the volume levels on your zoom? If you're cranking the monitor volume to hear it louder on the zoom and not actually adjusting the input signal, then it may appear louder when its not :)
That's generally unrecommended, but you can capture the sound from maybe headphones jack of Zoom if it has one, using external interface or even line-out in your computer if it has one
What about flipping the phase? Like when it comes to drum recording, sometimes, when a snare is being recorded with a top and bottom mic, the bottom mic sometimes needs its phase flipped so low end and or body is lost. Anything notes on this compared to other instrument application?
HEY BUDDY. Well in this scenario, both the bottom and top snare are still technically two different distances from the sound source, and the top mic is likely much closer. So the reason the bottom gets flipped is to better approximate those "peaks and troughs" as they appear on the top mic. You could achieve the same effect by simply nudging the snare bottom mic's timeline a millisecond or two in the DAW to align them better rather than flipping the phase in an EQ or something :)
The most common time I'm checking and adjusting phase and definitely the easiest to see and hear is when I'm layering different kick drums in a DAW. When they're out of phase, they will be much quieter, and sound much thinner with far less low end. So you zoom in and drag one around until some of the bigger peaks and valleys match. Playing them again they will be much louder and have much fuller low end. Sometimes that actually sounds worse though and you might prefer the thinner sound, but you usually don't. Quickest and easiest way to check phase at least in Ableton but I'm sure other DAWs are similar is to throw the Phase Invert Utility effect on one of the tracks in question and compare the master level with it on and off.
It's definitely a skill you learn! You'll learn to hear when something lacks body and low end, eventually you'll start to recognize what phase issues sound like. Its a pretty characteristic sound. Maybe try taking a project you have and doing what I showed in this video? Take a signal recorded by two microphones, and nudge one signal slightly over and analyze the results. You'll start to pick up on that distinct tone.
@audiohaze The reason 44.1kHz was picked is historical and an artifact of PCM adapters that were developed at the time. At the time the PCM adapters found in video recorders (tape) were somewhat crude but functional. Different manufacturers fought to establish this standard but ultimately PCM adapters that functioned in the following way were used. The PCM adapter worked by taking 3 samples per line for 245 lines @ 60Hz. So 3 x 245 x 60 = 44,100 = 44.1 khZ. This is also handy of course because it would easily allow for the same PCM adapter to be used in an application utilizing 44.1 audio and 60hz video at the time.
So in essence there is no good technical reason -- just an artifact of engineering at the time which was implemented in that way to keep the circuitry small. And of course all engineering is a trade off with complexity. The only requirement is that it is more than 2x 20khz.
I wonder if 50hz regions were doing... 36.75khz?
@@AfferbeckBeats In 50Hz video there are 37 lines of blanking, leaving 588 active lines per frame, which is 294 per field. So in a similar calculation you get: 50 x 294 x 3 = 44,100 Hz = 44.1 kHz
Exactly! Started to write that answer but scolled down and saw yours. ☺
Amazing thank you so much for sharing!
I will go back to watching my paint dry now.
That's sound recording 101 and that's such a needed video for all including professionals to revise a bit
Totally agree :)
Higher sample rates are useful for processing. Many plugins upsample the signal before processing because it delivers a better result. This is especially used to combat aliasing that occurs when harmonic distortion is introduced.
So it can make sense to record in higher sample rates than your delivery format.
Also, higher bitrates does not really give you more headroom in that sense (well 32bit float somewhat does). It only gives you more dynamic levels. But that will allow you to lower the input signal without raising the noise floor, thus giving you more headroom before clipping. If you record at 16 bits it's wise to keep the input level as high as possible without clipping to get the cleanest signal. 24 bits allows you to back it off a bit and still get a clean signal.
32 bit float is a bit special as it goes beyond 0. You can still in theory clip the signal, but in practice you won't. 32bit float converters are actually multiple ADCs in parallell with individual gain circuits. Pretty much the same way as HDR photography works. It samples the audio at different gain levels to deliver a signal with an extremely high dynamic range that is practically impossible to clip.
