No it doesn't that's a common misconception. While it was traditionally true that more bits gave you higher headroom if desired this advantage only goes up to around 21-bits due to the laws of physics because at room temperature no pre-amp technology available to humanity at the moment is operate at a low enough noise floor such that the extra bits beyond 21-ish bits doesn't encode noise.Thus having more than 24-bits in a signal doesn't really make any sense its mainly just industry marketing 32-bits gives absolutely no advantage its not even a 32-bit interger value that the circuit operates on usually its a floating point value so you don't actually get a full 32 bits of raw information either.
@@shayhan6227 You are neglecting signals above unity. Sure, a reliable noise floor below -120 dBu is basically non-existent. However, a dual gain input can help in situations where the upper bound of your signal is not known. That’s the advantage of the 32 bit field recorders: transient headroom.
@@warasilawombat I agree that this is possible by stacking dual ADCs at different gains to increase the dynamic range but during playback a human being normally wouldn't be able to take advantage of the full dynamic range without damaging their ear unless they always have their hand on the volume knob to lower it when the extremely high db portions of the song are played and raise it back up when the quiet portions are played back so that they can hear all the detail. There also would need to be an equal dual DAC and amp system that can somehow reproduce that 32-bit signal without inducing its own circuit noise floor when the signal is amplified for the speakers that can handle it without distortion.On top of this most microphones cannot record beyond 140dB of dynamic range without inducing a good chunk of distortion due to the diaphragm's physical non-linear excursion characteristics. You'd need some sort of multi-microphone system to do this properly. Also the other thing is that 32-bit floating point only has 23-bits of actual true data the rest is simply an exponent. Sure the “dynamic range” will be captured but the TRUE audible quality of the signal’s precision will still be trimmed to 23-bits. To put in perspective imagine a 16-bit floating point equivalent variable where you have an 8-bit mantissa (you'd technically on paper have a massive dynamic range but the true "quality" of the signal would be pretty poor and much worse than CDs). This is sort of like how OLED TV manufacturers claim that their TV can handle an “Infinite” contrast ratio, while its technically true since you are dividing by zero it doesn’t say much about the quality of the TV because technically as long as the TV can display any brightness above pitch black its technically now “infinite” simply because the pixel can turn off. Another thing I would argue (and this part is just my opinion) is that most of the pleasing signals being recorded and the ones with truly impressive/artistic value are usually recorded by a microphone instead of simply being directly synthesized by what is essentially a digital function generator. I think "live" recordings are where audiophiles can truly experience the value of the music (natural echoes and reverb of the unique material properties of the unique instruments and the distance they are from the mic, imperfections of the human voice, recording room characteristics, etc) recording as opposed to a synthetic and pure compilation of tonal signals and filters in electronic music. Its like listening to a pure crystal oscillator based sine wave, sure it will always be pure and perfect but there is not much to explore about it artistically.
@@shayhan6227 you are completely missing the point here. The feature we care about is that we don’t have to care about fiddling about with perfecting the gain settings when on location. If the mic can pick up the noise, it’ll go into the record. In post, you then have the freedom to correct the gain to whatever you need it to be, or compress it, or whatever else you want. I have a whole additional spiel about how the floating point format is basically perfect for dealing with the dynamic range of audio but thats best saved for another time. It alls revolves around the fact that the mantissa keeps your digital noise floor essentially 23ish bits below whatever signal you are respresenting PER SAMPLE, which is cool as hell.
Additionally to the benefit of not having to care about clipping while working, in the production, mixing and mastering stages we use plugins that process the audio in various ways. Since every process performed on the audio, be it an EQ, compressor, reverb, saturation or almost anything else is degrading the sound quality by definition, and a typical project can have more than 10 plugins on some channels and then more on the various buses and the master bus, it is better to have the audio stored in 32 bits. Newer DAW's even offer the option to work in 64 bits and as we all know 64bit ADC's and DAC's don't exist (for audio applications anyway) . When the final project is rendered 16 or 24 bits can be used.
Don't record in 32-bit (integer OR floating point), but process in 32-bit floating point, it allows you to go above 1.0 or below -1.0 (where it would normally clip) during editing and processing.
