*Everything you need to know for audio playback and production...* TLDR: 16 bits is enough for playback, and 24 bits is only helpful for production. Small correction: Bit-depth doesn't really get you "resolution". *1) Frequency gives you resolution for audio.* As far as frequency goes, 44-48kHz is enough because human hearing doesn't really go beyond 20-20k, and 44.1 gives you up to 22kHz accurately according to the Nyquist formula. 44.1kHz is the CD audio standard. 48kHz is DVD audio standard. *2) Bit depth gives you dynamic range.* 6dB of dynamic range per bit. This means 16-bit has 96dB DR and can easily cover from 20dB (silent room) to 116dB (front row at a rock concert, almost hearing damage). 24-bit has 144dB DR and can cover the range between a silent room and a jet engine. *3) Higher bit-depth lowers the effect of noise while mixing audio.* When you process audio, every operation adds a little bit of noise. With 16-bit, you have 65k steps. With 24-bit, you have 16M steps. A random error shift of 10-100 steps will be ever-so-slightly more significant if you have less steps to scale it. The absolute range means you have more room for error.
Well, you can think 24bit as higher "vertical" or signal amplitude resolution. Like 1920x1080 full hd being 1920 the sampling frequency, and 1080 the bit depth. Translating to an wave signal, higher bit depth means bigger headroom, snr and blablabla =D
Ah that's a GREAT way to put it. Vertical resolution vs Horizontal resolution. Dang it. I wish I had thought about putting it that way in the video. =) Thanks for the better explanation.
Victor Amicci To be more precise... bit-depth is more like the intensity. I've already explained how frequency is the resolution; the important thing is that sound isn't measured in 2D space, it's measured in frequency. The sampling frequency lets you "see" any frequencies that are 50% or less. So let's say 1920x1080 is like being able to see up to 20kHz (functionally enough), and 1280x720 is seeing up to 16kHz (practically enough), etc. Since bit-depth is intensity, think of how many colors can be displayed in a given sample (either a pixel, or a timecode). You may have heard of 24-bit color, which uses 8 bits for each of 3 color channels (0-255). The absolute dynamic range is the difference between the highest and lowest values you can communicate. In other words... what's 0% brightness? What's 100% brightness? How many steps are in between? The standard is to put 256 steps for each color, or 65,536 for the human-audible range. Sure, you can use more than 256 steps for color; HDR displays do this, and there are definitely more than 16 million colors. But there's a diminishing point of return. For our rods and cones, we've got a bit of room left to grow, but for the hairs in our ear, we've already hit the practical limit. We can resolve as far down as we can differentiate two different sounds as unique, and we can cover the range between a silent room and a rock concert. We don't really need to reach jet-engine levels for playback; it's just useful to have more room for error if we make mistakes while doing the math. We're going to divide by 256 anyway when we want to listen to it.
Here's a picture showing what bit depth is like for pictures: www.azooptics.com/images/Article_Images/ImageForArticle_1151(1).jpg And for audio: vignette.wikia.nocookie.net/digital-audio/images/f/f7/8_16_bit_depth.gif/revision/latest?cb=20140415224229
Thanks for the video! I’ve never realized the actual difference so clear, before I saw picture quality and numbers difference. In my understanding: record in 24 bits for better processing (noise reduction, compression, EQ, mixing) and publish in 16 because no one would ever notice the difference on TH-cam or in a typical podcast.
Today’s ADC units are far better than those from the past. A bit depth of 16 and a sample rate of 44.1 kHz is more than sufficient for the human cochlea to hear amplitudes and frequencies as accurately as our hearing apparatus is capable of. This is true of a straightforward record and playback. However, when we edit digital audio in any way, we introduce new quantization errors that build up in the lower bits the more we edit. Enough so that we can start to hear the artifacts introduced by editing. So 24 bit is recommended for professional recordings that will be edited so that the artifacts generated in the lower bits are rendered well nigh insignificant because a smaller portion of overall bits are corrupted by S/E errors due to editing. Therefore if the editing after recording is minimal, then 16 bit is all you really need.
This is the most important part! 16bit 44.1kHz really truly is indistinguishable from 24bit or 32bit 48kHz. But if you're messing with the recording a bunch, it helps reduce noise! Another vital point; most older hardware and even some more modern programs can only use 16bit samples, so there's that.
Higher bit depth means greater precision in mapping an analogue voltage to a corresponding digital value, and allows for a greater signal to noise ratio. 16 bit (2 bytes) means you have 65,536 possible digital values available for representing the analogue voltage. 24 bit (3 bytes) means you have 16,777,216 possible digital values for representing the analogue voltage. So, 24 bit can more precisely record what the analogue voltage was. It's important to know that the digital values are evenly spread out between maximum voltage and minimum voltage. This is important when you're dealing with very low voltage levels - there are fewer digital values between zero and the value closest to the one that represents the voltage if you're using a lower bit depth. That means any editing of the resulting digital audio has fewer values to change between. So it's always better to capture in the highest bit depth possible, and do any processing at an even higher bit depth, before mastering down to the bit depth needed for the final product. 16 bit allows for a maximum signal to noise ratio of 96dB, and 24 bit allows for up to 144 dB s/n ratio. Very few audio music recordings are made that have a dynamic range that cannot easily be accommodated within a 96dB noise floor, and almost no commercial music recordings have that sort of dynamic range. Quantization is the process of mapping the voltage to a specific digital value. Dithering is a way of doing the quantization that reduces digital artefacts caused by quantization. Bit Depth is about digital accuracy of the sampled analogue signal. Sample rate affects maximum frequency that can be digitally represented. Sample rate is a different matter that has it's own considerations that need to be understood.
bro I want to thank you for giving me exactly what I needed within the first minute of your video the rest of it I skimmed but the first minute was the info that I wanted. Thank you!
For the picture analogy, I think it might be closer to a picture format that can create 65,000 colors vs a picture format with 16-million different colors. But I'm no expert
Podcastage so a lot of different engineers that I send stuff off to are all so picky. Some people 16-441 mono some want 24-441/48 mono wav. So I guess it comes down to preference. But it’s sucks you can’t work with the sample rates back and forth with projects. For me, pro tools. If I created a session in 48, protocols or Windows or whatever will not open up the 44 project and of it does the system gets confused and all of sudden everything plays in slowmotion...including desktop audio. Yeah...questions. I learned all of this by serious trial and error. So I guess my main question is that when working with projects or systemsl audio drivers...why don’t they play well with each other and why can’t you jump back and forth we between them, with in them at times.
Its not the file itself, its the actual session. For example, if a created a protools session and what the devices are set to 48 or 44. In 48, it wont open on another system that is 44, i have to go into the option of the DAW and change it before opening the project. And if im using the same DAW as an audio card for my PC for my desktop audio and they dont match up, then i get the slowdown effect. YEah i know its confusing, but its real, and i found out the hard way through numerous trial and error
so this error would happen specifically when i was using my mixer as my Yamaha MG12XU as my sound card and ProTools DAW. My windows system would default to 48 but protools was set to 44. Once protools opened up it would try to convert the system that was 48 to 44 and every sound from the PC would play in slow motion, even the boot up screen. In the the end, i had to give up using the mixer as an all in one SoundCard and DAW..it just wasnt worth the constant hassle and uninstalling and reinstalling the drivers to fix the problem when ever it would come up
I think you’re wrong about bit depth. Sample rate resolution is like 1080 vs 4k tv resolution. But Bit depth is about the dynamic range of the sound; i.e. how low level a sound can be recorded at compared to the highest level.
I was thinking the exact same thing. I'm just getting into this subject, but my understanding and what I've read so far that its a lot more about dynamic range than a better sounding initial recording.
Good, informative video. One of the best ones done on this - would be cool to see more on why we need 24 bit vs 16 bit for recording music vs podcasts though.
Hi, I enjoyed your territorial on a BIT Recording, I have finally found out why some music sounds better on recording on TH-cam than others, Just recently I have been listening to a lot of music at 70-year-old dude retired enjoy listening to music just recently I rebuilt my transmission line loudspeaker cabinets , I’ve been listening to a lot of music, Classical to heavy metal , Play the same music recorded by different people obviously on different devices and you can notice when somebody record something in 16 to 24 BIT , It is surprising how much difference it can be ruin a piece of music, I hadn’t realised, Until now, Great territorial found something at my age of life, Phil FROM THE DARK SIDE OF THE MOULIN FRANCE.
Sampling theory tells us using fewer bits generates _quantization noise._ But quantization noise doesn't matter if you have already have significant noise in your signal; adding a little more won't make it sound worse. But honestly it's a stupid question. Skip the math and use 24 bits. This is 2018 people. As for sample rate, I doubt you need more than 44.1 kS/s (kilosamples per second). But there is an argument for using a higher rate which is that some signal sources (eg Brass) go higher than 20kHz and someone someday might want to hear that. Tape goes to about 30kHz so recording at 96 kS/s gives you somewhere north of 40kHz bandwidth - considerably more than tape. For mere mortals, 44.1 or 48 is probably fine. Hope that helps.