Appreciate the info and distinction! Really good stuff thank you Henrik :)
Higher sample rates like 88.2kHz and 96kHz are also really common in film post-production and sound design as the higher resolution allows for more options as you process the audio, since you have more information in your recorded audio you can really stretch and warp your clips with significantly less artifacting
Your videos are amazing. I hope you continue to grow the audience you deserve
LOL @ 6:09 Cat is trying to open the door in the background. That's so cool ^____^
Random note on sample rates: higher sample rates have less latency at the cost of higher cpu usage. Also, though we can’t hear above 20khz, I believe the myth that higher rates sounds better is a consequence of less aliasing in analog plugin emulations. So it’s not necessarily that higher sample rates make it sound better, it makes your plugins operate better because there’s less aliasing. If you were mixing mostly with analog outboard and linear plugins, like some fabfilter plugins, there would be no difference sound wise between 44.1 and 96khz.
Love this! Thank you for expanding the discussion on sample rates my friend
Agree. I sometimes find 96 kHz recordings of the same source less satisfying than 48 in many situations - probably because the computer is just struggling more. Call it computational distortion. And this is unnecessary, because the Nyquist Theorem doesn't suggest that twice the highest frequency you need to reproduce is "good enough", it states that nothing above that can offer any improvement.
The above is so because we are counting waves as they come "at us" - head on, such as when we sit in a chair at the beach and count waves per minute. This count is frequency, pure and simple and all that matters is peaks and troughs. So, in terms of counting frequency, very high sample rates don't translate to a smoother picture. This is counter-intuitive, so it takes some wrapping of the head around. Sample rate is not part of trying to reproduce a wave as we picture it from "the side" - such as they appear on a scope. It's weird but, to reproduce a sound, perfectly, all we need to measure is frequency of peaks over time and amplitude over time.
BTW, Ricky, IHMO the RE 20 continues to kill on your voice. Still my fave,
@@gr500music6 yeah, no, not what I was saying at all. Functionally this is no difference in sound quality. This is a bit to deep. A computer struggling (which btw doesn’t cause distortion unless it starts to screech and tear which means you are killing your CPU with plugins typically then) doesn’t create any type of distortion, audio or information wise.
What I am saying is the aliasing is what people perceive as sounding bad, not the sample rate. You can’t hear above 20khz. period. EQ curves above 20 can be heard but that’s only because part of the curve are audible. Recording in 96khz will literally sound no different than 44.1khz. It’ll capture the audible frequencies and the non-linearities of your analog front end without aliasing because aliasing is produced when a plugin is emulating stuff above nyquist.
So. No lol sample rate will not effect tracking and is much more a consideration in mixing and mastering because of analog emulations and maybe some linear eq cramping. You could make a small argument that interactions above 20khz are audible but I’ve never found those to make any significant impact on the sound.
@@AJOrpheo Thanks for the reply!
Higher sample DO sound better with DSD (direct stream digital) recording for sure. It is a different method than the "CD quality" PCM method. It uses a sampling rate of about 2.8 million (2.8224 MHz) samples per second vs. 44.1k (44,100 sample per second) or 48k (48,000 spc). but it uses a 1 bit depth...not 16 bit or 24 bit. The sound is incredibly detailed. Listen to a well mastered SACD and compare it to a standard CD of the same music...or watch RUclips videos with a live recording that switches back and forth...even through compressed RUclips audio, the difference is incredible. Who cares about Nyquist and "people can't hear above 20k anyway"? It sounds better WITHIN the normal range of human hearing! (DSD naysayers have got to be functionally tone deaf...check it out for yourself!) "CD quality" is fine for the car and earbuds...but in a good listening environment...it's DSD for the win!
44.1kHz was chosen because it fitted well with scan rates etc used with ntsc television. They knew it had to be over 44kHz to get the filter design to work well and cheaply enough with existing methods. The extra 0.1 was to do with making it play nicely with television standards in the US.
