3:00 acoustician here - SPL isn't normally expressed as 'dB SPL' - the correct index for sound power (LWA) or in this case, sound pressure (LPA) - however in environmental acoustics you'll usually use the LAeq,T index or the LAmax index to express the average and the maxima sound pressure levels. It's important to note that for sound pressure, a decibel is simply a ratio of the measured Pascals against the reference Pascal value. So for airborne sound, it's normally 20*LOG(Measured Pa/2*10^-5 Pa). It's also important to note that you can take sound pressure and sound power and use a lightbulb as an analogy. The sound power is like the watts of power the lightbulb is using - this is a constant and is the OUTPUT of the light (speaker/source) itself. The sound pressure is a relative point away from that source. It's important to also know that sources can be area, planar, line or point sources. And to add, weighting is very important too! A (for environmental acoustics, this emulates the effective weighting of the human ear) - C (similar to A, but allows room for lower frequencies, this is normally great for impulsive sounds or for music-related environmental sound) and then Z (this is just the unweighted sound). This is a great video, something I wish I had when I first started out!!
@@jrgroberts A weighting makes sure that single value broadband sound levels are most reflective of how humans actually perceive sound across the frequency spectrum. Look up the fletcher munson curve.
@@producermathew The Fletcher-Munson curve was created +90 years ago, when measurement equipment was nothing to par compared to what we use nowadays. Even if it has been was further revisions of the Fletcher-Munson curves, the A weighting curve is a scam. That helps politicians to avoid legal issues of millions of people contamined by out of the scope noise due to LAeq filters. It makes sense to revisit these concepts as @jrgroberts suggests, because the incremental low freq noise environments we suffer in the modern world.
@@PereRevert the A weighting curve is not inversely proportional to the fletcher munson curve. The A weighting correction is standardised within IEC 61672-1, ISO 226 and ANSI S1.4. A-weighting is generally considered representative of a typical curve for human hearing in so much that we do not perceive frequencies in the same way as an ultra flat type 1 measurement system. You say it’s a scam, yet don’t provide any context. This is false.
Your “low effort videos” are the ones I appreciate the most! No distracting background music, just your voice transporting expertise. It’s like technical ASMR 😄
I believe I'm speaking for the majority of us that genuinely appreciate what you're doing here, and we don't demand "high-effort videos", even though it feels like a out of time Christmas present to see you in my feed, what we most cherish is You, and your well being first, and then everything else, Dan! I hope you're re doing well and you take care of yourself.
I appreciate these no-fluff, erudite videos, that aren't padded out to 10:00 minutes just for the UA-cam algo. Respect to you Dan for your willingness to share your knowledge without wasting our time.
This. And it's actually a mark of smart people to respect their audience - they realise it's what's going to being them back instead of making them hesitate over whether a new video is another diluted waste-of-time video.
I understood the db are logarithmic, and that -6db halved the signal, but this was a great dive into the topic and covered some gaps in my knowledge. And like you said, no shame to the original commenter; you don't know what you don't know, and we're all here to learn.
32 bits fp are very, very beneficial for processing and design. The video does not state that 20/24 bits is end all be all, at all, and the "infinite dynamic range" is a very cheeky statement bordering on a lie. It is like saying that a range of numbers from 0 to 65000 is infinite. _Technically_, it is, but you still can't write 65001 to this range.
Absolutely 32/64 bit are necessary for dsp processing. But the idea that 24 bit and above are necessary for a/d and d/a is always a little tenuous. I just think that there is too much emphasis on tech sometimes and less talk about how music makes us feel. ✌️😊
The fact that you can explain such complex topic with just few sentences and make it understandable is the best proof you truly understand the context. Still ones of the best videos about audio engineering in the whole internet. Though, please don't stop making them higher effort videos even when these are absolutely brilliant too. Just the fact that you make banger music is a reason enough to watch them, but they bring so much to thw community.
Your "low-effort" videos are still amongst the most informative out there. Hope you're keeping well and keep up the great work (as long as your health/energy allows for it)!
2:55 0dBSPL is referenced to a pressure, which is 20 µPa. This is the "quietest sound" we can generally hear, as you correctly point out, but this refers to a 1KHz sine wave. Since our hearing is a bit more sensitive in the region from 2-4KHz, it's actually possible to hear negative dBSPL figures! Would've also loved to hear an explanation about what dBFS actually means (the reference to the digital maximum level), but it's a nice video overall :-) Looking forward to the next one!
From what I understood, full scale is just when you reach the integer limit when representing amplitude. However, the exact value depends on how the system you’re using works (I’m not sure if there are any standards). Also, DAWs are not clipping past full scale because they’re using floating point numbers instead of integers, which can store bigger numbers than integers.
Full scale integer is the largest number you can represent for positive cycles, or zero for negative cycles. Floating point audio represents full scale as 1.0 or -1.0. It used to mean the maximum amplitude that you could never exceed; now, with floating point processing, it means the maximum level that you're going to have to come back down to before releasing it.
Well, dBFS can be more complicated than that, if you read the wikipedia article on it, there are multiple conventions, and when following the AES convention (among others), a full scale square wave actually has +3dBFS. I work as a software engineer in an audio processing software company, and most of us use the other convention where we mean it to be the ratio to the maximum amplitude (of the (true or not) peak value of some signal). Though a DSP engineer once pointed me to the other convention (of which he is convinced is the right one, as it is defined by multiple standardization entities as such) and since I'm quite careful how I use the term, I try to avoid it if I can and always feel a bit conflicted :D. I guess it depends on the context which convention is more useful.
@@frydac This has to do with visual adjustment of RMS levels. A full scale square wave will have a level of 0dBpk and 0dBrms, which computes to a crest factor of 0. Many plugins&softwares adjust RMS levels to show a 3dB higher value, to reference to the crest factor of a clean sine wave, in which case you'll see 0dBpk and 0dBrms for a sine wave and 0dBpk and +3dBrms for a square wave as you point out, but simply due to this visual adjustment. You can however get theoretical peak values above 0dBFS, through so-called inter-sample peaks. Any fully scale sqaure wave will produce such inter sample peaks due to our processing being bandwidth limited, which makes an inifnitely short rise time impossible. As far as I know, modern DACs have a bit of headroom in the DA conversion to account for inter-sample peaks, but also modern mastering software like Ozone has inter-sample peak detection ("True peak"), which then makes sure this "digitally invisible" clipping can't happen.