I randomly found this video and I regret nothing. The explanation was SOLID and made me watch through to the end. I suppose I have to subscribe now. :)
Very interesting thank you. But your analogy about bit depth and camera resolution is somewhat incorrect. Bit depth translates to dynamic range not resolution. For example an iPhone can film in 4k resolution while an Arri Alexa ( industry standard for a while) shoots "only" at 2.8k resolution. Still you don't see many hollywood movies shot on an iPhone. Its because dynamic range of an Alexa is 15 stops while newer iphones are 8 to 10 stops. at least this should explain that part :)
Good explanation for the resolution, I make a little electronics on a daily basis and if you take a line signal 0-2 v approximately then you can divide 2v/16bit and get the resolution. 2v/4 bit == 2v / 15 = 0,133v, We can see that at low volume input, the resolution is not that good. 2v/5 bit == 2v / 31 = 0,065v, 5 bit is better. but we also need a resolution on time, the faster we can take a sample the better the high frequencies are sampled. 50 Hz = 0.020 sec 500 Hz = 0.002 sec 5000 Hz = 0.2 msec 44100 Hz = 0.0227 ms 48KHz = 0.0208 ms 96KHz = 0.0104 ms Maybe it's wrong, but I think so
As I understand it the high frequencies are either sampled or aren’t. You record at 44.1 kHz because 22kHz is the extreme limit of human hearing. The extra 0.1Khz is cushion for the frequencies we want.
96db SNR is actually very impressive, it's like the signal is the sound of a train passing by, while the noise is people not conversing, but whispering!
Hi Bandrew! Enjoyed the video. Your hunch is correct about starting with better quality before you compress. I work for a well know satellite TV company and the signals flowing into our facility are high bandwidth, high quality before they hit our compression system to make sure we get a good result for customers. Recording studios record at much higher quality than we normally see in the end product. Garbage in, garbage out, Awesomeness in, pretty good out...
Someone was claiming that there was no difference between 16-bit and 24-bit in a article I read and they also included the quantization numbers? Thank you for the explanation.
You have to have recording hardware that's well guarded against electrical noise to worry about the noise floor of 16-bit. In most systems the unwanted noise is well above it. Even silence cutting filters default to far above it as detected as silence.
4:26 and now im down to 4 bit and as you can hear it’s just atrocious I highly doubt your able to understand anything im saying at all there’s just such little information here its not even useful.
Another benefit of 24bit audio is this,when you mix audio signals together such as voices and instruments, they can PARTIALLY cancel out eachother resulting in duller voices and instruments,so mixed xoices with instruments in 16bit will more sounding like 8bit,but mixed voices and instruments in 24bit will sounding more like 16bit,so that’s also the reason why 24bit sounds so clear because you will hear the starting and endong point of each sound and very smoothly.
Every time you mix an instrument in, that one track starts losing bits from the original as you mix in more tracks. It goes by powers of 2. 2 tracks, you lose 1 bit per track. With 3-4 tracks, you lose 2 bits per track
@@JoeStuffzAlt wow that’s an interesting note, so if i will record my voice at 16bit and i will myx it with 8 instruments with it,then my voice will become 8bits or less?? In such case no wonder 24bit should be better since it has more bandwide,especially with 32bit recording since in theory you have to worry about accidentally recording overamplified audio and get declipped audio,because they say that 32bit audio recordings should never suffer from declippilg because of the higher bandwide.
@@johneymute It'll be more nuanced than that. Powers of 2, so in theory it could be 13 bits, which is still a good amount and most people won't hear it. Not to mention the more audio sources going on, the less you'll be able to pick things out. You need to record around 256 layers to drop it to "8-bit per track if you manage to digitally extract it" quality. You can record softer with 24-bit and have excellent quality. With 256 times the resolution, if the waveform is a sliver on the screen, you still have a lot of data to work with. If amplifying 100x gives you a workable waveform, you still have over 16 bits of data to work with. There other factors like noise floors to worry about though
To put things into perspective, a good modern condenser microphone may support a dynamic range in excess of 130 dB (e.g. Rode NT1a), occasionally even approaching 140 dB (e.g. AKG C4000B, 137 dB(A)). That's way beyond the capabilities of 16-bit recording, even approaching the limits of 24 bit. Granted, you are not likely to have to record both a bumblebee and a jackhammer with the same settings and obtain decent results for both - but you could, assuming your preamp and ADC are up to snuff. On the playback side, human hearing struggles when confronted with more than about 70 dB all at once (even though it'll go down to ~0 dB SPL when it's quiet and accept 100+ if need be), so assuming you have a volume knob, 16 bits basically do the job just fine. It's just when you try to brute-force things and want to run a DAC directly into a power amp with some potent speakers that you'll find 96 dB don't really cut it. Related to this is why having "too little" gain on your mic preamp isn't necessarily an issue. As long as it gets the mic's internal noise well above the ADC noise floor (ideally by 10+ dB) while not adding too much of its own, that's all that's really needed. All the rest is convenience and "out of the box" usefulness of recordings, but ultimately functionally equivalent to adding gain in software. Did you know that you can determine a good estimate for the (best-case) equivalent input noise level of a *dynamic* mic entirely from its specifications? That's because its noise will be dominated by its voice coil resistance (which generally is close to nominal impedance), as all resistors generate thermal noise in a predictable and well-documented manner (see Johnson noise) - so you know your electrical noise level. The microphone's sensitivity spec then allows you to translate noise voltage back to dB SPL (as 1 Pa = 94 dB SPL). For example, a 200 ohm, -54 dBV / Pa (2 mV / 1 Pa) mic at room temperature (295 K) would have 0.26 µV of noise over a 20 kHz bandwidth, which in turn is the equivalent of 16 dB SPL. Remember, that's the best-case scenario - no amp noise, no nothing. Even so it's roughly the equivalent of a good small (0.5") condenser with a capsule typically twice the size or more. If you were to build a dynamic mic with the noise floor of a SOTA large condenser (3-5 dB(A)), it would have to be quite large with the associated very narrow directivity in the highs, not what you want in a general purpose mic. So why don't they build high efficiency microphones, you say, just like high efficiency (PA) speakers? Simple, at a given transducer size there is a tradeoff between efficiency and low-end response, and it's not like dynamic mics tend to be real bass kings to begin with (unlike condensers which can record very low as long as their buffer stage has very high input impedance). I mean, what would you expect from a 1" wideband speaker? These things just have to be very inefficient to have any bass at all, and there are very few parameters to tweak. Narrower air gap (beware of production tolerances), stronger magnet (e.g. neodymium instead of ferrite), choose impedance to make voice coil lighter and as such reduce moving mass (but don't make it so high as to get in trouble with input current noise on typical preamps)... that's about it. The "hottest" dynamic mic I can think of offhand would have to be the AT BP40, -48 dB/Pa @ 450 ohms with what I assume is a 40 mm (1.5") capsule.
At 4:04 I could hear the difference even using cheap headphones while listening with only my right ear. It's subtle, but around that mark your words sounded thinner. I believe you're right, downsampling a higher quality file will yeld better results, even for TH-cam.
@5:20 , Hi, great video, Just one point, If you get a picture with lots of pixels but with 4 bits, you still will have a blurry image! Because bits of an image does not have anything to do with its pixels. In fact, each pixel has a number of bits that tells you how is it colored.
Funny analogy to use, comparing bit depth of digital audio to the pixels of a digital camera, when a digital camera actually works with bit depth as well to form an image.
4 ปีที่แล้ว
I just fell in love with the Zoom f6 and it's 32 bit mode... I think the biggest difference is if you record or deliver audiofiles. 16bit is more than enough for delivering a well mastered audio signal but recording in 16 is pretty bad ... because if you don't compress the audio on the way in you waste even more bits for the headroom you have to leave so that you don't clip the signal ... with 32 bit you are always on the safe side ...