Nice job Audio Haze! I made it through and my mixes dont suck anymore
I couldn’t focus on the topic when I saw how great solo this trumpet player at 9:13 is throwing at me. Nothing suspicious at all. 😂😂😂
Something with phase: If you have inverted phase in a stereo recording (One channel canceling the other channel) you will end up with a very strange sound (Basically it'll sound like the noise is coming from inside your head and it will sound completely different between speakers and headphones. Also singular speakers that mix both channels down into one, like smart speakers, won't handle it well)
I'm an electrical engineer so this was cool to listen to. Good job. Nyquist Theorem states that, by sampling a signal at a rate that is at least two times the maximum frequency or bandwidth B, the original analog signal can be re-created from the samples, however, to achieve this lower limit, an ideal, brick-wall lowpass filter would be required, which is not realizable, therefore, by sampling at a rate a little higher than 2B, we relax the lowpass re-construction filter requirement into something that is practically realizable. I once applied this theory to relationships and came up with "The Sampling Theorem for Chicks". Good times.
What a great video, really. Keep being one of the best music channels of yt.
Good video. I think you should be looking at aliasing when you talk about sample rate. Really important with digital synths, digital saturation, digital clipping and digital compression. I guess the other thing is large pitch changes. 96khz will do all 4 of these things better than a low sample rate. Internal oversampling isn’t always an option. We can’t afford to always work at 96khz but in most situations it will sound better.
Bit depth, 32 but is always desirable. Accidentally clip a performance, well not really in 32 bit, you can turn it down. Noise floor is lower again.
Now playback on the other hand is a different story. For that 44.1 and 24bit is significantly greater quality than everything on Spotify, and more than adequate.
But I think Phase and DB points were interesting and adequate. Good points for the majority of listeners.
@audiohaze Not boring at all, but I wanted to add that the reason for increasing the bit depth, and thereby the dynamic range is not really as you mentioned to increase the high end, but to reduce distortion of low level sounds.... since human perception of volume is a logarithmic scale (the reason for using dB of course), very quiet sounds are necessarily very low level, so while you have 16 bits at the top of the volume scale, it could be that you only have 10 or 12 bits for quiet sounds, which will increase distortion greatly... thus using 24 or 32 bit samples leaves more resolution on the table for quiet sounds.
Also, increased sample rate, I believe does offer advantages for digital processing, leading to less distortion being created when pure digital effects and mixing are used.
When you recreate your signal from digital back to analog you get multiple specters of the original signal. And their mid freqs are spaced one from another by the whole the spectrum range of the original signal. So if you use only a double freq to sample your signal you will get these copies going back to back. Then you must filter out all copies and leave the first one, your recreated signal. BUT you cant filter perfectly. You will get aliasing in high freqs. That why you need some extra space and more higher sample rate.
love the anti-clickbait title... I did make it through to the end, although I was hoping to find out some ways in which misunderstandings of these concepts relates to mixes sounding bad
Making the imponderable ponderable! Really well done.
Over the years (I'm 57) I've picked much of this up via trial and so much error. This will give people a head start if they want to delve deeper.
I had to learn phase without the digital benefit of being able to move the waves with my bro's 2 track. Ugh!
I’m a musician, but my job is as a sound designer. Most film and TV shows are recorded at 24 or 23.976 frames per second, in the US. This pairs very nicely with 48kHz. The standard spec I am asked to provide sound for film and TV is 48kHz/24-Bit. I actually record most of my SFX at 96kHz/32-bit. This is because when using a lot of processing, the tools we use can more effectively manipulate the higher sample rate/bit depth then convert later. Basically, an explosion recorded at 96kHz/32-bit FP will sound better when processed and played back at 48kHz/32-bit FP than a signal originally recorded at 48kHz/32-bit FP. Human voices usually sound horrible at 96kHz. I would argue most musical instruments sound worse at 96kHz as well.