By the way, avoiding intersample peaks is sort of an overzealous measure. Basically all DACs these days either have sufficient headroom not to produce audible clipping when converted to analogue, or it doesn't matter because they are crap anyway. 😂
First and foremost, your health is top priority. Take care of yourself the best way you can. That will not only benefit you, but all of us as well. 🙂 Wishing you the best of luck and quick relief. Secondly, it's not often somebody with extensive knowledge and experience takes the time to share it. Finding someone to share it with such integrity, morals and straight forward common sense is even harder to find. Then you do all that in a pleasing, humorous, respectful way. This video is a prime example, by not calling out the person asking the question and not being degrading in the response. You sir, have a gift with your presentation, thought process, and your voice. When I say "voice" I'm referring to your speaking voice, not your singing voice. That's a whole other subject. 😃 Had to leave you with a smile. Thank you for all you do and just being you.
Well, now I know. I never did before. 45 years working audio-adjacent and it never clicked, but after 5 minutes of a “low effort” video from Dan and it did. Damn fine job!
Once again something I think i sort of knew explained so clearly and concisely that I could probably explain it to someone else (except I'll probably forget)...nice one as ever. Cheers!
What do you mean higher efford? This is a complex topic explained shortly and very well. It does not need to be an hour long, or have amazing graphics, if it can be explained simply in 5 minutes. As always, thank you for making something complex so simple and understandable!
This video triggered my will to revisit my knowledge (or lack thereof) about this measure. And I learned how summing two uncorrelated sources result in terms of gain in dB, which is not so straightforward: summing 2 uncorrelated signal of the same level results in a 3 dB gain, while summing signals 6 dB apart (i.e. one is twice lower) results on a gain of 1 dB. Interesting stuff. Thank you Mr Worrall.
This is an interesting topic! What if I have 40 tracks but I’d like to lower my out by 1-2 db, how much do I need to lower the others?? Not expecting an answer 😅😅 but definitely something I’m going to research. I’d love Dan’s take on it 😁
Just a precaution.... How do you consider 2 unrelated signal as "the same level"? Basically what did you use to measure them? The loudness of 2 unrelated signal will be heavily dependent on how they overlap, so I wonder how useful it is to consider how louder they can become if added together?
Yep, this vid sent me back to the dB pages and calculators! I always like to remind myself that however loud one instrument is, adding a second one makes things 3dB louder. So when I go to eq something, like the high end of guitar strumming, and boost by 3 dB that’s like adding a second guitarist (at that frequency range)
"Ce qui se conçoit bien s'énonce clairement, et les mots pour le dire arrivent aisément". Mr Worrall, You're the clearest, yet most concise, on any audio related topic you discuss. And the metaphores are just on point. Perfect explaination
You've already put in the effort with the entirety of your career and knowledge. It may be basic and low effort to you now but it's vast to those who didn't know, and a refresher for those who did know. Hope everything is going good and thanks for all the videos
Well done again, Dan. Clear and to the point. Also, for posterity for up and coming engineers: LUFS has *no* bearing on how loud a record can be cut. Make it sound good. Short enough side lengths. Nothing sibilant or clipping.
Always very appreciative of your videos, regardless of whether they're shorter and simpler or longer and more in depth. There's always very well articulated and helpful information to be found and I'm super grateful for it!
Something related to this I have always got some amusement from is given the same input and speakers doubling the RMS power of the amplifier gives a 3 dB increase in volume which is only a little louder.
Fantastic explanation! I was just researching this subject to make sense of the various options on a mixer that has line level, Hi-z, and -26db input settings, compared to an interface that has line and instrument settings. I think I've finally figured out how they relate to my gain adjustment ranges. I believe you are correct that this subject should be introduced early in audio education, and would greatly appreciate a more detailed video on this if you ever find the time and energy to produce one. I've learned so much from you sharing you knowledge with us, As a result, I have been inspired to research many topics that I wasn't aware excised. You are such a valuable asset to the audio community! Thank you!
"How could he usefully explain the inordinately complicated subject of a decibel in five minutes," I asked myself. "Well, probably in a way that is both surprising, and entertaining," I replied. Curse you, Mr Worrall, you've failed to disappoint me yet again!
Holy crap this video closed so many knowledge gaps for me! I've played the guitar for over 20 years, studied music for 7... Never had a proper explanation of the qualifiers of dB scales until now
These low-effort videos are just as high effort as your usually well thought-out graphical content. Making things easy to understand is not always the easiest thing to do at all. That's quite high effort in itself :P
The first chapter of the amazing “Modern Recording Techniques” covers how decibels work. Musicians like to self learn but skip the books, this one is worth the time
A neat bonus is that decibels are great for napkin maths. You can loosely estimate a radio signals' pathloss with just a measurements built into your phone. For example, a typical 4G radio outputs 20 Watts into an antenna, which is 43dBm (from 1 milliwatts), and an antenna can easily have gain of 15-23dBi (gain from no directivity). So if the radio output is 43dBm, and the antenna has 15dBi, then the emitted radio power is 58dBm. The phone receives the signal at -61dBm, so the signal dropped roughly 119dBm along this path. Adding or subtracting the dB's is way easier than scaling by 10^12 imo. Looking forward to more videos! They're great regardless of format! :)
Too complicated, I just turn up Sausage Fattener until my tracks are a perfect square juicy sausage (I push it just enough for the the sausage to get angry but not TOO angry)
@@ArtiOfficialif you apply the Pointy Metal effect to the Sausage Fattener you can stop your tracks from splitting the dBFS down the middle. Remember to turn on the Background Woo setting like Captain Disillusion does on his videos and you'll be golden. Will, technically Fattened Sausage coloured, but you know how it goes.
Honestly Dan. The only thing I want from you anytime soon is to get better and get relief. Don't worry about high or low effort videos. Just worry about getting better and for the record I dig the low effort shorter videos.