Nice and informative video! But a quick calculation which is highly suggestive. Not necessarily precise, but it gives a fair estimate. (I'll be using the European decimal point, which is a comma.) Say you record 16-bit 44.100 KHz audio for 10 minutes. 16 * 44.100 = 705.600 We'll divide the number by 8, since the data on your storage device is displayed in bytes. 1 byte = 8 bits 705.600 / 8 = 88.200 Now that we have the storage in bytes written per second, we'll simplify it to megabytes (MB), so MB/s instead of B/s. 88.200 / 1.000 = 88,2 Now we have the kilobytes per second (KB). 88,2 / 1.000 = 0,0882 Now we have the final result, 0,0882 MB/s. Meaning that per second recorded of 16/44.1 audio takes 0,0882 megabytes. That would mean that 10 minutes would use 52,92 MB of your storage device. Now let's take that in high-resolution recording to see if it's actually just for "pro"-use. Our recording will take place in 24-bit 192.000 KHz. 24 * 192.000 = 4.608.000 Now we could be lazy and just divide 4.608.000 with 705.600, since that is the 2 different rates of bits written per second. But we'll do the math. 4.608.000 / 8 = 576.000 Now we have the number of bytes written per second of recording. 576.000 / 1.000 = 576 Now we have the number of Kilobytes written per second. 576 / 1.000 = 0,576 0,576 is the number of Megabytes written per second of HD audio. That would mean that 10 minutes would use 345,6 MB of your storage device. Now that we have done the math and found out that on the storage side 16/44.1 versus 24/192 means 52,92MB/10min. versus 345,6MB/10min. The most popular HDD (Hard Disk Drive) is a 1 TB Western Digital Blue. Now due to the drive being divided into 1024 KB blocks by the OS (Operating System), that means we have 931 GB (Gigabytes) of storage available to us. Which is 931.000 MB. So the number of 10 minute "low" vs "high" definition audio files we have room for on a completely empty 1 TB (1.000 GB) drive, is: Low (16/44.1): 931.000 / 52,92 = 17.592,59 High (24/192): 931.000 / 345,6 = 2.693,87 This means that you can store about 17,5 thousand 10 minute files of "LD" and you can store about 2,5 thousand 10 minute files of "HD". To put that into perspective, let's calculate how long that is in number of days. (24 * 60) / 10 = 144 We can play or record a 10 minute audio file, or any 10 minute file, 144 times in a day. 17.592,59 / 144 = 122,17 2.693,87 / 144 = 18,71 This means that it would take you, me or anyone 122,17 days to fill up an entire 1TB drive with 10 minute "LD" audio files, if we imagine that the computer automatically splits the file every 10th minute. It also means that it would take you 18,71 days to fill up an entire 1TB drive with 10 minute "HD" audio files. Now that is if you were to leave your computer alone for that amount of time, recording non-stop to an empty 1TB drive. So to summarize and conclude, aka. *TL;DR*: It would take you more than 18 days of non-stop recording to fill up a 1TB drive with 24-bit 192.000 KHz files. That's not a normal use-case. Also, the average storage space per storage device naturally goes up, as we continue to develop them. Soon, 4TB or more will be the norm for external HDDs. So all this basically leads me to conclude that there is absolutely *no reason* to record in lesser quality, if your audio interface allows you to record in higher quality. This calculation is entirely based on the situation that you record and store files... That means that if you use VOiP services like Discord or Skype, you don't use up space on your computer (other than perhaps cache, but I digress). So again: *Use the highest quality setting available on your audio interface*.
no, if you do not hear the difference between 8 vs.16 bits depth, your speakers degrade the sound quality. You need to get better HQ speakers, not active PC or bluetooth speakers
I might be naive, but I listen to basically all techno music. To my understanding, synthesized techno music starts and is digital, so there is no ADC needed, correct? It's more about getting the same quality of the master out of my phone/player into my car head unit, through it, out, into the amps, and then speakers. I would presume that the head unit sends an analog signal out to the amps, thus the head unit DAC quality is pretty important, but I could be wrong. --I had to edit. I thought that 24 bit is used because there is no middle, e.g., 18 bit, 20 bit. It has to be in increments of 8, thus 16 bit to 24 bit to 32 bit. Also, as the bit depth goes up, the noise floor changes ... gets moved, which ultimately changes the sound perceived. CD quality was basically invented by Sony, and 16 bit 44.1khz was determined because it resulted in a high cut off of just about 20kHz, which most humans don't even hear above 16kHz. Therefore, we can't hear any difference. 16 bit proved to end in a quality that could not be discerned from prior media types' quality. Most people can't hear the difference between 16 bit and 24 bit. ---But, I'm still interested if anyone has any input about techno music specifically, as it's synthesized.
Best explanation, too many maths guys on the internet that don't bother mentioning the practical use of greater sound floor and head room then mixing down, sampling and trying to edit out redundant noise or artefacts from home studio recordings. At home you don't usually have the best sound proofing so greater bit depth is that sharpener for every home producers biggest ally...the mix
I seen another popular TH-camr he recorded in floating 32 bit with the Zoom F6 and he demonstrated lowering clipped audio and it sounded great, then he took audio that was recorded way too low and he raised it about 40dB and once again sounded great. He said that is because he recorded in the 32 bit. When you demonstrated the 4 bit it sounded like clipping audio. I guess I'll stop recording in 16 bit and go 24 bit myself.
I know this is a late question, but what about people who just want to do live podcasting? Will 24bit give any real advantage over 16bit? And what about if someone connects a 16bit mixer into a 24bit audio interface, will it record at 16bit or 24bit?
Thanks for the video and your knowledge on bit rate! I produce music and work sound production for a production company so I wanted to review the actual definition of bit rate.
float means you can clip past 0dbfs and signal can still be turned down without clipping effect. it's useful to render rough mixes in float 32 rather than old and outdated "-6db of headroom" because end result is the same, you can turn down audio for final mix. the -6db of headroom is old safety measure for 24-bit renders before 32 float was widespread
I think I can hear the difference between 16 and 24 bits in your voice when you say the "s", "p", "f"... but in a mix I probably can't hear the difference
How record in 24 bit in fl studio? My interface is set to 24 bit / 44.1Khz. But my wav files still say 16 bit. I can set the wav to 32 bit float, but that is just how wav format is written to disk. How do I know that the input signal is actually 24 bit?
Higher bit depths mean there's less rounding errors in palcing the xy corrdinates.. the rounding errors end up being noise. Lower bit will result in more noise, but honestly the noise floor at 16 bit is still good enough to where people really won't hear the noise unless the volume is so high in quiet parts it causes hearing damage. 24 bit levels are a bit overkill and use up too much drive space if storing. I do blind tests all the time with friends, they can never tell the difference between 44khz/16bit and 44khz/32bit. Yes.. highr but depth is better.. bit overkill
14 bit sounds really good and more analogue like for some unknown reason; 48 bit sounds clearer and sharper but the higher the bit rate, the more of the soul of the music is removed (not sure why, but this is my experience working in professional audio since 1988).
@Delatronics Fun Fact: The earliest models of CD Players that were first released by Sony and others in 1984 actually had 14-bit D/A converter chips. It took Sony about a year to produce true 16-bit D/A chips in mass quantities for the consumer market because they were so advanced and difficult to produce at that time, hence too expensive for the consumer market. I still have a working Sony CDP-30 that my dad gave me for my birthday in 1984. Sounds great even now, haha.
Audio bit depth in no way relates to video resolution, but is directly related to pixel bit depth, with 16 bit pixels have far fewer possible colors. However, the human visual system can't distinguish more colors so 16 bit looks identical to 1000+ bit (except for a handful of females with quad cones). Similarly, except for the 96 dB noise floor (we can detect up to 120) there's no audible difference between 16 and 1000+ bit. The only reason 24 bit is superior for capture and mixing is additive losses accumulate. But after mixing outputting to 16bit 44100 hz is indistinguishable from 24bit 48000 hz+ except for the higher noise floor which is virtually silent. Also, sampling rate does relate to framerate, but at best we can only hear up to 20khz, and exactly 2 samples are needed for full capture (mathematically proven and confirmed by hearing tests), so sampling above 40khz in absolutely no way improves audio quality for anyone.
"Monty" Montgomery who created OGG container and Vorbis audio codec made an excellent technical explanation over at xiph.org/video/vid2.shtml The chapter "bit-depth" and "dither" is relevant here. TL;DW, The higher the bit depth, the lower the noise and the higher the dynamic range. Dynamic range could be expanded within the same bit depth with dithering. From another article here at people.xiph.org/~xiphmont/demo/neil-young.html 16-bit is enough for hearing range. For recording, 24-bit's lower noise could give an extra headroom for boosting the signal in post and apply effects after effects in post.
The actual problem is that we have to avoid digital clipping. In order to be on the safe side, we don't use the full dynamic range, leaving some headroom for unexpected peaks. So we use, let's say 14 bits out of the available 16 bits, or 20 bits out of 24. Since the hearing dynamic is about 16 bits, there is some loss if you boost your 14 bits to 16, no loss if you reduce 20 to 16. If you can control your recording dynamic to 16 bits precisely, using analog compressor/limiter for example, you can stick to 16 bits.
Very good video and I record classical music in 24 bit and in low frequencies you can hear the difference. It´s very very small difference and it happen in some little points of the music, but exist. With attention on the bass, like me, hehehe... you can "feel" the difference of 16 bit to 24 bit. In classical music and some wen age musics, 24 bit is better. Congratulation for your video.