LOVE IT! But still, I was waiting for more mixing tips :P (I studied Audio Engineering in Vancouver 3 years ago but still fighting with my mixes... :( ) Great Channel btw :)
I think it's good to keep in mind that if you experiment with phase on purpose e.g. with bass using some stereo plugins which does such things, when you mono the signal like the music is played e.g. in clubs, it might cancel what you have done similarly how you showed, so bass suddenly disappears when played in mono.
Coming from film sound design perspective, we sometimes record sound (for effects use) in 96kHz. I've found that material recoded that way responds better to time stretching and pitch shifting than 48kHz (the standard). This is because of better resolution, like you mentioned. And I'm not an expert in this area, but I've heard that classical music tends to be recorded in 96kHz.
The phase thing I knew, it’s happened to me way too many times. I actually prefer using a single microphone when I can. It’s easier to get a full sound with that, than 5 microphones and you’re not sure about phase cancellation. And this wasn’t boring! The most boring music exercise Ive done is listen to white noise in frequency ranges, to train my ear to hear different zones of pitch. Very useful though, much like this video.
This presentation points out why I recommend that everyone learning recording start out with ONE microphone, LEARN how to record an entire group that way, then graduate to two mics.
Most people these days start out with 50+ channels 'hot' on a DAW and wonder where the problems are.
That's how I started learning in 1961, one mic, one mono tape machine, one basement and friends that could actually PLAY music.
Best regards,
Bill P.
Studio 'A' nonlinear
Love your videos dude, always helpful, thought provoking, and inspiring
Meh, I do understand phase. I’m at the phase where I like watching boring videos. Hah. He thinks he could punk me. The part I didn’t get were the sign waves, yield? Stop? Speed limit? I’ve never seen them wave at me.
Oh, I’m also at the phase where I just tell bad jokes to waste time between meetings.
I like this phase. This is a good phase.
Thank you for posting mate 🤙
your videos are very calming to me for some reason i can’t put my finger on it tho
Oh oh! I also wish you talked about the buffer size. Even if it wasn't super relevant, touching on it woulda been interesting.
I like your channel man, really informative. Have a good day!
Hello,
Nice presentation, let me clear up the dB haze for you.
A decibel is the smallest CHANGE is sound amplitude that can be detected by the average human being in either direction (louder or softer).
Example --- A signal raised to 78 dB from 77 dB or lowered to 76 dB from 77 dB is the minimum detectable change in loudness.
Which explains why I crack up when these youtube cowboys talk about increasing or decreasing a filter response curve by 1/10 dB.
By definition, the change is INAUDIBLE !
Best regards,
Bill P.
Great videooo!!! Super interesting to hear about and i love learning about audio
Thanks!
96k was a hot debate when I was shopping for my first interface. In 2002.
I've been told that although you can't hear 96kHz, you can feel it. If you have the proper equipment, and all the proper systems, etc. But at the moment that's still quite an investment. I think monitors that can capture 96kHz is still in the multiple thousands.
I personally don't think 96kHz will even catch on unless the hardware for it gets cheaper. (to my understanding, that also includes wires, monitors, systems, etc.)
6:10 cat want to leave,i stay tuned
The term decibel has always been a bit confusing to me. Thanks for clarifying!
Really good video. However I found bit-depth explanation a bit confusing. One way to look at bit-depth as you say is dynamic range - on that I agree. But I prefer to view bit-depth more as precision - with higher bit-depth you can measure smaller differences in signal amplitude. So, for example, when you have extremely low volume recording with 16-bit depth, after amplification (normalization) you will get a lot of loud digital noise and with 24-bit there will be much less of such noise. TLDR - higher bit-depth good for processing audio and bad for bigger size files.
Oh shit 2:00 This is good info for making ASMR audio mixes as well.
this is something i needed to watch. boring eh idk i quite enjoyed it. explained systematically clear. thank you.
Glad you liked it!