Even though I already knew everything that was mentioned, I still feel like I learned something, like my understanding is now more whole, more unified. Also, thanks for the frequency analogy, that's a great one I'll use from now on when explaining dB to people
This one rings my Bel. 🤪 Seriously, this is an excellent plain-language explanation with some key concepts that can be quite difficult to grasp. Due to content like this, things have improved over the years and thankfully it's a long time since I heard anyone say "you've gotta use all the bits" - that one used to drive me up the wall in the 16-bit digital multitracker days! 🙄
I mean.. Why do i watch your videos? Not because i'm seeking some dopamine and want so flashy lights. When i click on one of your videos i know i need to be focused because there is coming something valuable! I kinda like this ''low efford videos'' how you call them. They are more ''hyper focused teaching videos'', no distraction, no BS. Just someone that tells you the answer to a question you have. If you would offer a masterclass i would buy it! ;) Love from Switzerland! Mike
It's worth noting that in the case of dithering, it can't even be argued that the 'dynamic range' of the audio is being truncated or quantized in any way. Dithering provides that infinite dynamic range with infinite _precision_ as well. Regardless of whether we're working with 32bit or 16bit or even 2bit audio, dithering allows the audio to be statistically encoded into the signal with the exact precision that it had originally, with the only consequence being a higher noise floor.
I think the fact that you can double the amplitude and it won't even sound that much louder is pretty amazing. It does also explain that when mixing single dB's can still have a big influence due to how much these waves might interfere with each other. It's also another pretty great showcase for why pushing for loudness is so pointless. It will destroy the balance but it won't even be experienced as that much louder.
I'm not sure you need to be worried about waves interfering when mixing unless you have phase issues between mics/etc. And loudness war is dumb but don't underestimate the psychological effect of something being even 1-1.5 dB louder! It can really fool you when A/Bing during mixing I think a lot of electronic musicians do go for loud just because they think they're supposed to, it makes the track's waveform "look right"
@@JoshWiniberg I think there's definitely ways to do that without actually doing the loudness war. There's ways to imitate blowing up the speakers without actually blowing up the speakers. Some smart sidechaining, you can let the bass distort your cymbals, just generally good sound design and mixing where they seem to result in rumbling. Actually really pushing towards loudness just feels like it's resulting in a bad mix.
I think this one needs a high Effort video The Rockbottom, fundamental of the dB Is it is a measure of power gain, not a measurement of voltage Yes, 6dB increase could mean the voltage has doubled, but it actually means there is four times power 3dB doubling or halving of the power is more interesting, because your hearing is logarithmic, and it is the smallest increase decrease you can easily perceive if you concentrate You can hear 1-2 dB difference I’m old school and this is how I was taught The Logarithm of 2 Is 0.3010😊 A complete history of the decibel is what is needed go back to the beginning when it was always stated in a 600 ohm line and work up to the modern day and show with ohms law how the Power formulas are derived😊😊 Because when you talk about the voltage in dB, you are using a power formula
no video you do could be considered "low effort" when you've put decades of work into perfecting your craft; you've put in years upon years of effort already. if these videos are "low effort", then so is LockPicking Lawyer, but they're not, as only years upon years of hard work can make something so hard look so easy, and the same goes for you❣
Great video, thanks! While I knew that turning up the level by 6 db (or isn't it actually 6.03 db or smt like that?) means doubling the volume and turning down by 6 db means halving the volume, I didn't even think about that and its implications when you posed the question regarding dynamic range. Somehow decibels are pretty abstract even though we deal with them daily. I'm all for more low effort videos!
Indeed. The only people that should be allowed to be in the pop charts are people with PhDs in electrical engineering. Any producer that won a Grammy without graduating in science should send back their award.
In audio it is always about references. For similar reasons we do not record "clean" signals for Music in an anechoic chambers, do we? Also, your "24 bit" audio interface is more likely "20-ish" bit (at best!) for Music\Audio related recordings. 4~6 bit are always occupied by the internal (self) noise of the device's own circuit noise. Yes, that noise is really "quiet" (because of the fairly low voltages… again "low voltage" is a reference to us humans) but it is still there.
Do you think you can make a video explaining the difference between +/-3dB and +/-6dB? Ie. From my understanding, +3dB is doubling of power, but doubling of amplitute is 6dB?
Db of what? If you now have 100 Db SPL from 100 watts of power and you want to increase to 103 Db SPL you'll need 200 watts. An increase of 6db demands 4 times the power. Edit: forgot...4 times the power is 2 times the voltage, is how you get +6Db Someone correct me if I'm wrong
How does that translate into sound power vs sound pressure levels? Eg. If I push a fader up by 3dB, does my SPL increase by 3dB? If I sum 2 channels of exactly the same audio together, is it a 3dB or a 6dB increase?
@@XenoeJin yes your spl will go up the same amount. If you add two identical signals (100% correlated) you've doubled the amplitude = +6dB. Adding two uncorrelated signals typically increases the level by 3dB.
Great video until the infinite dynamic range part... Suffers from the same issue as the sample rate video (which didn’t consider the computational approximations involved when implementing sinc functions in the reproduction of digital PCM audio; yes, a sinc function can perfectly reproduce the audio as long as the sample rate is above the Nyquist rate, but the sinc function is only approximated computationally and not implemented operationally in digital audio systems so that is never the case), namely that the actual computational implementation aspects are not considered and are assumed reducible to the mathematical. “At least theoretically” is a fine caveat if “theory” only ever meant math. The mathematical representation of an unbroken logarithmic function will tend to negative infinity as the lower limit, yes, but a computational representation of the same is irreducible to the former, even if the mathematical basis for the representation is the same. Additionally, most real-world phenomena that may be modelled logarithmically will only exhibit the same behaviour as a logarithmic function across certain specified ranges and cannot be arbitrarily extended. The computational approximations of the calculations involved (especially in fixed-point computational-arithmetic space, as is the case for 16-bit FP PCM audio) do not evaluate to the same result as the mathematical operations effectuated without approximation (if that is even possible, depending on the specific case). Negative infinity below -98.04dBFS (for 16 FP PCM audio) is nothing more than a symbolic representation of the fact that the mapping is discontinuous after that point. “Total silence” in digital audio is not negative infinity… it is just the actual loudness level of the quantization noise (just generate a “silent” 16-bit file and amplify it enough). The dB-to-Scaling factor calculations are only valid for data spaces that can contain, store, and represent those values; a 16-bit fixed point sample cannot store an infinitesimally small quantization value and cuts off well above that, which is what generates the quantization noise. The water glass analogy is also ill-suited to make the point you’re trying to - especially for how far it’s pushed - because there aren’t actually an infinite number of water molecules in the glass.