Beyond 24-bit 192Khz I don't notice 99% of the difference. Other than it 'sounds better somehow' (maybe sounds less compressed/condensed/limited?). But I do hear it.
holly shit, this video has some of the cleanest audio i ever listen on youtube, minus some popping here and there, the mic is amazing, really good explanation
OK response for a listener's perspective. However, for working in a DAW, 32 (or even 64) is vastly preferred for various reasons. Agreed about 88.2kHz when it comes to working sample rate and yes for the divisibility reason.
yes, but 96kHz and 192kHz is just a multiple of that. it increases the resolution, especially in stereo depth. another advantage is the analog low pass filtering in DACs which does not need to be that agressive.
Well 44.1/88.2/176 were music standards. And 48/96/192 were cinema/video standards. Today it doesnt really matter like back than, when we had cd's - they had a specific requirement for the media to be burned on it.
The 2 boundaries in audio are the signals noise floor and clipping, each on the opposite end of the scale, from which you want to keep clear of. If the noise floor is around 90dB, all the extra dynamic rang will give you is more detailed noise floor. The last couple of most significant bits are often unused to prevent clipping. You want to make sure the recording gain is set so that you just prevent clipping, and have the strongest signal vs noise floor. Least significant bits of 24 bits represents nanovolts, 32 bits in the picovolts and smaller. A single resistor produces way more noise due to thermal effects. 32 bit float can be useful for editing, but be aware that every operation you apply degrades the signal by a tiny amount due to rounding errors. Floats can be considered lossy. Same goes for integer calculations. Conversions often happen between the two in software, especially in domain conversions like amplitude/time to amplitude frequency domain such as an equalizer, or notch filters.
As usual, another interesting topic and discussion by people that are way more experienced and smarter than I. I am neither a recording engineer nor artist. I just enjoy listening to good music, both digital and analog; digital for convenience (streaming music stations) and analog (LPs) for a more emotional and enjoyable experience. That is just my personal experience and ears, please do not start the analog vs. digital debate. Any way, I have a Rega phono stage with a analog to digital converter and usb digital output that I have been using to digitize my LPs. I like having my favorite LPs on my DAP. The sound quality of those rips is pretty good. The ADC is limited to 16 bit 48khz. Does anyone think that moving up to a converter with 24 bits and 96khz might yield a better sound quality?? I have a decent Cayin N3 Pro DAP and some nice JVC IEMs. Thanks for any input.
If I understand this correctly, the 44.1 kHz sampling rate came to be a standard because this was what the readily available Sony PCM-1600 U-matic digital recorder used. This worked reasonably well back in the day. Specialized digital audio recorders (usually with diverse, proprietary encoding techniques) were rare on the ground, not to mention being really expensive. When the powers that be made things official with the Red Book for CDs, this became the standard.
It all depends on what they are talking about, if it's A/D and D/A, then 24 bits is it, However if you are talking about a DAW, or any digital audio processing system, don't limit it to 24bits, as you will be doing heaps maths on that audio (even just changing the levels) and you want to keep any artefacts as far away from the audio as possible, particularly as you will inevitably be doing transform after transform. Of course it will probably be cropped back to 24 or 16 bits in the end.
I don’t disagree with recording at that resolution today, but I always wonder what film and music would be like today, had producers had technology or resolution from the future. I think of classic film or music from the 20s and 30’s had use of cameras and mics and studios from a decade or two in their future. This is an opportunity to record at a level that could be utilized a decade from now.
The CEO of Sony wanted a particular recording of Beethoven's 9th to fit on a single disc, so Sony engineers dropped the sample rate down to 44.1K to achieve this. They did not know they could have tightened the track pitch to achieve this goal yet, which came later to increase the capacity of CD-ROM.
By default, my (B.M.C.) DAC upsamples the bit rate of every PCM (16 or 24 bit) signal to 32 bit before it will be converted into analog. When a 32 bit (floating point) is more ‘CPU-friendly’, that would make sense. Less noise and jitter, maybe? Some audiophile recording labels offer 32 bit recordered DXD downloads nowadays in WAV format (as FLAC codec does not support floating point). It is proclaimed that these ultra big files might sound better depending on the quality of the DAC. In case the DAC has to work much harder to handle these big files, it could produce more internal noise. Then, it will work out in the opposite direction - resulting in a more distorted sound reproduction. For me, using a Mac-Mini (model late 2012) with Audirvāna software while bypassing Apple’s internal audio core to send these files to my DAC the differences between 32 and 24 bit of the same file at the same DXD sampling rate in the same (WAV) format, seems to be extremely small. At least at my home system. It is more like a “trade off”indeed. But it keeps interesting to check out all these new advantages and follow the discussions in the comment section.