I never noticed that MS-20 back there, I'd love to see a video with fhat!
One of the best vids on youtube ❤️(subbed)
Jeez dude. Your videos are so good.
Ey thanks dude!! Love your vids as well actually :) I’m a sub!
0dB spl isn't really no sound. 0dB is the limit of our hearing at 1kHz, some people can even hear a good bit into the negatives at around 3-5kHz
When you record to vinyl you can't call no further than ten thousand kilohertz. So that should be the golden number
To expand on the topic of kilohertz: if you are writing music for CD or the movie, it is better to use 48 kHz because when you need to convert to 44.1 kHz, the loss will be small (divide 48 by 44.1 and you will see the result), but if you are writing music in 44.1 kHz, it is better to stay at this frequency and not change it.
This was super interesting. Very well explained as well. Thanks!
Glad I could help thanks Ethan!
higher bit depth actually has no real impact on dynamic range, it just determines how loud your noise floor is, and 16 bits or more will get you a noise floor that's entirely inaudible unless you're playing it RIDICULOUSLY loud or are dithering multiple times (at 24 bit even that doesn't matter): technically higher bit depths do increase your headroom relative to the noise floor but that metric is basically useless
10 decibels (or 1 bel, actually) means a ratio of 10:1 in a value. When you double a value, that's the same as increasing it by 3.0103 dB. Doubling the amplitude of a sine wave *quadruples* its power, so its *power* goes up by 6.0206dB (it's doubled twice). So the dynamic range (in power) of 16-bit audio is 6.0206*16, or 96.3296dB.
Comming into mixing from math, I press like just for mentioning Shannon-Nyquist-Kotelnikov theorem :))
I would be interested in seeing a video with this phase cancellation on a real track. I mean, I'm not sure I'm having those issues with Garageband, but it would be cool to see if I'm messing something up, and of couerse, an example of how to fix it. I'm a noob when it comes to mixing, and most of the time, I don't even know where to start.
What I've heard is that 88.2/96kHz and above benefits from moving any quantization noise above the hearing threshold. What my practical take is, higher frequencies (think, percussion/hi-hat) sound worse at 44.1 vs 96kHz. Not sure about 48kHz though (maybe it's good enough?).
Fantastic video!!
Making me really happy i record 1channel directly from my keyboard for all my music lol
How many times does it take to remember the Phase lesson?
Phase CANNIBALIZATION!
Perfect! That phase phrase might stick. Thanks!
One note about bit depth: It’s true that bit depth impacts dynamic range and that dynamic range is a product of bit depth, but the “24 bit” in “24 bit, 48 khz” isn’t talking about dynamic range as such. What bit depth actually means is how much data we have available to describe each of those 48,000 samples. I.e. each and every one of those samples contain 24 bits of information.
Then to further confuse things, “8 bit music” doesn’t really have anything to do with bit depth. It’s just an aesthetic inspired by the sounds from 8 bit game consoles and computers like the Nintendo Entertainment System or Commodore 64. And the “8 bit” here refer to the capabilities of the CPU, and so we’re getting into computer science rather than audio engineering 😅 Those machines couldn’t really do sampled or recorded audio at all, the sounds they made were generated by synthesis in real time.
Thanks for the distinction! And yes regarding 8bit I was referring to the true 8bit compositions from retro consoles rather than the modern digital recreations :) absolutely 8bit music these days is just a style, but back in the day it wasn't!
Thanks, Billy. Thanks, I was about to add a comment like this, but you did a better job!
Its amazing how he used reverse psychology to make us actually want to watch the whole video despite how boring it is
It all makes sense now
SUPERB MATERIAL! Thank you!.
Glad I could help!!
You're kinda the Rhett Shull of audio and I like it
Love that thanks :)
I left the khz crazy talk long time ago and settled on 44.1 forever.
Bit depth seems far more important. I leave it at 24
I mean yes, these are super boring reasons.... but the Nyquist Theorem excites me deep down in my basal ganglia
It is a good theorem :))
Pretty cool. One question; do you use VU meters?