This used to confuse the hell of out me as I learned image processing before audio, so when I started writing audio code it looked a lot like graphics functions. When people would moan about the dynamic range of a 16/44 WAV file, I said it was not a trait of the file format but the interpretation thereof. I could scale the signal to any range I wanted, even apply non-linear transforms like a sigmoid curve to really mess things up. A digital stream is just numbers. Only when you translate them to/from analog does it make sense to think in dB.
Hey Dan. Was wondering if you are planning on doing a video on ai and algorythm based mixing and plugins like soothe, RX and surfer EQ at any point? Hope your health situation is getting better. Best wishes, from Denmark!
Decibels is for loudness? Hertz is for (rental cars) frequencies or pitch? The question should be what's the lowest loudness the human ear can hear? And what's the maximum loudness, before it becomes painful or dangerous to listen to. What's the lowest and highest frequency the human ear can hear. Mixing audio isn't interested in infinity anything, it's interested in what limitations are there to mixing music. What are the parameters of frequencies and loudness we should concern ourselves with?
Hello Dan! can you do one of these quick ones for clipping? (on a clipper plugin i mean) and why it is not just compression? as far as i know clipping actually lowers the sample value thus reshaping the wave instead of just lowering gain on the sample leaving you with the same wave , but i still have trouble wrapping my head around it! i'd love to know how the plugin does it internally in asking from my ignorance so i welcome anyone to answer btw!
Saying 16 bit audio is infinite is not correct. Its not a defined concept to begin with, but in the digital domain you are simply dealing with ones and zeroes. 0 doesnt equal negative infinity. In 16 bit audio, you can have 2^16 possible values in two's complement.
I was fooling around in Audacity one day and found I could cut around 800-850dB (then boost back to the original level) before any audible quantization noise was apparent, with major degradation as you approach -900 dB; trying the same thing in Ableton (with a bespoke Max4Live patch, because Ableton jumps straight to 0 WAY sooner than that) resulted in the audio becoming silent any time it dropped below a similar level rather than any quantization noise. So you potentially have more headroom than your DAW is letting on, since DAW developers seem to agree that sounds near -100 dB are inaudible and can be dropped.
let me correct you, as far as many people confuse quantization noise with phase noise. What you get below -800 dB in 32 bit float formats is not the quantization noise, but the phase noise. The quantization noise in 32 bit float starts from -139 dB FS as the 23 bit mantissa is not capable of delivering the accuracy down below the entire dynamic range.
A step-looking waveform is only the mode of the visual representation of the descrete signal, but after you decode it into analog form it is never staired as far as the high order harmonics are filtered out leaving your signal wave smooth. But even if you ever have to deal with a square looking wave quantization errors have nothing in common as far as they can only appear in float point formats beyond mantissa's range, so nothing to worry about.
3:00 acoustician here - SPL isn't normally expressed as 'dB SPL' - the correct index for sound power (LWA) or in this case, sound pressure (LPA) - however in environmental acoustics you'll usually use the LAeq,T index or the LAmax index to express the average and the maxima sound pressure levels. It's important to note that for sound pressure, a decibel is simply a ratio of the measured Pascals against the reference Pascal value. So for airborne sound, it's normally 20*LOG(Measured Pa/2*10^-5 Pa).
It's also important to note that you can take sound pressure and sound power and use a lightbulb as an analogy. The sound power is like the watts of power the lightbulb is using - this is a constant and is the OUTPUT of the light (speaker/source) itself. The sound pressure is a relative point away from that source. It's important to also know that sources can be area, planar, line or point sources.
And to add, weighting is very important too! A (for environmental acoustics, this emulates the effective weighting of the human ear) - C (similar to A, but allows room for lower frequencies, this is normally great for impulsive sounds or for music-related environmental sound) and then Z (this is just the unweighted sound).
This is a great video, something I wish I had when I first started out!!
great post man. everyone should read this immediately.
The A weighting curve adjustment should be scrapped and replaced by something actually relevant to noise nuisance value.
@@jrgroberts A weighting makes sure that single value broadband sound levels are most reflective of how humans actually perceive sound across the frequency spectrum. Look up the fletcher munson curve.
@@producermathew The Fletcher-Munson curve was created +90 years ago, when measurement equipment was nothing to par compared to what we use nowadays. Even if it has been was further revisions of the Fletcher-Munson curves, the A weighting curve is a scam. That helps politicians to avoid legal issues of millions of people contamined by out of the scope noise due to LAeq filters. It makes sense to revisit these concepts as @jrgroberts suggests, because the incremental low freq noise environments we suffer in the modern world.
@@PereRevert the A weighting curve is not inversely proportional to the fletcher munson curve. The A weighting correction is standardised within IEC 61672-1, ISO 226 and ANSI S1.4.
A-weighting is generally considered representative of a typical curve for human hearing in so much that we do not perceive frequencies in the same way as an ultra flat type 1 measurement system.
You say it’s a scam, yet don’t provide any context. This is false.
Your “low effort videos” are the ones I appreciate the most! No distracting background music, just your voice transporting expertise. It’s like technical ASMR 😄
Like wise. Might be an idea do try some long form live/podcast style content, maybe q&a?
this
right! Without background music is even better and it's easier to follow and understand!
Agreed❤
I believe I'm speaking for the majority of us that genuinely appreciate what you're doing here, and we don't demand "high-effort videos", even though it feels like a out of time Christmas present to see you in my feed, what we most cherish is You, and your well being first, and then everything else, Dan!