An argument like this comes up in photography and photo editing. The eventual consensus is that you want to capture as much information as you can (given storage and performance constraints) so that what you lose in editing isn't important. It's not about the raw stuff as a final product as much as how much you can munge it up before it gets ugly.
There is no 32-bit PCM. 24-bit already goes beyond the noise floor possible with electronics. "32-bit" in audio just means that the 24-bit PCM is represented in 32-bit floating point format (CPU friendly) - it is still just 24-bit resolution.
That might be true for playback, but not recording. For recording and sound editing, 32-bit is a real thing (but quite new, at least in pro-sumer tech). A microphone signal can go to 2 different pre-amps, then the signal can be combined by the recorder into a 32-bit PCM.
@@emeryththeman basically yes, that's my understanding that it goes through 2 ADCs and is then combined. It's true that microphones don't generally provide more dynamic range, though. Mine has a maximum SPL of 130dBSPL and self noise of 13dBA, that still makes it below the range possible to capture in 24-bit. But that doesn't mean that while recording to 24-bit I can just set the proper gain level once and be prepared for any kind of situation. In filmmaking and field recording one always has to set a proper gain level for the situation with 24-bit. To be honest I'm not sure why, the numbers being what they are. Maybe just that setting the gain level in the recorder provides better distance of signal from noise than setting it in post? I should actually try to understand this better... And then with 32-bit recording, there is no gain setting at all.
@@emeryththeman maybe the need to set the right gain with 24-bit comes from the precision of integers at different volume levels. Recording at lower levels, when you then bring the volume up in post, the difference between samples will have less precision than if it would have been recorded at higher gain in the first place?
Many audiophiles conflate bits and sample rates used in the recording process with bits and sample rates in the playback files, this is a mistake. High sample rates and more bits are used in the recording process to enable editing but are pointless for playback. If you actually had a 24 bit recording with a true 144dB dynamic range and the equipment able to handle it (which is impossible) and your quiet listening room had a 30dB background then a peak sound above background would be 174dB which would be lethal. No, I am not exaggerating, this sound level kills. Even a true 16 bit 96dB recording would be unplayable as the quiet passages would be inaudible or the loud passages screamingy loud. Most modern recordings can be delivered with 10 or 12 bits. A full symphony orchestra giving it all it's got (ƒƒƒƒ) peaks at about 104 dB SPL. Let's give the orchestra 105 dB, and 105 dB - 30 dB = only 75 dB real dynamic range. Most recordings are compressed dynamically to a sensible range so the listener is not constantly having to twiddle the volume control. Higher sample rates on playback produces ultrasonic noise and does not improve audio in the audible range.
Only correction i would add is higher samplerates for playback do matter if you desire a smoother sound (less digital grain and grit) or if you have a great DAC that supersamples, that's fine too. DSD playback is my favorite though.
@@LunarLightLtd1 This is a common misunderstanding of how digital audio works. A 44.1KHz CD can deliver audibly perfect waveforms up to 20KHz, the upper limit of human hearing. No graininess or stair steps. Higher sample rates on playback increase bandwidth above audibility , this is all it does. This is why so called hi-res audio is marketing madness. On the other hand, DSD can produce stair steps and distortion and huge sample rates are required to make this effect inaudible.
@@SanceShaji Probably impossible and highly undesirable. The best electronics struggle to give 120dB (20 bits) dynamic range. A true 144dB dynamic range recording would be unplayable. If you set volume as loud as you could stand on the loud passages then the quiet sounds would be far too faint to hear.
Audiophiles be like: “My ears and my opinions have empirically determined that there is no logical scientific reason for the artists or engineers to improving the resolution of this recording” ..Pure nonsense… Many ppl thought like this at every advent throughout the many evolutions of recorded audio… The weird bigotry of saying “it’s good enough for *MY* ears” while simultaneously romanticizing “Audio Nirvana” truly confounds me.. One half of every audiophile is obsessed with resolution while the other half seems to shun it like a vampire in full sun. 😂🤦🏼♂️🤷🏼♂️
i don't agree, during the mixing part the headroom is very nice to have. it make's tracks better to mannage and give space. for the final mix 24 bit is the way to go. after some good dithering
I wish he answered this question by telling me what his produced gear can do. Are his amps preamps ……. Capable of 24 or 32 bit resolution. Generally a system capable of 120 db range is off the chart with design and expense. Most people are happy with 105.