Fantastic breakdown...love your explanations as i get to pass it on to clients and friends - on a side note some people have even suggested 32 bit JUST IN CASE you clip in recordings it will still capture more sound information and you can bring the amptitude down and capture all the sound - theres one big youtuber who made a video on this and i thought although somewhat true it was overkill and anyone driving things that high are in a small minority but i see a few deciding to do 32 just becoz - although the file size increase really starts pushing practicality :) - just a side note, again great vid!
If you have a 32-bit converters and an analog input that can handle high levels. Otherwise recording in 32-bit FP on a 24 bit interface is just a waste of hard drive space and RAM.
If you have a 32-bit converters and an analog input that can handle high levels. Otherwise recording in 32-bit FP on a 24 bit interface is just a waste of hard drive space and RAM.
I’m not sure that 1080p vs 4K video is a good comparison, since that’s clearly noticeable to most people. 16bit vs 24bit is more like watching 1080p vs 1080p, taking a placebo pill, and telling everyone that you now have special eyes which can see the difference.
I work with Voice Over artists, and when they record at 16dB with low gain I get tons of background noise, it is a gamble recording at 16bit, you would need to have a stable good gain.... or just record at 24bit, which is what I ask people to do.
Try this. Generate a 4 kHz signal in Cool Edit or Audition using 96k/24 bit. Downsample it twice, one file to 96K/16 and another to 44.1K/24. Compare them yourself. See that the 24 bit depth does nothing for the signal resolution. On frequency analysis you can see that It only reduces the noise floor. The 24 vs 16 bit graph in the movie is only a drawing and is incorrect. So then why do editors and producers use 24 bit - to keep the noise floor low. They are processing music. You are not. So if you have a limited disc space, choose 96K over 24 bit. That also gets rid of the brick wall AD converter, something you can hear. If you have disc space, a good system and good hearing use 192/24. Your source material may vary.
I'm a musician, I work really hard on everything. I've got years of experience and I'm still learning. Up until now I have recorded in 16bit 44,1 and I feel like crying, all the work I've done and only to learn I've been missing out on production benefits... :(
Joel Firey to be fair, 16, 44.1 was the standard for CDs back in the 90's and 00's. So worst case scenario your music could potentially sound as low quality as anything on a professionally produced CD.
When you laughed was it because it was a tedious topic, or was it because it you were excited? Pretty sure we all got excited when talking about extra variables. Yes - Using 24bit before spotify and youtube compresses is beneficial. In my neighborhood I'm the only one who notices, and due to my "dontgiveafuggifitclipsinuteromilkit" attitude - I've definitely taken full advantage of those extra frequencies allotted for filling waves. When you go back to 16, and try to do what you did in 24...that's when you notice and go..."Oh helllllll no!"
@ReaktorLeak Way more powerful. Though for online uploads, kinda pointless, I agree with dudeman...having the highest possible quality before conversion definitely matters.
@ReaktorLeak I'm not a premium member on reverbnation, so my upload limit has been 8 megabytes. Only recently did I realize that I'd have to doctor before uploading overamped 24bit Waves as MP3's online. Your CD copy of In Utero can clip by 2.4 decibels...in your house, but not converted into as low as 128 or 96kbps on reverbnation. (So much garbage to fix or replace) Alot of the the time i would cut large projects into parts to get the highest possible bitrate with only the shortest 3 minute tracks achieving 320. 24bit. Great for home, but rattles the average listeners devices to sh!t. Just recently remastered some of my match-ups (mashups😡) and uploaded them on mixcloud to take advantage of 320 mp3, and found they now support 16bit wave. Very exciting. When I upload to spotify now via distrokid...I'll still upload 24bit...but...I'll make sure it sounds good as an mp3 as well, since all sites stream at 16bit. My system can handle it, and I gotta listen to this stuff forever, not just once casually so I'll always have my overamped waves. Haha - Hate neutering and converting them. Most can't tell a difference, dare I say it you'd practically have to have both waves up on your screen switching off between the two to notice. www.mixcloud.com/Flow_Pdisco/dj-analysis-gamechangers/
had a few Years ago the M-Audio Delta 1010LT and i loved, then my pc got stolen! NOW i use for a few Months again a M-Audio Delta 1010LT PCI Card, its superb for non USB audiosignalstudio also for a good Audio Homestudio without any latencyroblems!!
Actually i feel huge difference, like dynamics from your voice disappeared at 16bit and it was less open sound and not crisp at all.. I guess i bought a good pair of cheap Sony headphones 😉
Good Stuff. Can you do a 24-bit USB mic comparison: Samson G-Track Pro vs. Apogee Mic plus vs. Shure MV88 iOS I believe you've already done the Samson individually, but not the Apogee. ps. I have noticed watching your videos that XLR mics on the U-Phoria box don't sound as crisp as the XLR mics on basically every other USB plugin you've tried. You may want to do a comparison of all the box packages... TASCAM, FocusRite & Steinberg
TH-cam uses AAC for sound . This compression is very lossy I think it's on the same level as MP3 Most recordings are in WAV or FLAC I personally record in FLAC . On TH-cam due to the heavy compression it's pretty pointless IMO to record 24-bit Spotify uses Ogg Vorbis which is the best compression so far . The losses are minimum and it sounds much better . So if you're aiming for Spotify you have benefits to record 24-bit I may be wrong . I don't know much . But I hope I helped
So if an music album is remastered to 24-bit, does it make the album better? Is it worth buying? I have the original CDs, so is it any differences? Is it any better? Is it worth it? Is it like 5.1 surround sound remixes? The Beatles’ Sgt. Pepper’s, White Album & Abbey Road was remastered to 5.1 surround sound. What does 24-bit means? Is it 5.1 surround sound remixes? The Dutch prog rock band Focus is also releasing a 50th Anniversary box set now with remastered material. It says it is a «24-bit remaster». Is it worth it when I already have the original CD albums?
24-bits just means you have 8 extra bits, and every time you add a bit , the audio resolution doubles. 24-bit has 256 times more resolution than 16-bit. Honestly, I found it depends on the album. Sometimes it's not perceivable, other times, I can really hear it in the vocals. I also think it depends on how many instruments are being recorded.
*Everything you need to know for audio playback and production...*
TLDR: 16 bits is enough for playback, and 24 bits is only helpful for production.
Small correction: Bit-depth doesn't really get you "resolution".
*1) Frequency gives you resolution for audio.*
As far as frequency goes, 44-48kHz is enough because human hearing doesn't really go beyond 20-20k, and 44.1 gives you up to 22kHz accurately according to the Nyquist formula.
44.1kHz is the CD audio standard.
48kHz is DVD audio standard.
*2) Bit depth gives you dynamic range.*
6dB of dynamic range per bit.
This means 16-bit has 96dB DR and can easily cover from 20dB (silent room) to 116dB (front row at a rock concert, almost hearing damage).
24-bit has 144dB DR and can cover the range between a silent room and a jet engine.
*3) Higher bit-depth lowers the effect of noise while mixing audio.*
When you process audio, every operation adds a little bit of noise.
With 16-bit, you have 65k steps.
With 24-bit, you have 16M steps.
A random error shift of 10-100 steps will be ever-so-slightly more significant if you have less steps to scale it.
The absolute range means you have more room for error.
Appreciate the correction.
Well, you can think 24bit as higher "vertical" or signal amplitude resolution.
Like 1920x1080 full hd being 1920 the sampling frequency, and 1080 the bit depth.
Translating to an wave signal, higher bit depth means bigger headroom, snr and blablabla =D
Ah that's a GREAT way to put it. Vertical resolution vs Horizontal resolution. Dang it. I wish I had thought about putting it that way in the video. =) Thanks for the better explanation.
Victor Amicci
To be more precise... bit-depth is more like the intensity. I've already explained how frequency is the resolution; the important thing is that sound isn't measured in 2D space, it's measured in frequency. The sampling frequency lets you "see" any frequencies that are 50% or less. So let's say 1920x1080 is like being able to see up to 20kHz (functionally enough), and 1280x720 is seeing up to 16kHz (practically enough), etc.
Since bit-depth is intensity, think of how many colors can be displayed in a given sample (either a pixel, or a timecode). You may have heard of 24-bit color, which uses 8 bits for each of 3 color channels (0-255). The absolute dynamic range is the difference between the highest and lowest values you can communicate. In other words... what's 0% brightness? What's 100% brightness? How many steps are in between? The standard is to put 256 steps for each color, or 65,536 for the human-audible range.
Sure, you can use more than 256 steps for color; HDR displays do this, and there are definitely more than 16 million colors. But there's a diminishing point of return. For our rods and cones, we've got a bit of room left to grow, but for the hairs in our ear, we've already hit the practical limit. We can resolve as far down as we can differentiate two different sounds as unique, and we can cover the range between a silent room and a rock concert. We don't really need to reach jet-engine levels for playback; it's just useful to have more room for error if we make mistakes while doing the math. We're going to divide by 256 anyway when we want to listen to it.