I hope you're re doing well and you take care of yourself.
indeed :)
Well said 👏🏼👏🏼
I appreciate these no-fluff, erudite videos, that aren't padded out to 10:00 minutes just for the UA-cam algo. Respect to you Dan for your willingness to share your knowledge without wasting our time.
To be fair, Dan can waste my time whenever. Would enjoy these videos even if I wasn't interested in audio engineering at all lol
the 10 minute thing is just an outdated myth anyway, it doesn't actually help with the algorithm directly.
@donit. My concession to the algorithm was to pointlessly render this at 4k resolution.
This. And it's actually a mark of smart people to respect their audience - they realise it's what's going to being them back instead of making them hesitate over whether a new video is another diluted waste-of-time video.
The pitch analogy was really effective and quite amusing. "This note is three semitones"! :)
I understood the db are logarithmic, and that -6db halved the signal, but this was a great dive into the topic and covered some gaps in my knowledge. And like you said, no shame to the original commenter; you don't know what you don't know, and we're all here to learn.
Now send this to the guy who told me 24 bit WAV wasn't enough because computers are 64 bits now and that's why vinyl sounds better.
Absolutely loved this. This goes beautifully along with George Massenburg when he says ‘anything above 20 bits is “marketing bits” - brilliant’
32 bits fp are very, very beneficial for processing and design. The video does not state that 20/24 bits is end all be all, at all, and the "infinite dynamic range" is a very cheeky statement bordering on a lie. It is like saying that a range of numbers from 0 to 65000 is infinite. _Technically_, it is, but you still can't write 65001 to this range.
Absolutely 32/64 bit are necessary for dsp processing. But the idea that 24 bit and above are necessary for a/d and d/a is always a little tenuous. I just think that there is too much emphasis on tech sometimes and less talk about how music makes us feel. ✌️😊
I pity the fool that wants to write 65001 when 32000 is close enough for jazz.
The fact that you can explain such complex topic with just few sentences and make it understandable is the best proof you truly understand the context. Still ones of the best videos about audio engineering in the whole internet.
Though, please don't stop making them higher effort videos even when these are absolutely brilliant too. Just the fact that you make banger music is a reason enough to watch them, but they bring so much to thw community.
Your "low-effort" videos are still amongst the most informative out there. Hope you're keeping well and keep up the great work (as long as your health/energy allows for it)!
2:55 0dBSPL is referenced to a pressure, which is 20 µPa. This is the "quietest sound" we can generally hear, as you correctly point out, but this refers to a 1KHz sine wave. Since our hearing is a bit more sensitive in the region from 2-4KHz, it's actually possible to hear negative dBSPL figures!
Would've also loved to hear an explanation about what dBFS actually means (the reference to the digital maximum level), but it's a nice video overall :-) Looking forward to the next one!
From what I understood, full scale is just when you reach the integer limit when representing amplitude. However, the exact value depends on how the system you’re using works (I’m not sure if there are any standards).
Also, DAWs are not clipping past full scale because they’re using floating point numbers instead of integers, which can store bigger numbers than integers.
Full scale integer is the largest number you can represent for positive cycles, or zero for negative cycles. Floating point audio represents full scale as 1.0 or -1.0. It used to mean the maximum amplitude that you could never exceed; now, with floating point processing, it means the maximum level that you're going to have to come back down to before releasing it.
Well, dBFS can be more complicated than that, if you read the wikipedia article on it, there are multiple conventions, and when following the AES convention (among others), a full scale square wave actually has +3dBFS.
I work as a software engineer in an audio processing software company, and most of us use the other convention where we mean it to be the ratio to the maximum amplitude (of the (true or not) peak value of some signal). Though a DSP engineer once pointed me to the other convention (of which he is convinced is the right one, as it is defined by multiple standardization entities as such) and since I'm quite careful how I use the term, I try to avoid it if I can and always feel a bit conflicted :D.
I guess it depends on the context which convention is more useful.
@@frydac This has to do with visual adjustment of RMS levels. A full scale square wave will have a level of 0dBpk and 0dBrms, which computes to a crest factor of 0. Many plugins&softwares adjust RMS levels to show a 3dB higher value, to reference to the crest factor of a clean sine wave, in which case you'll see 0dBpk and 0dBrms for a sine wave and 0dBpk and +3dBrms for a square wave as you point out, but simply due to this visual adjustment.
You can however get theoretical peak values above 0dBFS, through so-called inter-sample peaks. Any fully scale sqaure wave will produce such inter sample peaks due to our processing being bandwidth limited, which makes an inifnitely short rise time impossible. As far as I know, modern DACs have a bit of headroom in the DA conversion to account for inter-sample peaks, but also modern mastering software like Ozone has inter-sample peak detection ("True peak"), which then makes sure this "digitally invisible" clipping can't happen.
By the way, avoiding intersample peaks is sort of an overzealous measure. Basically all DACs these days either have sufficient headroom not to produce audible clipping when converted to analogue, or it doesn't matter because they are crap anyway. 😂
First and foremost, your health is top priority. Take care of yourself the best way you can. That will not only benefit you, but all of us as well. 🙂 Wishing you the best of luck and quick relief. Secondly, it's not often somebody with extensive knowledge and experience takes the time to share it. Finding someone to share it with such integrity, morals and straight forward common sense is even harder to find. Then you do all that in a pleasing, humorous, respectful way. This video is a prime example, by not calling out the person asking the question and not being degrading in the response. You sir, have a gift with your presentation, thought process, and your voice. When I say "voice" I'm referring to your speaking voice, not your singing voice. That's a whole other subject. 😃 Had to leave you with a smile. Thank you for all you do and just being you.
Watching this to pass the time while I'm bedbound with IBD pain and fatigue ❤️🤒☺️
Hope it helps, get better soon...
@@DanWorrall Cheers Dan 🥰!!
Well, now I know. I never did before. 45 years working audio-adjacent and it never clicked, but after 5 minutes of a “low effort” video from Dan and it did. Damn fine job!
Once again something I think i sort of knew explained so clearly and concisely that I could probably explain it to someone else (except I'll probably forget)...nice one as ever. Cheers!