Nope, it's only about dynamic range/signal to noise ratio. Since 16-bit is transparent, more bit doesn't do any good for consumer audio formats. Studios might need 24 bits for headroom when recording and mixing though. There's no more nuance or resolution to be heard by upping the bits from 16 to 24 or 32 bits. Check out Monty Montgomery's explainers, he's the best: th-cam.com/video/cIQ9IXSUzuM/w-d-xo.html and th-cam.com/video/FG9jemV1T7I/w-d-xo.html .
114dB is massive. Also that is very small mV signal, also resistors in DACs become very difficult to be both accurate and stable with temperature at the level of 32bit LSB (we're talking 0.01% or below tolerance and
There is only one dac in the world at the moment that can playback almost full 24 bits, 130db no distortion, it's so good you can "listen" to actual data stream signal. Topping d90se
48K is a natural multiple when no external reference nor compatibility was required. But 44.1K was chosen because it allowed 3 x 2 channel samples into each video line as video recorders were first used. The original Digital Converters like the Sony PCM-100 did not have storage, pro-video decks were used. Often 3/4" U-matic. Or naturally with Sony, later Beta. And technically electronics can not resolve voltage differences past 22 bits. Even 20 bits is stretching the abilities of a ADC to convert. Nor a DAC to resolve. Even the best ladder DACs.
This is one of those subjects you could argue about all day, everyday. Most say that there's nothing beyond 96/24 in PCM, after that go DSD. When they record in 192 it's so they can zoom in as much as possible to cut and edit. So 24bit and 96k is all PCM can do as far as quality. Most say - Don't argue with me personally
People forget that sample rate directly influences latency. It's going to need more CPU for crackle-free playback, but cranking it up will mean better latency, as unintuitive as this may sound. Almost no applications in the box require anything abive 44.1 or 48, but if you're trying to record e-drums into your vst and monitor yourself too, that is going to matter. As far as dynamic range is concerned... ya no, it really doesn't matter. Oh and SR matters when slowing down samples too, for obvious reasons.
Most say 24bit/96k for playback, everything above that is for editing. Every format has a limit, and pressing X2 on a calculator doesn't improve sound quality
Correct me I am wrong (and there is a good chance I might be!) , from what you are saying a bit gives you a fixed amount of dynamics ie 1 bit will give you x Dbs while 10 bits will give 10x ? That is to say your max and min levels can be dynamic but you need sufficient bits to express that. My Maths lecturer said to me many years ago that any model that does not work at its limits is probably wrong. So using that premise : 1 bit resolution will only define the lowest part of the sound you are representing using binary, everything else is not defined and there for lost. So also working at 32bit means most of the high end bits are also redundant as there is no sound at that level to set the bits to 1? This would automatically set the limit of resolution you can go to regardless of sample rate. Real music (as I am sure you know) is more complicated then a sine wave. In a real environment sound (and the recorded wave form) will be the superposition of all the instruments ( each of their fundamentals and harmonics), the reflections from the environment. In-bedded in all of this will be phasual (cant spell!) information. If you look at the wave form at a particular point in space and a particular point in time- you will have a voltage detected by you sensor. If you increase you time base from an instant to a couple of seconds you will have info. This at the present is analogue. The fine detail on that wave form will tell the listener about the instruments and where it is ( with reference to another sensor position). If you want to use digital representation of this you need to drill down to the point where it beyond our ability to detract any difference.By increasing the sample rate you can get closer to pulling these nuances out. But is this not true of the amplitude of the signal as well ? If we have a fixed bit to amplitude Db ratio then any voltage fluctuation with in this ratio will not be seen. This will translate into a harmonic/phase difference. Where am I going wrong with this observation ??
I read something about 48khz easier to use for film (video) purposes. Something about easier syncing 48khz stream with the video or tv use. But can't remember the exact details.
44.1Khz/16bit is kinda good enough for hifi consuming.. double or quadruple the rates for authoring or mastering... yep i agree that DSD is great as a source format. DSD64 is like 192khz source in storage size. a 4-minute song takes 170MB of space
It's a VPI direct drive turntable with the 'Fatboy' gimbal tonearm, possibly the HW-40 Anniversary model, No idea what phono cartridge Paul McG uses, but I read somewhere that Harry Weisfeld (the HW in HW-40) was partial to the Kiseki Purple Heart moving coil.