You tricked us all to watch this😅 loving the vids
Why 44.1kHz sampling rather than 40kHz? Because proper recording of digital audio requires using a low-pass filter to prevent *aliasing*. Without the filter, a (supersonic) 43.1kHz tone would end up sounding (audibly!) like a 1kHz tone. The supersonic content needs to be filtered out of the audio before it is sampled.
Filters cause phase shift. The higher the frequency of the low-pass filter, the less it will affect the phase of the audible frequencies. So 44.1 is better than 40.
Wait bro, I don't recall but do you have a video on how to record vocals?
Not boring, very interesting
not an audiophile nor can afford to be one but i enjoyed that somehow...maybe the cat contributed a bit, thanks, thumb up
Not boring, and also, I didn't make it through because I already knew it. 😀 And no, that's not why my mixes suck at all. I think I don't really understand how to use compressors on a mix, and also, my room sucks so anything acoustic sounds like shit. Or perhaps it's my mics.
Brilliant video - love the humour (apologies humor....probably 🙂)
Thanks dude!
I play guitar, electric one. So when I use 44.1 rate I hear some dirt in high-end, It goes away only using 96
I would challenge that this may be in your head. That said it seems as though sample rate may have an effect on latency? Maybe this is what you're experiencing?
I think depth doesn't matter outside of mixing/mastering and SR is just for oversampling or whatever it is. Going from 320 to 1,411 does seem like quite a jump though.
Dumb question about phase issues: how do you actually check it? Nudge one recording a smidge and listen to hear if it gets louder or softer?
there is a free plug; span by voxengo that u can check if its in phase or not. its on the bottom right. if u do smidge a lil, it does sound like you're "double tracking" (proper term is duplicating track) it, but it will be an issue when it is summed into mono. also it'll be +3db louder
I like to record voice over at 96khz so if I need to use a pitch shifter for voice effects I dont degrade the source too much when I need to slow it down but thats juste for voice effect post production.
The title made me think “You’re so bored, you probably think this videos about you”
I watched until the end hoping to find out why my mixes suck. I am an electronic engineer and I know all these concepts really well, but still, my mixes suck.
I’m sorry man :( probably not as bad as you think though! We’re our own greatest critic
AUDIOHAZE, Would you buy a Shure MV7X or a Shure SM57 as a "1 mic for everything" type of deal?
SM57 for me!
thanks for this
I've been using 88.2kHz sample rate lately for the higher resolution and because it is mathematically double 44.1. I hear very subtle differences between 88.2 and 44.1 but I have no idea if I am actually hearing it or if I am just telling myself there is a difference lol.
I would challenge that you're hearing distinct differences but who knows!
Thinking about the question in reverse. Could you detect 22k? 11k? What about the choices made for MP3 of a quick recording? When would 320m be used and when might 128m be just fine?
Factors for considering the use of higher sampling:
Quality of the source
Quality of gear.. mic, cables, pres, etc
Hearing ability (which changes over time and frequency)
Disk space
And with a bit of humor, it may depend on whether you wish to also “taste” the ice cream slurp… gotta say the new Raspberry 192k interface sounds delicious ;)
Wait, why wouldn't moving the other mic farther away fix the phase?
It won’t necessarily! Check phase before hand, if the mics are out of phase, adjust from there and listen to the signals again and repeat :)
not me sampling at 96 kHz for the birds and nature to listen to my music loollll
what happens if I record a vocal take, copy and paste it onto a new track, so there's two of the same take. pan one left and one right and then slightly nudge one take, it wont phase right? and saves me doing two takes?