What do you mean higher efford? This is a complex topic explained shortly and very well. It does not need to be an hour long, or have amazing graphics, if it can be explained simply in 5 minutes. As always, thank you for making something complex so simple and understandable!
The simplicity is so sophisticated
I for 1 really appreciate your low effort, but objective, content. :)
This video triggered my will to revisit my knowledge (or lack thereof) about this measure. And I learned how summing two uncorrelated sources result in terms of gain in dB, which is not so straightforward: summing 2 uncorrelated signal of the same level results in a 3 dB gain, while summing signals 6 dB apart (i.e. one is twice lower) results on a gain of 1 dB. Interesting stuff. Thank you Mr Worrall.
This is an interesting topic! What if I have 40 tracks but I’d like to lower my out by 1-2 db, how much do I need to lower the others?? Not expecting an answer 😅😅 but definitely something I’m going to research. I’d love Dan’s take on it 😁
Just a precaution....
How do you consider 2 unrelated signal as "the same level"? Basically what did you use to measure them? The loudness of 2 unrelated signal will be heavily dependent on how they overlap, so I wonder how useful it is to consider how louder they can become if added together?
@@brettclarinmusic1-2db!
Reducing all tracks by 1-2db is the same as reducing the master output by 1-2db (assuming no fx)
Yep, this vid sent me back to the dB pages and calculators! I always like to remind myself that however loud one instrument is, adding a second one makes things 3dB louder. So when I go to eq something, like the high end of guitar strumming, and boost by 3 dB that’s like adding a second guitarist (at that frequency range)
Thank's for everything, Dan. Hope you're doing well
You always manage to take something fundamental (that we have all thought about) and still shed light on it.
Hope your doing well! Great work as ever.
"Ce qui se conçoit bien s'énonce clairement, et les mots pour le dire arrivent aisément". Mr Worrall, You're the clearest, yet most concise, on any audio related topic you discuss. And the metaphores are just on point. Perfect explaination
You've already put in the effort with the entirety of your career and knowledge. It may be basic and low effort to you now but it's vast to those who didn't know, and a refresher for those who did know. Hope everything is going good and thanks for all the videos
The Shumann resonance reference got a laugh out of me.
Excellent as always, Dan.
Well done again, Dan. Clear and to the point.
Also, for posterity for up and coming engineers: LUFS has *no* bearing on how loud a record can be cut. Make it sound good. Short enough side lengths. Nothing sibilant or clipping.
No worries about the quality of the video, this is actually an incredible value video
Always very appreciative of your videos, regardless of whether they're shorter and simpler or longer and more in depth. There's always very well articulated and helpful information to be found and I'm super grateful for it!
Something related to this I have always got some amusement from is given the same input and speakers doubling the RMS power of the amplifier gives a 3 dB increase in volume which is only a little louder.
Thank you Dan! Praying for you to get well.
Fantastic explanation! I was just researching this subject to make sense of the various options on a mixer that has line level, Hi-z, and -26db input settings, compared to an interface that has line and instrument settings. I think I've finally figured out how they relate to my gain adjustment ranges.
I believe you are correct that this subject should be introduced early in audio education, and would greatly appreciate a more detailed video on this if you ever find the time and energy to produce one. I've learned so much from you sharing you knowledge with us, As a result, I have been inspired to research many topics that I wasn't aware excised. You are such a valuable asset to the audio community!
Thank you!
Hope you're feeling better Dan. Thanks always for sharing information.
Surprised that a simple graph of logarithmic function v. linear function wasn't used.
Love your channel. You have a great way about you.
i just love how all these "low effort" videos are just VERY IMPORTANT THINGS that i've should've known from the start
"How could he usefully explain the inordinately complicated subject of a decibel in five minutes," I asked myself.
"Well, probably in a way that is both surprising, and entertaining," I replied.
Curse you, Mr Worrall, you've failed to disappoint me yet again!
One of Music’s GOATs graces us with his knowledge once again 🤟🫡🤟
Thank you for the new video, hope everything is all right with you, sending positivity and my thanks for you uploading quality content.
ALL your videos are very insightful Dan! 👍
I hope you're feeling better & better Dan (at least 6 dB per day! 😎) I keep learning more and more 🙏
Thanks Dan, another great one. Take your time. We'll be here.
Great explanation!
Holy crap this video closed so many knowledge gaps for me!
I've played the guitar for over 20 years, studied music for 7... Never had a proper explanation of the qualifiers of dB scales until now
These low-effort videos are just as high effort as your usually well thought-out graphical content. Making things easy to understand is not always the easiest thing to do at all. That's quite high effort in itself :P
The first chapter of the amazing “Modern Recording Techniques” covers how decibels work. Musicians like to self learn but skip the books, this one is worth the time
Thanks fren
A neat bonus is that decibels are great for napkin maths.
You can loosely estimate a radio signals' pathloss with just a measurements built into your phone. For example, a typical 4G radio outputs 20 Watts into an antenna, which is 43dBm (from 1 milliwatts), and an antenna can easily have gain of 15-23dBi (gain from no directivity).
So if the radio output is 43dBm, and the antenna has 15dBi, then the emitted radio power is 58dBm. The phone receives the signal at -61dBm, so the signal dropped roughly 119dBm along this path.
Adding or subtracting the dB's is way easier than scaling by 10^12 imo.
Looking forward to more videos! They're great regardless of format! :)
Too complicated, I just turn up Sausage Fattener until my tracks are a perfect square juicy sausage (I push it just enough for the the sausage to get angry but not TOO angry)
yeah! who needs dbSPL or dbA or even LUFS? my track is 20 sausages loud! WITH MUSTARD!
sausage fatten the main mix then add the transients of the kicks in for serrated sausage
You're both so right, personally I know it's loud enough when I hear it cracking, just like sausage on a grill. Don't want my tracks raw.
@@ArtiOfficialif you apply the Pointy Metal effect to the Sausage Fattener you can stop your tracks from splitting the dBFS down the middle. Remember to turn on the Background Woo setting like Captain Disillusion does on his videos and you'll be golden. Will, technically Fattened Sausage coloured, but you know how it goes.
glad to hear you back, hope you are doing okay Dan
Honestly Dan. The only thing I want from you anytime soon is to get better and get relief. Don't worry about high or low effort videos. Just worry about getting better and for the record I dig the low effort shorter videos.