32 bit doesn't sound better than regular 16-bit audio for consumer formats. First off, about 90 percent of all available digital music is 16-bit. Upsampling to 32 bit doesn't change anything, it just wastes computer space. More bit doesn't mean you get a higher or better resolution. It means you get a better signal-to-noise ratio. Theoretically 32-bit would mean a SNR of 192dB (not possible today). But what good does that do? You can't listen to music at 192dB, you'd die. 24 bit in theory means a SNR of 144dB, but in practice closer to 118-120. That is also way more than you need. Everything above 90-95 is transparent (assuming a flat frequency response and low distortion which all competent dacs have acheived at a reasonable price since the late 80's), you can't hear that noise floor if you don't kill the music and max out the volume on your amp. Also, most music we listen to have a dynamic range of 30db tops (lower than that for most pop music). So it's all overkill and doesn't make any audible differences to have more bits. In a professional setting like a studio it's preferable to work in 24 or 32 bits so you'll have more headroom when mixing and so on though.
Another reason why 32 bit float can be better than 24 bit for recording and post-production : with PCM integer audio like 24 bit, not all the bits are equal, and the higher bits are more significant than the lower ones. You need to record louder for getting better quality, but you have a hard ceiling. For example, in the early days, when recording in 8 bit, the loudest signals in the maximum 6dB range had only 127 possible values. The quieter signals had even less possible values. Needless to say, the quality was very bad and full of quantization noise and errors. With 24 bit audio, even when above quantization noise, the subtle but important signals like natural reverberation are badly digitized if the recording level is too low, and brought up later. Let's say, theoritically, that no important signal should be quieter than the 9th bit. The 9th bit can have 255 different values for that low 6 dB range, and it may be enough for recording the subtle, natural reverberation. If you record properly at 24 bit, with peak levels under -12dBfs, then you already don't make any use of the two most significant bits, the 23rd and 24th, which make for 3/4 of the datas in the file, and the usable range left is around 13 bits. That's 78 dB of usable dynamic range, which is quite high, but may not be enough. With 32 bit float audio all the bits are equal. Please listen to the demonstration here : www.sounddevices.com/low-signal-32-bit-float/
If I understand pcm correctly, 24 bit is 12 bits per channel. 16 is 8 bits per channel. I think the confusion is everyone thinks more is better. Not necessarily. Look at DSD. It’s only 1 bit but the sampling rate is about 2 or 3 MHz! Of course DSD is PWM digital, pulse width modulation. Not PCM - pulse code modulation.
16bit PCM audio recording means 16bit audio per channel, regular audio CD is 16 bits x 44100Hz x 2 channel = 1411200 bps DSD64 has a data rate 4 times of CD, 5644800 bps
Speaking about not knowing what to do with dynamic range: Don't you love how the automatic dynamic range limiter starts to fill the silence with background noise (like at 2:58)?
Curious as to the lowest reliable frequency on can get online? I am thinking about some GOOD Towers that go down to 30 HZ + to - 3db. Had a pair before and miss the bottom end on some music. Thanks
@@AnimusInvidious Not. Why would anyone want to hear much of anything less than 30 Hz, mind u +/- 3 db so likely down to a realistic and solid 25 Hz. ??
@@AnimusInvidious Well high quality speakers that are flat down to 30 Hz don't need a Sub. The speakers I am talking about are Tower Studio Monitors that in todays money would retail for $5,300 US with a shipping weight of 100 lbs each
So. For wild recording, if you have a true 32, or 64 bit, or 32/64 floating bit recorder, you can never ever clip on the recording. In post, you can make a whisper, and an atomic bomb, be normalized.
Here we go again with not really understanding how things work. The bit depth determines the resolution of waveform capture at an instantaneous point on the time domain. Therefore, the higher the bit depth the greater the waveform resolution. So yes 32bit contains more detail. And the faster it collects them the better. Because everything in between get electrically averaged out.
Bit depth determines total dynamic range (as well as resolution). The higher the bit depth, the higher the possible dynamic range as well (and here's where you're correct) the resolution as well.