Here's a picture showing what bit depth is like for pictures: www.azooptics.com/images/Article_Images/ImageForArticle_1151(1).jpg
And for audio: vignette.wikia.nocookie.net/digital-audio/images/f/f7/8_16_bit_depth.gif/revision/latest?cb=20140415224229
4:27 = My teammate mic on discord
Lol
More like my zoom classes
@@RocKnMetaL97, more like the astronauts in space xD
This video taught me McDonalds drive thru uses 4 bit audio....
😂
@Preston Tuttle 🤣👍
😂
Lmaoooo 😂
😂😂😂
l record in 32-bit float, at 192Khz WAV just because l hate having all of that extra hard drive space.
how does that thing even power on?
try 384Khz
😂😂 that was good
sorry i only talk to people who record at 128 bit
@Victrola 1926 wow are you not special, I record on a VHS and that is way more superior.
Thanks for the video! I’ve never realized the actual difference so clear, before I saw picture quality and numbers difference. In my understanding: record in 24 bits for better processing (noise reduction, compression, EQ, mixing) and publish in 16 because no one would ever notice the difference on TH-cam or in a typical podcast.
I am now 64yrs..A student Of you Sir...That 16 bit to 4 bits example with your own voice is Awesome .👏👏...With regards- HRR.
Man, this is one the best explanation that I've ever heard about this topic. You gotta be an amazing teacher. Well done.
Oh man, sorry to say I was laughing when I heard your 4 bit sound. By the way thanks for teaching me this bit depth.
Fucking same🤣🤣
4 bits is like 16 possible values, pretty funny indeed
Bro, I understood you in 4 bit.
You are very familiar with all those NASA recordings?
Today’s ADC units are far better than those from the past. A bit depth of 16 and a sample rate of 44.1 kHz is more than sufficient for the human cochlea to hear amplitudes and frequencies as accurately as our hearing apparatus is capable of. This is true of a straightforward record and playback. However, when we edit digital audio in any way, we introduce new quantization errors that build up in the lower bits the more we edit. Enough so that we can start to hear the artifacts introduced by editing. So 24 bit is recommended for professional recordings that will be edited so that the artifacts generated in the lower bits are rendered well nigh insignificant because a smaller portion of overall bits are corrupted by S/E errors due to editing. Therefore if the editing after recording is minimal, then 16 bit is all you really need.
This is the most important part! 16bit 44.1kHz really truly is indistinguishable from 24bit or 32bit 48kHz. But if you're messing with the recording a bunch, it helps reduce noise! Another vital point; most older hardware and even some more modern programs can only use 16bit samples, so there's that.
Without a shadow of a doubt, this is the best Chanel I ve subscribed to, u are amazing bro, keep up the good work
Higher bit depth means greater precision in mapping an analogue voltage to a corresponding digital value, and allows for a greater signal to noise ratio.
16 bit (2 bytes) means you have 65,536 possible digital values available for representing the analogue voltage.
24 bit (3 bytes) means you have 16,777,216 possible digital values for representing the analogue voltage.
So, 24 bit can more precisely record what the analogue voltage was.
It's important to know that the digital values are evenly spread out between maximum voltage and minimum voltage.
This is important when you're dealing with very low voltage levels - there are fewer digital values between zero and the value closest to the one that represents the voltage if you're using a lower bit depth. That means any editing of the resulting digital audio has fewer values to change between. So it's always better to capture in the highest bit depth possible, and do any processing at an even higher bit depth, before mastering down to the bit depth needed for the final product.
16 bit allows for a maximum signal to noise ratio of 96dB, and 24 bit allows for up to 144 dB s/n ratio.
Very few audio music recordings are made that have a dynamic range that cannot easily be accommodated within a 96dB noise floor, and almost no commercial music recordings have that sort of dynamic range.
Quantization is the process of mapping the voltage to a specific digital value. Dithering is a way of doing the quantization that reduces digital artefacts caused by quantization.
Bit Depth is about digital accuracy of the sampled analogue signal. Sample rate affects maximum frequency that can be digitally represented. Sample rate is a different matter that has it's own considerations that need to be understood.
bro I want to thank you for giving me exactly what I needed within the first minute of your video the rest of it I skimmed but the first minute was the info that I wanted. Thank you!
Great analogies, simple explanations and easy to follow. You hit it out of the park with this one.
For the picture analogy, I think it might be closer to a picture format that can create 65,000 colors vs a picture format with 16-million different colors. But I'm no expert
Bro...that was awesome. I had so many questions answered but at the same time now have new ones. Great video, had me hooked from start to finish!
What kinds of questions did it raise?
Podcastage so a lot of different engineers that I send stuff off to are all so picky. Some people 16-441 mono some want 24-441/48 mono wav. So I guess it comes down to preference. But it’s sucks you can’t work with the sample rates back and forth with projects. For me, pro tools. If I created a session in 48, protocols or Windows or whatever will not open up the 44 project and of it does the system gets confused and all of sudden everything plays in slowmotion...including desktop audio.
Yeah...questions. I learned all of this by serious trial and error. So I guess my main question is that when working with projects or systemsl audio drivers...why don’t they play well with each other and why can’t you jump back and forth we between them, with in them at times.
Its not the file itself, its the actual session. For example, if a created a protools session and what the devices are set to 48 or 44. In 48, it wont open on another system that is 44, i have to go into the option of the DAW and change it before opening the project. And if im using the same DAW as an audio card for my PC for my desktop audio and they dont match up, then i get the slowdown effect. YEah i know its confusing, but its real, and i found out the hard way through numerous trial and error
so this error would happen specifically when i was using my mixer as my Yamaha MG12XU as my sound card and ProTools DAW. My windows system would default to 48 but protools was set to 44. Once protools opened up it would try to convert the system that was 48 to 44 and every sound from the PC would play in slow motion, even the boot up screen. In the the end, i had to give up using the mixer as an all in one SoundCard and DAW..it just wasnt worth the constant hassle and uninstalling and reinstalling the drivers to fix the problem when ever it would come up
I think you’re wrong about bit depth. Sample rate resolution is like 1080 vs 4k tv resolution. But Bit depth is about the dynamic range of the sound; i.e. how low level a sound can be recorded at compared to the highest level.
I was thinking the exact same thing. I'm just getting into this subject, but my understanding and what I've read so far that its a lot more about dynamic range than a better sounding initial recording.
Good, informative video. One of the best ones done on this - would be cool to see more on why we need 24 bit vs 16 bit for recording music vs podcasts though.
Hi, I enjoyed your territorial on a BIT Recording, I have finally found out why some music sounds better on recording on TH-cam than others,
Just recently I have been listening to a lot of music at 70-year-old dude retired enjoy listening to music just recently I rebuilt my transmission line loudspeaker cabinets , I’ve been listening to a lot of music, Classical to heavy metal , Play the same music recorded by different people obviously on different devices and you can notice when somebody record something in 16 to 24 BIT , It is surprising how much difference it can be ruin a piece of music, I hadn’t realised, Until now, Great territorial found something at my age of life, Phil FROM THE DARK SIDE OF THE MOULIN FRANCE.
4:27 So this is what Microsoft did with the Xbox live parties/mics.
WOW man tnx Im going to 4 bit no doubt after this! I didn't know I had a hidden HD audio feature in my PC. TNX
Sampling theory tells us using fewer bits generates _quantization noise._ But quantization noise doesn't matter if you have already have significant noise in your signal; adding a little more won't make it sound worse. But honestly it's a stupid question. Skip the math and use 24 bits. This is 2018 people.
As for sample rate, I doubt you need more than 44.1 kS/s (kilosamples per second). But there is an argument for using a higher rate which is that some signal sources (eg Brass) go higher than 20kHz and someone someday might want to hear that. Tape goes to about 30kHz so recording at 96 kS/s gives you somewhere north of 40kHz bandwidth - considerably more than tape. For mere mortals, 44.1 or 48 is probably fine.
Hope that helps.
Mike Page great insight! I appreciate you sharing that information about brass and tape response. :)
I randomly found this video and I regret nothing. The explanation was SOLID and made me watch through to the end. I suppose I have to subscribe now. :)
Very interesting thank you. But your analogy about bit depth and camera resolution is somewhat incorrect. Bit depth translates to dynamic range not resolution. For example an iPhone can film in 4k resolution while an Arri Alexa ( industry standard for a while) shoots "only" at 2.8k resolution. Still you don't see many hollywood movies shot on an iPhone. Its because dynamic range of an Alexa is 15 stops while newer iphones are 8 to 10 stops. at least this should explain that part :)
Ive heard a lot of people try to make bit depth only about SNR, but I think youre right that it also does impact accuracy/audio quality
Good explanation for the resolution, I make a little electronics on a daily basis and if you take a line signal 0-2 v approximately then you can divide 2v/16bit and get the resolution.