Even though I already knew everything that was mentioned, I still feel like I learned something, like my understanding is now more whole, more unified.
Also, thanks for the frequency analogy, that's a great one I'll use from now on when explaining dB to people
This one rings my Bel. 🤪
Seriously, this is an excellent plain-language explanation with some key concepts that can be quite difficult to grasp. Due to content like this, things have improved over the years and thankfully it's a long time since I heard anyone say "you've gotta use all the bits" - that one used to drive me up the wall in the 16-bit digital multitracker days! 🙄
„rings my Bel“ 😂
These low effort videos are as good as your normal ones. I think if you stuck to this format, your audience would be equally as happy
Xeno's Bow toward infinity. This was wonderful. Thanks, Dan. Will be sharing as needed.
I mean.. Why do i watch your videos?
Not because i'm seeking some dopamine and want so flashy lights.
When i click on one of your videos i know i need to be focused because there is coming something valuable!
I kinda like this ''low efford videos'' how you call them. They are more ''hyper focused teaching videos'', no distraction, no BS. Just someone that tells you the answer to a question you have.
If you would offer a masterclass i would buy it! ;)
Love from Switzerland!
Mike
You know that feeling when a new thing comes out you feel the urge to watch all the previous episodes from the start or you might miss the impact
It's worth noting that in the case of dithering, it can't even be argued that the 'dynamic range' of the audio is being truncated or quantized in any way. Dithering provides that infinite dynamic range with infinite _precision_ as well. Regardless of whether we're working with 32bit or 16bit or even 2bit audio, dithering allows the audio to be statistically encoded into the signal with the exact precision that it had originally, with the only consequence being a higher noise floor.
I think the fact that you can double the amplitude and it won't even sound that much louder is pretty amazing. It does also explain that when mixing single dB's can still have a big influence due to how much these waves might interfere with each other.
It's also another pretty great showcase for why pushing for loudness is so pointless. It will destroy the balance but it won't even be experienced as that much louder.
Strictly speaking destroying the balance is what increases the perceived loudness (LUs). Loudness is a mixture of volume and dynamic range.
Well they're not pushing for objective loudness on paper, they want it to feel big and loud.
I'm not sure you need to be worried about waves interfering when mixing unless you have phase issues between mics/etc. And loudness war is dumb but don't underestimate the psychological effect of something being even 1-1.5 dB louder! It can really fool you when A/Bing during mixing
I think a lot of electronic musicians do go for loud just because they think they're supposed to, it makes the track's waveform "look right"
@@JoshWiniberg I think there's definitely ways to do that without actually doing the loudness war.
There's ways to imitate blowing up the speakers without actually blowing up the speakers.
Some smart sidechaining, you can let the bass distort your cymbals, just generally good sound design and mixing where they seem to result in rumbling.
Actually really pushing towards loudness just feels like it's resulting in a bad mix.
@@georgerosebush9754 I agree, and going for objective loudness isn't the way to get there.
Well done, professor Dan (from a recently retired mathematics professor).
I love the audio only content. Longer form would be even better 😁
Thank u as always ! Appreciate any effort
I think this one needs a high Effort video The Rockbottom, fundamental of the dB Is it is a measure of power gain, not a measurement of voltage
Yes, 6dB increase could mean the voltage has doubled, but it actually means there is four times power
3dB doubling or halving of the power is more interesting, because your hearing is logarithmic, and it is the smallest increase decrease you can easily perceive if you concentrate You can hear 1-2 dB difference I’m old school and this is how I was taught
The Logarithm of 2 Is 0.3010😊
A complete history of the decibel is what is needed go back to the beginning when it was always stated in a 600 ohm line and work up to the modern day and show with ohms law how the Power formulas are derived😊😊
Because when you talk about the voltage in dB, you are using a power formula
Found myself laughing out loud at the schumann resonance jab. Well payed!
Great analogies, as always. Thanks for this video
Love the fundamentals. I do wireless networking which also uses dB
Thanks mate, be well.
Excellent video.
no video you do could be considered "low effort" when you've put decades of work into perfecting your craft; you've put in years upon years of effort already. if these videos are "low effort", then so is LockPicking Lawyer, but they're not, as only years upon years of hard work can make something so hard look so easy, and the same goes for you❣
Keep up the excellent education content!
Great video, thanks! While I knew that turning up the level by 6 db (or isn't it actually 6.03 db or smt like that?) means doubling the volume and turning down by 6 db means halving the volume, I didn't even think about that and its implications when you posed the question regarding dynamic range.
Somehow decibels are pretty abstract even though we deal with them daily.
I'm all for more low effort videos!
The poetry of Dan’s language 😅
This is very scientific of you. Most modern self proclaimed producers don't have a clue about electricity and scientific measurements.
Indeed. The only people that should be allowed to be in the pop charts are people with PhDs in electrical engineering. Any producer that won a Grammy without graduating in science should send back their award.
I understand because I already knew, but I don't think someone seeing this for the first time will fully get the message
Nice job, professor!
Dan always comes with receipts haha! Brilliant stuff mate!
This is the correct format for you. However I would suggest you should write a book, and release it as an audiobook as well.
This is the guy to follow, kids.
One of my physics professors told the class, "don't try to remember all these formulas, just remember how to look them up".
I think for your videos a "thanks for listening" would be more appropriate as a closer :D
In audio it is always about references. For similar reasons we do not record "clean" signals for Music in an anechoic chambers, do we?
Also, your "24 bit" audio interface is more likely "20-ish" bit (at best!) for Music\Audio related recordings. 4~6 bit are always occupied by the internal (self) noise of the device's own circuit noise. Yes, that noise is really "quiet" (because of the fairly low voltages… again "low voltage" is a reference to us humans) but it is still there.
Do you think you can make a video explaining the difference between +/-3dB and +/-6dB? Ie. From my understanding, +3dB is doubling of power, but doubling of amplitute is 6dB?
Db of what? If you now have 100 Db SPL from 100 watts of power and you want to increase to 103 Db SPL you'll need 200 watts. An increase of 6db demands 4 times the power.