Zero dBSPL is NOT no sound! Similarly, Zero degrees Fahrenheit ( or Celsius ) is not no temperature. Zero dBSPL is very quiet, near the threshold of human hearing (Your Mileage May Vary). There is a an Anechoic room with a background noise level of -20.6 dBA. Actual Silence would be minus Infinity dBSPL or dBA
I’m fw your channel
Thanks dude :)
Nice explanation, next time someone ask me to explain sample rate to them i'll just send this XD better to explain it thant me XD. That DB stuff is still confusing for me even after all this time though 😵💫🤣
Hahaha well thanks for sending it to them! Yeah decibels will never not be confusing
You underestimate my power!
HAHAH
Bingo! 2nd video I watched of your channel. Im returning, often. Oh I subd
How do we get this video to as many people as possible?!
Liking and subscribing I guess?? lol
Not boring at all!
nice video and thanks for that explanation! In case of which sample rate to choose I learned the following: Most video productions use a sample rate of 48khz and most audio productions like on a CD use 44,1khz. so it's not a question of quality but a question of for what format to produce. Does anyone know if this statement is still true or relevant?
Never heard this one! I do know CDs are 44.1kHz but I've never really known the reason?
@@AudioHaze I heard these were arbitrary choices for standards like you said in your video.
Fell asleep. Still didn't learn anything.
Could you do this with a cat or other cute furry animal?
Hahahaha will make a mental note for next time
Thanks for the video. Really interesting, wasn't boring at all.😊
P.S.: I use a Rode VideoMicro hooked into my Zoom H1n to record to Podcast episodes. The audio sounds 'loud' enough when I play it on my H1n but not loud enough when I open it in Audacity. Not sure why this happens. Sorry for being a bit off topic, but I would be grateful if you could suggest a solution.
P.P.S: I do 'Amplify' the sound but still not very satisfied with the final result.
It may be a product of the volume levels on your zoom? If you're cranking the monitor volume to hear it louder on the zoom and not actually adjusting the input signal, then it may appear louder when its not :)
@@AudioHaze Thanks for the reply. Yes, I thought so too. The volume level on my Zoom H1n is 80.
That's generally unrecommended, but you can capture the sound from maybe headphones jack of Zoom if it has one, using external interface or even line-out in your computer if it has one
@@---pp7tq Thanks for the reply.
Very helpful
thanks
What about flipping the phase? Like when it comes to drum recording, sometimes, when a snare is being recorded with a top and bottom mic, the bottom mic sometimes needs its phase flipped so low end and or body is lost. Anything notes on this compared to other instrument application?
HEY BUDDY. Well in this scenario, both the bottom and top snare are still technically two different distances from the sound source, and the top mic is likely much closer. So the reason the bottom gets flipped is to better approximate those "peaks and troughs" as they appear on the top mic. You could achieve the same effect by simply nudging the snare bottom mic's timeline a millisecond or two in the DAW to align them better rather than flipping the phase in an EQ or something :)
@@AudioHaze ah! This was super helpful, thanks!
DBW sounds kinda freaky ngl 😳
No, this was not boring!
What I did not understand is wow you check Phase and how you adjust? Probably I am just too stupid to get it.
how
The most common time I'm checking and adjusting phase and definitely the easiest to see and hear is when I'm layering different kick drums in a DAW. When they're out of phase, they will be much quieter, and sound much thinner with far less low end. So you zoom in and drag one around until some of the bigger peaks and valleys match. Playing them again they will be much louder and have much fuller low end. Sometimes that actually sounds worse though and you might prefer the thinner sound, but you usually don't.
Quickest and easiest way to check phase at least in Ableton but I'm sure other DAWs are similar is to throw the Phase Invert Utility effect on one of the tracks in question and compare the master level with it on and off.
It's definitely a skill you learn! You'll learn to hear when something lacks body and low end, eventually you'll start to recognize what phase issues sound like. Its a pretty characteristic sound. Maybe try taking a project you have and doing what I showed in this video? Take a signal recorded by two microphones, and nudge one signal slightly over and analyze the results. You'll start to pick up on that distinct tone.
Nice synth
Thanks man!
you were right i’m sorry
HAHA well hey thanks for trying lol
Depth is NOT about clipping.