2v/4 bit == 2v / 15 = 0,133v, We can see that at low volume input, the resolution is not that good.
2v/5 bit == 2v / 31 = 0,065v, 5 bit is better.
but we also need a resolution on time, the faster we can take a sample the better the high frequencies are sampled.
50 Hz = 0.020 sec
500 Hz = 0.002 sec
5000 Hz = 0.2 msec
44100 Hz = 0.0227 ms
48KHz = 0.0208 ms
96KHz = 0.0104 ms
Maybe it's wrong, but I think so
As I understand it the high frequencies are either sampled or aren’t. You record at 44.1 kHz because 22kHz is the extreme limit of human hearing. The extra 0.1Khz is cushion for the frequencies we want.
96db SNR is actually very impressive, it's like the signal is the sound of a train passing by, while the noise is people not conversing, but whispering!
Hi Bandrew! Enjoyed the video. Your hunch is correct about starting with better quality before you compress. I work for a well know satellite TV company and the signals flowing into our facility are high bandwidth, high quality before they hit our compression system to make sure we get a good result for customers. Recording studios record at much higher quality than we normally see in the end product. Garbage in, garbage out, Awesomeness in, pretty good out...
I sort of kind of maybe understood this stuff, until now. Finally! As always, thanks for the work and genius that goes into these videos.
Someone was claiming that there was no difference between 16-bit and 24-bit in a article I read and they also included the quantization numbers? Thank you for the explanation.
You have to have recording hardware that's well guarded against electrical noise to worry about the noise floor of 16-bit. In most systems the unwanted noise is well above it. Even silence cutting filters default to far above it as detected as silence.
Video man! I've been wondering this for a while, love your channel and keep it up!
I've decided to use auto-tune in every podcast I do for that warm "I am a robot here to kill humanity and take over world" sound everyone wants!
4:26 and now im down to 4 bit and as you can hear it’s just atrocious I highly doubt your able to understand anything im saying at all there’s just such little information here its not even useful.
Thanks! Your video is so clear it made someone who had no idea about all this suddenly get it
I just wanted to tune my microphone for discord and teamspeak, now I am here
Another benefit of 24bit audio is this,when you mix audio signals together such as voices and instruments, they can PARTIALLY cancel out eachother resulting in duller voices and instruments,so mixed xoices with instruments in 16bit will more sounding like 8bit,but mixed voices and instruments in 24bit will sounding more like 16bit,so that’s also the reason why 24bit sounds so clear because you will hear the starting and endong point of each sound and very smoothly.
Every time you mix an instrument in, that one track starts losing bits from the original as you mix in more tracks. It goes by powers of 2. 2 tracks, you lose 1 bit per track. With 3-4 tracks, you lose 2 bits per track
@@JoeStuffzAlt wow that’s an interesting note, so if i will record my voice at 16bit and i will myx it with 8 instruments with it,then my voice will become 8bits or less??
In such case no wonder 24bit should be better since it has more bandwide,especially with 32bit recording since in theory you have to worry about accidentally recording overamplified audio and get declipped audio,because they say that 32bit audio recordings should never suffer from declippilg because of the higher bandwide.
@@johneymute It'll be more nuanced than that. Powers of 2, so in theory it could be 13 bits, which is still a good amount and most people won't hear it. Not to mention the more audio sources going on, the less you'll be able to pick things out. You need to record around 256 layers to drop it to "8-bit per track if you manage to digitally extract it" quality.
You can record softer with 24-bit and have excellent quality. With 256 times the resolution, if the waveform is a sliver on the screen, you still have a lot of data to work with. If amplifying 100x gives you a workable waveform, you still have over 16 bits of data to work with. There other factors like noise floors to worry about though
To put things into perspective, a good modern condenser microphone may support a dynamic range in excess of 130 dB (e.g. Rode NT1a), occasionally even approaching 140 dB (e.g. AKG C4000B, 137 dB(A)). That's way beyond the capabilities of 16-bit recording, even approaching the limits of 24 bit. Granted, you are not likely to have to record both a bumblebee and a jackhammer with the same settings and obtain decent results for both - but you could, assuming your preamp and ADC are up to snuff.
On the playback side, human hearing struggles when confronted with more than about 70 dB all at once (even though it'll go down to ~0 dB SPL when it's quiet and accept 100+ if need be), so assuming you have a volume knob, 16 bits basically do the job just fine. It's just when you try to brute-force things and want to run a DAC directly into a power amp with some potent speakers that you'll find 96 dB don't really cut it.
Related to this is why having "too little" gain on your mic preamp isn't necessarily an issue. As long as it gets the mic's internal noise well above the ADC noise floor (ideally by 10+ dB) while not adding too much of its own, that's all that's really needed. All the rest is convenience and "out of the box" usefulness of recordings, but ultimately functionally equivalent to adding gain in software.
Did you know that you can determine a good estimate for the (best-case) equivalent input noise level of a *dynamic* mic entirely from its specifications? That's because its noise will be dominated by its voice coil resistance (which generally is close to nominal impedance), as all resistors generate thermal noise in a predictable and well-documented manner (see Johnson noise) - so you know your electrical noise level. The microphone's sensitivity spec then allows you to translate noise voltage back to dB SPL (as 1 Pa = 94 dB SPL). For example, a 200 ohm, -54 dBV / Pa (2 mV / 1 Pa) mic at room temperature (295 K) would have 0.26 µV of noise over a 20 kHz bandwidth, which in turn is the equivalent of 16 dB SPL.
Remember, that's the best-case scenario - no amp noise, no nothing. Even so it's roughly the equivalent of a good small (0.5") condenser with a capsule typically twice the size or more. If you were to build a dynamic mic with the noise floor of a SOTA large condenser (3-5 dB(A)), it would have to be quite large with the associated very narrow directivity in the highs, not what you want in a general purpose mic.
So why don't they build high efficiency microphones, you say, just like high efficiency (PA) speakers? Simple, at a given transducer size there is a tradeoff between efficiency and low-end response, and it's not like dynamic mics tend to be real bass kings to begin with (unlike condensers which can record very low as long as their buffer stage has very high input impedance). I mean, what would you expect from a 1" wideband speaker? These things just have to be very inefficient to have any bass at all, and there are very few parameters to tweak. Narrower air gap (beware of production tolerances), stronger magnet (e.g. neodymium instead of ferrite), choose impedance to make voice coil lighter and as such reduce moving mass (but don't make it so high as to get in trouble with input current noise on typical preamps)... that's about it. The "hottest" dynamic mic I can think of offhand would have to be the AT BP40, -48 dB/Pa @ 450 ohms with what I assume is a 40 mm (1.5") capsule.
i'm guessing it helps most notably when you're compressing and EQ'ing the audio as well.
At 4:04 I could hear the difference even using cheap headphones while listening with only my right ear. It's subtle, but around that mark your words sounded thinner. I believe you're right, downsampling a higher quality file will yeld better results, even for TH-cam.
makes sense to me as an engineer. great video. much better than all the web pages out there
I love the last part of your summary! nice one thanks
Awesome explanation. It's great you had all those examples, especially when you recorded your voice with 4-bit.
@5:20 , Hi, great video, Just one point, If you get a picture with lots of pixels but with 4 bits, you still will have a blurry image! Because bits of an image does not have anything to do with its pixels. In fact, each pixel has a number of bits that tells you how is it colored.
Funny analogy to use, comparing bit depth of digital audio to the pixels of a digital camera, when a digital camera actually works with bit depth as well to form an image.
I just fell in love with the Zoom f6 and it's 32 bit mode... I think the biggest difference is if you record or deliver audiofiles. 16bit is more than enough for delivering a well mastered audio signal but recording in 16 is pretty bad ... because if you don't compress the audio on the way in you waste even more bits for the headroom you have to leave so that you don't clip the signal ... with 32 bit you are always on the safe side ...
Only vid that made any sense about the matter
THANK YOU 🤝
Nice and informative video!
But a quick calculation which is highly suggestive. Not necessarily precise, but it gives a fair estimate.
(I'll be using the European decimal point, which is a comma.)
Say you record 16-bit 44.100 KHz audio for 10 minutes.
16 * 44.100 = 705.600
We'll divide the number by 8, since the data on your storage device is displayed in bytes. 1 byte = 8 bits
705.600 / 8 = 88.200
Now that we have the storage in bytes written per second, we'll simplify it to megabytes (MB), so MB/s instead of B/s.
88.200 / 1.000 = 88,2
Now we have the kilobytes per second (KB).
88,2 / 1.000 = 0,0882
Now we have the final result, 0,0882 MB/s. Meaning that per second recorded of 16/44.1 audio takes 0,0882 megabytes.