Edit: forgot...4 times the power is 2 times the voltage, is how you get +6Db
Someone correct me if I'm wrong
This is what I am confused about, when is doubling of something 3dB or 6dB?
Amplitude is the height of the curve. Power is the area under the curve. 6dB doubles amplitude. 3dB doubles power.
How does that translate into sound power vs sound pressure levels?
Eg. If I push a fader up by 3dB, does my SPL increase by 3dB?
If I sum 2 channels of exactly the same audio together, is it a 3dB or a 6dB increase?
@@XenoeJin yes your spl will go up the same amount. If you add two identical signals (100% correlated) you've doubled the amplitude = +6dB. Adding two uncorrelated signals typically increases the level by 3dB.
thanks for the vid, dan the man!
You and Andrew Zeleno are the smartest audio engineers
Ok. I'll come back to this video in a couple of years. It's not the first time. 😂 I know I'm not alone.
Great video until the infinite dynamic range part... Suffers from the same issue as the sample rate video (which didn’t consider the computational approximations involved when implementing sinc functions in the reproduction of digital PCM audio; yes, a sinc function can perfectly reproduce the audio as long as the sample rate is above the Nyquist rate, but the sinc function is only approximated computationally and not implemented operationally in digital audio systems so that is never the case), namely that the actual computational implementation aspects are not considered and are assumed reducible to the mathematical. “At least theoretically” is a fine caveat if “theory” only ever meant math.
The mathematical representation of an unbroken logarithmic function will tend to negative infinity as the lower limit, yes, but a computational representation of the same is irreducible to the former, even if the mathematical basis for the representation is the same. Additionally, most real-world phenomena that may be modelled logarithmically will only exhibit the same behaviour as a logarithmic function across certain specified ranges and cannot be arbitrarily extended. The computational approximations of the calculations involved (especially in fixed-point computational-arithmetic space, as is the case for 16-bit FP PCM audio) do not evaluate to the same result as the mathematical operations effectuated without approximation (if that is even possible, depending on the specific case). Negative infinity below -98.04dBFS (for 16 FP PCM audio) is nothing more than a symbolic representation of the fact that the mapping is discontinuous after that point. “Total silence” in digital audio is not negative infinity… it is just the actual loudness level of the quantization noise (just generate a “silent” 16-bit file and amplify it enough). The dB-to-Scaling factor calculations are only valid for data spaces that can contain, store, and represent those values; a 16-bit fixed point sample cannot store an infinitesimally small quantization value and cuts off well above that, which is what generates the quantization noise. The water glass analogy is also ill-suited to make the point you’re trying to - especially for how far it’s pushed - because there aren’t actually an infinite number of water molecules in the glass.
Dean Loves Decibel
I feel like this guy always know what he is talking about
This used to confuse the hell of out me as I learned image processing before audio, so when I started writing audio code it looked a lot like graphics functions. When people would moan about the dynamic range of a 16/44 WAV file, I said it was not a trait of the file format but the interpretation thereof. I could scale the signal to any range I wanted, even apply non-linear transforms like a sigmoid curve to really mess things up. A digital stream is just numbers. Only when you translate them to/from analog does it make sense to think in dB.
If the world ends anytime, please make sure this guy leaves the messages for potential surviviors...
Thanks as always. and… Look after your health. We will be here when your ready. :+)
Hey Dan. Was wondering if you are planning on doing a video on ai and algorythm based mixing and plugins like soothe, RX and surfer EQ at any point? Hope your health situation is getting better. Best wishes, from Denmark!
Thanks Dan
Decibels is for loudness? Hertz is for (rental cars) frequencies or pitch? The question should be what's the lowest loudness the human ear can hear? And what's the maximum loudness, before it becomes painful or dangerous to listen to. What's the lowest and highest frequency the human ear can hear. Mixing audio isn't interested in infinity anything, it's interested in what limitations are there to mixing music. What are the parameters of frequencies and loudness we should concern ourselves with?
Wowww....thanks Dan🙏🏻🙏🏻
Wonderful.
Thanks. Get well.
Checks out, I actually learned that in audio school - maybe not first day, but within the first 2 months
Hello Dan! can you do one of these quick ones for clipping? (on a clipper plugin i mean) and why it is not just compression? as far as i know clipping actually lowers the sample value thus reshaping the wave instead of just lowering gain on the sample leaving you with the same wave , but i still have trouble wrapping my head around it! i'd love to know how the plugin does it internally
in asking from my ignorance so i welcome anyone to answer btw!
Saying 16 bit audio is infinite is not correct. Its not a defined concept to begin with, but in the digital domain you are simply dealing with ones and zeroes. 0 doesnt equal negative infinity. In 16 bit audio, you can have 2^16 possible values in two's complement.
So dB goes into a glass, then?
I was fooling around in Audacity one day and found I could cut around 800-850dB (then boost back to the original level) before any audible quantization noise was apparent, with major degradation as you approach -900 dB; trying the same thing in Ableton (with a bespoke Max4Live patch, because Ableton jumps straight to 0 WAY sooner than that) resulted in the audio becoming silent any time it dropped below a similar level rather than any quantization noise. So you potentially have more headroom than your DAW is letting on, since DAW developers seem to agree that sounds near -100 dB are inaudible and can be dropped.
let me correct you, as far as many people confuse quantization noise with phase noise. What you get below -800 dB in 32 bit float formats is not the quantization noise, but the phase noise. The quantization noise in 32 bit float starts from -139 dB FS as the 23 bit mantissa is not capable of delivering the accuracy down below the entire dynamic range.
@@CyrilleBoucanogh In the instance I described, it's definitely quantization noise because it resulted in a stair-stepping waveform
A step-looking waveform is only the mode of the visual representation of the descrete signal, but after you decode it into analog form it is never staired as far as the high order harmonics are filtered out leaving your signal wave smooth.
But even if you ever have to deal with a square looking wave quantization errors have nothing in common as far as they can only appear in float point formats beyond mantissa's range, so nothing to worry about.
THE BIG TALKING W HAS SPOKEN
good morning, thank you