@@Paulmcgowanpsaudio Digital photography bit depth analogy: dynamic range=#of colors&saturation (ie intensity) For audio, dynamic range is frequency&litude thats' it. When desiring and dealing with dynamic range captures this microscopically detailed, reproduction equipment slew rates are critical. We go to further and further lengths to extract more from our sources, when it's the sources that are lacking. If it wasn't captured, it's not there. The best microphones in the world are still lacking in a major and obvious way (at least to me). I am working on a couple new concept microphones to fix that if you're interested to listen under NDA but willing to share. :)
Oh, and 32-bit is not completely useless - it can be useful for situations where the required dynamic range is not known in advance, such as documentary filmmaking, or even non-documentary filmmaking in a scene where you don't know how loud the actors or other production sounds are going to get.
You can still clip 32 bit, just turn up the gain on the recorder. The improvement in S/N is on the quiet end, not the loud end. Beside the noise floor of a wire is higher than what 32 bit will yield, even 24 bit is at the limit of what can be achieved in noise.
24 bits is already more than enough for capture , transport, and playout. Just as 96khz is also a practical limit. The three LSBs are just noise. 32bits or more are needed in DSP systems due to the multiplication needed to alter gain and frequency response.
MASSIVE MISUNDERSTANDING OF THE DIFFERENCE BETWEEN 24 BIT FIXED VERSUS 32 BIT FLOAT! Everyone needs to do their homework here. The float is an extra 8 bit internal exponent, that has nothing to do with ProTools converters and only applies to the internal mix bus. All converters for ProTools are 24 bit and automatically converted to 32 for the internal mix bus; it is pointless to record 32 bit float files. However bouncing 32 bit float files for mastering provides the mastering engineer more headroom, but all converters are fixed converters and ONLY decode 24 bit. (There is an exception given that Steinberg has recently released 32 bit fixed point converters, but these do not apply to ProTools and are not float files).
32 bits is totally unnecessary. 16 bit gives all the dynamic range you could ever use. The dynamic range of 32 bit would be like from one molecule hitting the eardrum up to the power of the big bang.
I just hope that James does not end up equalizing and compressing his creations, like nearly all engineers do (it is like a drug that makes nearly all recording engineers climax to vandalizing sonic works of art). If James does end up equalizing and compressing his creations, then his question is pointless, and he might as well stick to the .mp3 format while dining at McDonalds.
even the noise on a battery - which is very stable - is louder than -144dB. that means, nobody has a power supply that gives you noise free power for you amp circuits to reach -144dB accurate signals. so, 24 bits is accurate enough
Since DACs cant resolve past 21 bits, 24 vs 32 bit is moot. Besides, the rest of the equipment in the recording and playback chains all have S/N ratios higher (worse than) the S/N ratio for 24 bit (144dB), so still pointless. 16 vs 24 is worth it, but 32? Nope.
Bit depth is about volume, not frequency. Are you able to identify which is louder, A car horn & a jet engine? If so that means 16 bit is not beyond your range..
Just ask yourself: „would I rather have a 24 year old girl or a 32 year old. So you’re able to easily remember the essence of what Paul really well explained 😎
No reason to go to 32 bit FOR WHO. If i buy a 32 bit song or mix with 32 bit. Just because you dont need it in theory doesnt mean the 32 bit file isnt the better music file. This is never discussed.
32-bit float is great when you're unsure of the dynamic range such as field recording. It effectively eliminates digital clipping.
I like to use the new 32 bit recorders for unattended field work. No worries about clipping!
No it doesn't that's a common misconception. While it was traditionally true that more bits gave you higher headroom if desired this advantage only goes up to around 21-bits due to the laws of physics because at room temperature no pre-amp technology available to humanity at the moment is operate at a low enough noise floor such that the extra bits beyond 21-ish bits doesn't encode noise.Thus having more than 24-bits in a signal doesn't really make any sense its mainly just industry marketing 32-bits gives absolutely no advantage its not even a 32-bit interger value that the circuit operates on usually its a floating point value so you don't actually get a full 32 bits of raw information either.
@@shayhan6227 You are neglecting signals above unity. Sure, a reliable noise floor below -120 dBu is basically non-existent. However, a dual gain input can help in situations where the upper bound of your signal is not known. That’s the advantage of the 32 bit field recorders: transient headroom.