That would mean that 10 minutes would use 52,92 MB of your storage device.
Now let's take that in high-resolution recording to see if it's actually just for "pro"-use.
Our recording will take place in 24-bit 192.000 KHz.
24 * 192.000 = 4.608.000
Now we could be lazy and just divide 4.608.000 with 705.600, since that is the 2 different rates of bits written per second. But we'll do the math.
4.608.000 / 8 = 576.000
Now we have the number of bytes written per second of recording.
576.000 / 1.000 = 576
Now we have the number of Kilobytes written per second.
576 / 1.000 = 0,576
0,576 is the number of Megabytes written per second of HD audio.
That would mean that 10 minutes would use 345,6 MB of your storage device.
Now that we have done the math and found out that on the storage side 16/44.1 versus 24/192 means 52,92MB/10min. versus 345,6MB/10min.
The most popular HDD (Hard Disk Drive) is a 1 TB Western Digital Blue. Now due to the drive being divided into 1024 KB blocks by the OS (Operating System), that means we have 931 GB (Gigabytes) of storage available to us. Which is 931.000 MB.
So the number of 10 minute "low" vs "high" definition audio files we have room for on a completely empty 1 TB (1.000 GB) drive, is:
Low (16/44.1): 931.000 / 52,92 = 17.592,59
High (24/192): 931.000 / 345,6 = 2.693,87
This means that you can store about 17,5 thousand 10 minute files of "LD" and you can store about 2,5 thousand 10 minute files of "HD".
To put that into perspective, let's calculate how long that is in number of days.
(24 * 60) / 10 = 144
We can play or record a 10 minute audio file, or any 10 minute file, 144 times in a day.
17.592,59 / 144 = 122,17
2.693,87 / 144 = 18,71
This means that it would take you, me or anyone 122,17 days to fill up an entire 1TB drive with 10 minute "LD" audio files, if we imagine that the computer automatically splits the file every 10th minute.
It also means that it would take you 18,71 days to fill up an entire 1TB drive with 10 minute "HD" audio files.
Now that is if you were to leave your computer alone for that amount of time, recording non-stop to an empty 1TB drive.
So to summarize and conclude, aka. *TL;DR*:
It would take you more than 18 days of non-stop recording to fill up a 1TB drive with 24-bit 192.000 KHz files. That's not a normal use-case.
Also, the average storage space per storage device naturally goes up, as we continue to develop them. Soon, 4TB or more will be the norm for external HDDs.
So all this basically leads me to conclude that there is absolutely *no reason* to record in lesser quality, if your audio interface allows you to record in higher quality.
This calculation is entirely based on the situation that you record and store files... That means that if you use VOiP services like Discord or Skype, you don't use up space on your computer (other than perhaps cache, but I digress). So again: *Use the highest quality setting available on your audio interface*.
I said a quick -calculation-, not a quick -explanation-.
I'd also like to add, that this is if you use the maximum "resolution" all the time.
So realistically, you'd use even less.
What’s 32 bit float? Can human ears hear the difference between 16, 24 or 32? Tks
8 bit seems OK, actually. Could my speakers or youtube be improving the sound automatically?
no, if you do not hear the difference between 8 vs.16 bits depth, your speakers degrade the sound quality. You need to get better HQ speakers, not active PC or bluetooth speakers
@@sevcaczech5961 So you're implying that active PC or bluetooth speakers are always bad / not good enough? Let me tell you: they aren't.
I might be naive, but I listen to basically all techno music. To my understanding, synthesized techno music starts and is digital, so there is no ADC needed, correct?
It's more about getting the same quality of the master out of my phone/player into my car head unit, through it, out, into the amps, and then speakers. I would presume that the head unit sends an analog signal out to the amps, thus the head unit DAC quality is pretty important, but I could be wrong.
--I had to edit. I thought that 24 bit is used because there is no middle, e.g., 18 bit, 20 bit. It has to be in increments of 8, thus 16 bit to 24 bit to 32 bit. Also, as the bit depth goes up, the noise floor changes ... gets moved, which ultimately changes the sound perceived.
CD quality was basically invented by Sony, and 16 bit 44.1khz was determined because it resulted in a high cut off of just about 20kHz, which most humans don't even hear above 16kHz. Therefore, we can't hear any difference. 16 bit proved to end in a quality that could not be discerned from prior media types' quality. Most people can't hear the difference between 16 bit and 24 bit.
---But, I'm still interested if anyone has any input about techno music specifically, as it's synthesized.
Best explanation, too many maths guys on the internet that don't bother mentioning the practical use of greater sound floor and head room then mixing down, sampling and trying to edit out redundant noise or artefacts from home studio recordings. At home you don't usually have the best sound proofing so greater bit depth is that sharpener for every home producers biggest ally...the mix
I seen another popular TH-camr he recorded in floating 32 bit with the Zoom F6 and he demonstrated lowering clipped audio and it sounded great, then he took audio that was recorded way too low and he raised it about 40dB and once again sounded great. He said that is because he recorded in the 32 bit. When you demonstrated the 4 bit it sounded like clipping audio. I guess I'll stop recording in 16 bit and go 24 bit myself.
Excellent, helped me to completed understand the entire concept. Will pass it on to fellow classmates studying audio production. Thanks
I know this is a late question, but what about people who just want to do live podcasting? Will 24bit give any real advantage over 16bit?
And what about if someone connects a 16bit mixer into a 24bit audio interface, will it record at 16bit or 24bit?
Thanks for the video and your knowledge on bit rate! I produce music and work sound production for a production company so I wanted to review the actual definition of bit rate.
What about 32bit float though. Im currently in audacity and just was curious...what does float mean?
float means you can clip past 0dbfs and signal can still be turned down without clipping effect. it's useful to render rough mixes in float 32 rather than old and outdated "-6db of headroom" because end result is the same, you can turn down audio for final mix. the -6db of headroom is old safety measure for 24-bit renders before 32 float was widespread
I think I can hear the difference between 16 and 24 bits in your voice when you say the "s", "p", "f"... but in a mix I probably can't hear the difference
How record in 24 bit in fl studio? My interface is set to 24 bit / 44.1Khz. But my wav files still say 16 bit. I can set the wav to 32 bit float, but that is just how wav format is written to disk. How do I know that the input signal is actually 24 bit?
Awesome vid man. I swear every word coming out of your mouth was information, and that's awesome. I learned a lot.
That’s one of the highest compliments I’ve received. Thank you very much.
Higher bit depths mean there's less rounding errors in palcing the xy corrdinates.. the rounding errors end up being noise. Lower bit will result in more noise, but honestly the noise floor at 16 bit is still good enough to where people really won't hear the noise unless the volume is so high in quiet parts it causes hearing damage. 24 bit levels are a bit overkill and use up too much drive space if storing. I do blind tests all the time with friends, they can never tell the difference between 44khz/16bit and 44khz/32bit. Yes.. highr but depth is better.. bit overkill
I sense a TLM103 review coming soon...
Gonna have to review cheap mics for a while to save up for that one.
Amazing video! Explanation was so brilliantly simple that I understood with ease :)
My question is which one is better to listen song , 16 bit or 24bit ?
Thanks for the info I can understand what you're getting across its like there's a way in but not all ways in the front door.
Thank you from Hong Kong, ... clear and professional presentation.
14 bit sounds really good and more analogue like for some unknown reason; 48 bit sounds clearer and sharper but the higher the bit rate, the more of the soul of the music is removed (not sure why, but this is my experience working in professional audio since 1988).
@Delatronics
Fun Fact: The earliest models of CD Players that were first released by Sony and others in 1984 actually had 14-bit D/A converter chips. It took Sony about a year to produce true 16-bit D/A chips in mass quantities for the consumer market because they were so advanced and difficult to produce at that time, hence too expensive for the consumer market. I still have a working Sony CDP-30 that my dad gave me for my birthday in 1984. Sounds great even now, haha.
Awesome video man! Precisely what I needed to know! You really made this video understandable! Thanks man!
Audio bit depth in no way relates to video resolution, but is directly related to pixel bit depth, with 16 bit pixels have far fewer possible colors.
However, the human visual system can't distinguish more colors so 16 bit looks identical to 1000+ bit (except for a handful of females with quad cones). Similarly, except for the 96 dB noise floor (we can detect up to 120) there's no audible difference between 16 and 1000+ bit. The only reason 24 bit is superior for capture and mixing is additive losses accumulate. But after mixing outputting to 16bit 44100 hz is indistinguishable from 24bit 48000 hz+ except for the higher noise floor which is virtually silent. Also, sampling rate does relate to framerate, but at best we can only hear up to 20khz, and exactly 2 samples are needed for full capture (mathematically proven and confirmed by hearing tests), so sampling above 40khz in absolutely no way improves audio quality for anyone.
You explain very well and quickly I love and appreciate that a lot
Always wondered this thanks for the video.
"Monty" Montgomery who created OGG container and Vorbis audio codec made an excellent technical explanation over at xiph.org/video/vid2.shtml
The chapter "bit-depth" and "dither" is relevant here.
TL;DW, The higher the bit depth, the lower the noise and the higher the dynamic range. Dynamic range could be expanded within the same bit depth with dithering.
From another article here at people.xiph.org/~xiphmont/demo/neil-young.html
16-bit is enough for hearing range. For recording, 24-bit's lower noise could give an extra headroom for boosting the signal in post and apply effects after effects in post.
I'm going to have to give this a check out.
The actual problem is that we have to avoid digital clipping. In order to be on the safe side, we don't use the full dynamic range, leaving some headroom for unexpected peaks. So we use, let's say 14 bits out of the available 16 bits, or 20 bits out of 24. Since the hearing dynamic is about 16 bits, there is some loss if you boost your 14 bits to 16, no loss if you reduce 20 to 16. If you can control your recording dynamic to 16 bits precisely, using analog compressor/limiter for example, you can stick to 16 bits.
Very good video and I record classical music in 24 bit and in low frequencies you can hear the difference. It´s very very small difference and it happen in some little points of the music, but exist. With attention on the bass, like me, hehehe... you can "feel" the difference of 16 bit to 24 bit. In classical music and some wen age musics, 24 bit is better. Congratulation for your video.
Fantastic breakdown...love your explanations as i get to pass it on to clients and friends - on a side note some people have even suggested 32 bit JUST IN CASE you clip in recordings it will still capture more sound information and you can bring the amptitude down and capture all the sound - theres one big youtuber who made a video on this and i thought although somewhat true it was overkill and anyone driving things that high are in a small minority but i see a few deciding to do 32 just becoz - although the file size increase really starts pushing practicality :) - just a side note, again great vid!
If you have a 32-bit converters and an analog input that can handle high levels. Otherwise recording in 32-bit FP on a 24 bit interface is just a waste of hard drive space and RAM.
If you have a 32-bit converters and an analog input that can handle high levels. Otherwise recording in 32-bit FP on a 24 bit interface is just a waste of hard drive space and RAM.
Which is the ideal parameters to record electric guitar with distortion effect?
Great stuff! I learnt something.. glad I subscribed!
Thanks for making it Black and White.
......THUMBS UP
Thanks for watching.
I’m not sure that 1080p vs 4K video is a good comparison, since that’s clearly noticeable to most people. 16bit vs 24bit is more like watching 1080p vs 1080p, taking a placebo pill, and telling everyone that you now have special eyes which can see the difference.
thank you for this! finally understand it more because of the comparisons. thank you very much
I work with Voice Over artists, and when they record at 16dB with low gain I get tons of background noise, it is a gamble recording at 16bit, you would need to have a stable good gain.... or just record at 24bit, which is what I ask people to do.
you should do this test of these two microphones to see who wins: audio technica at2020usb + vs blue yeti pro in versus series
Try this.
Generate a 4 kHz signal in Cool Edit or Audition using 96k/24 bit. Downsample it twice, one file to 96K/16 and another to 44.1K/24. Compare them yourself. See that the 24 bit depth does nothing for the signal resolution. On frequency analysis you can see that It only reduces the noise floor.
The 24 vs 16 bit graph in the movie is only a drawing and is incorrect.
So then why do editors and producers use 24 bit - to keep the noise floor low. They are processing music. You are not. So if you have a limited disc space, choose 96K over 24 bit. That also gets rid of the brick wall AD converter, something you can hear. If you have disc space, a good system and good hearing use 192/24. Your source material may vary.
I'm a musician, I work really hard on everything. I've got years of experience and I'm still learning. Up until now I have recorded in 16bit 44,1 and I feel like crying, all the work I've done and only to learn I've been missing out on production benefits... :(
Joel Firey to be fair, 16, 44.1 was the standard for CDs back in the 90's and 00's. So worst case scenario your music could potentially sound as low quality as anything on a professionally produced CD.
Except, not really
You should do a versus series video of the TLM 102 and TLM 103.
Very informative, and concise. Thank you for creating and sharing this video.
When you laughed was it because it was a tedious topic, or was it because it you were excited? Pretty sure we all got excited when talking about extra variables. Yes - Using 24bit before spotify and youtube compresses is beneficial. In my neighborhood I'm the only one who notices, and due to my "dontgiveafuggifitclipsinuteromilkit" attitude - I've definitely taken full advantage of those extra frequencies allotted for filling waves. When you go back to 16, and try to do what you did in 24...that's when you notice and go..."Oh helllllll no!"
@ReaktorLeak Way more powerful. Though for online uploads, kinda pointless, I agree with dudeman...having the highest possible quality before conversion definitely matters.
@ReaktorLeak I'm not a premium member on reverbnation, so my upload limit has been 8 megabytes. Only recently did I realize that I'd have to doctor before uploading overamped 24bit Waves as MP3's online. Your CD copy of In Utero can clip by 2.4 decibels...in your house, but not converted into as low as 128 or 96kbps on reverbnation. (So much garbage to fix or replace) Alot of the the time i would cut large projects into parts to get the highest possible bitrate with only the shortest 3 minute tracks achieving 320. 24bit. Great for home, but rattles the average listeners devices to sh!t. Just recently remastered some of my match-ups (mashups😡) and uploaded them on mixcloud to take advantage of 320 mp3, and found they now support 16bit wave. Very exciting. When I upload to spotify now via distrokid...I'll still upload 24bit...but...I'll make sure it sounds good as an mp3 as well, since all sites stream at 16bit. My system can handle it, and I gotta listen to this stuff forever, not just once casually so I'll always have my overamped waves. Haha - Hate neutering and converting them. Most can't tell a difference, dare I say it you'd practically have to have both waves up on your screen switching off between the two to notice.
www.mixcloud.com/Flow_Pdisco/dj-analysis-gamechangers/
U r a man who brings practically everything 💯 nic bro
That was a really good video. Thank you for explaining 4:06
4:27 "houston we have a problem"
eh the 4 bit was the best in my opinion really shows how great ur microphone is and the true depth in it 4:26
there is no audible difference on listening from 24 to 16, you can do blind test of 24 to 16 and you will never know which is one or another ...
Ni siquiera hablo inglés y entender tu explicación bro... Crack 👍
32 coz more volume space
192khz coz its latency (then more khz you have in ur daw..then lower latency monitoring) ..not just a sound...
Agreed. The higher the sample rate, the better the latency performance.
Really good explanation man. Hats off.
had a few Years ago the M-Audio Delta 1010LT and i loved, then my pc got stolen!
NOW i use for a few Months again a M-Audio Delta 1010LT PCI Card, its superb for non USB audiosignalstudio also for a good Audio Homestudio without any latencyroblems!!
That’s what I use but the non LT version. Still works, no need for anything else.
Actually i feel huge difference, like dynamics from your voice disappeared at 16bit and it was less open sound and not crisp at all.. I guess i bought a good pair of cheap Sony headphones 😉
Thanks! I mean it, great video and description. Hope 2 see more.
Good Stuff.
Can you do a 24-bit USB mic comparison: Samson G-Track Pro vs. Apogee Mic plus vs. Shure MV88 iOS
I believe you've already done the Samson individually, but not the Apogee.
ps. I have noticed watching your videos that XLR mics on the U-Phoria box don't sound as crisp as the XLR mics on basically every other USB plugin you've tried. You may want to do a comparison of all the box packages... TASCAM, FocusRite & Steinberg
TH-cam uses AAC for sound . This compression is very lossy I think it's on the same level as MP3
Most recordings are in WAV or FLAC I personally record in FLAC .
On TH-cam due to the heavy compression it's pretty pointless IMO to record 24-bit
Spotify uses Ogg Vorbis which is the best compression so far . The losses are minimum and it sounds much better . So if you're aiming for Spotify you have benefits to record 24-bit
I may be wrong . I don't know much . But I hope I helped
So if an music album is remastered to 24-bit, does it make the album better? Is it worth buying? I have the original CDs, so is it any differences? Is it any better? Is it worth it? Is it like 5.1 surround sound remixes? The Beatles’ Sgt. Pepper’s, White Album & Abbey Road was remastered to 5.1 surround sound. What does 24-bit means? Is it 5.1 surround sound remixes? The Dutch prog rock band Focus is also releasing a 50th Anniversary box set now with remastered material. It says it is a «24-bit remaster». Is it worth it when I already have the original CD albums?
24-bits just means you have 8 extra bits, and every time you add a bit , the audio resolution doubles. 24-bit has 256 times more resolution than 16-bit.
Honestly, I found it depends on the album. Sometimes it's not perceivable, other times, I can really hear it in the vocals. I also think it depends on how many instruments are being recorded.