@@warasilawombat I agree that this is possible by stacking dual ADCs at different gains to increase the dynamic range but during playback a human being normally wouldn't be able to take advantage of the full dynamic range without damaging their ear unless they always have their hand on the volume knob to lower it when the extremely high db portions of the song are played and raise it back up when the quiet portions are played back so that they can hear all the detail. There also would need to be an equal dual DAC and amp system that can somehow reproduce that 32-bit signal without inducing its own circuit noise floor when the signal is amplified for the speakers that can handle it without distortion.On top of this most microphones cannot record beyond 140dB of dynamic range without inducing a good chunk of distortion due to the diaphragm's physical non-linear excursion characteristics. You'd need some sort of multi-microphone system to do this properly.
Also the other thing is that 32-bit floating point only has 23-bits of actual true data the rest is simply an exponent. Sure the “dynamic range” will be captured but the TRUE audible quality of the signal’s precision will still be trimmed to 23-bits. To put in perspective imagine a 16-bit floating point equivalent variable where you have an 8-bit mantissa (you'd technically on paper have a massive dynamic range but the true "quality" of the signal would be pretty poor and much worse than CDs). This is sort of like how OLED TV manufacturers claim that their TV can handle an “Infinite” contrast ratio, while its technically true since you are dividing by zero it doesn’t say much about the quality of the TV because technically as long as the TV can display any brightness above pitch black its technically now “infinite” simply because the pixel can turn off.
Another thing I would argue (and this part is just my opinion) is that most of the pleasing signals being recorded and the ones with truly impressive/artistic value are usually recorded by a microphone instead of simply being directly synthesized by what is essentially a digital function generator. I think "live" recordings are where audiophiles can truly experience the value of the music (natural echoes and reverb of the unique material properties of the unique instruments and the distance they are from the mic, imperfections of the human voice, recording room characteristics, etc) recording as opposed to a synthetic and pure compilation of tonal signals and filters in electronic music. Its like listening to a pure crystal oscillator based sine wave, sure it will always be pure and perfect but there is not much to explore about it artistically.
@@shayhan6227 you are completely missing the point here. The feature we care about is that we don’t have to care about fiddling about with perfecting the gain settings when on location. If the mic can pick up the noise, it’ll go into the record. In post, you then have the freedom to correct the gain to whatever you need it to be, or compress it, or whatever else you want.
I have a whole additional spiel about how the floating point format is basically perfect for dealing with the dynamic range of audio but thats best saved for another time. It alls revolves around the fact that the mantissa keeps your digital noise floor essentially 23ish bits below whatever signal you are respresenting PER SAMPLE, which is cool as hell.
Additionally to the benefit of not having to care about clipping while working, in the production, mixing and mastering stages we use plugins that process the audio in various ways. Since every process performed on the audio, be it an EQ, compressor, reverb, saturation or almost anything else is degrading the sound quality by definition, and a typical project can have more than 10 plugins on some channels and then more on the various buses and the master bus, it is better to have the audio stored in 32 bits. Newer DAW's even offer the option to work in 64 bits and as we all know 64bit ADC's and DAC's don't exist (for audio applications anyway) . When the final project is rendered 16 or 24 bits can be used.
Don't record in 32-bit (integer OR floating point), but process in 32-bit floating point, it allows you to go above 1.0 or below -1.0 (where it would normally clip) during editing and processing.
Definitely record in 32 bit if you plan on processing with it!
This actually makes sense, since all editing would be below the noise floor.
Beyond 24-bit 192Khz I don't notice 99% of the difference. Other than it 'sounds better somehow' (maybe sounds less compressed/condensed/limited?). But I do hear it.
holly shit, this video has some of the cleanest audio i ever listen on youtube, minus some popping here and there, the mic is amazing, really good explanation
OK response for a listener's perspective. However, for working in a DAW, 32 (or even 64) is vastly preferred for various reasons.
Agreed about 88.2kHz when it comes to working sample rate and yes for the divisibility reason.
from what I understand, the question of choosing the 48 KHz and other values has to do with the synchronization with the video files
yes, but 96kHz and 192kHz is just a multiple of that. it increases the resolution, especially in stereo depth. another advantage is the analog low pass filtering in DACs which does not need to be that agressive.
Well 44.1/88.2/176 were music standards. And 48/96/192 were cinema/video standards. Today it doesnt really matter like back than, when we had cd's - they had a specific requirement for the media to be burned on it.
48kHz (or multiple) was/is used because it is easier to synchronize with video.
The 2 boundaries in audio are the signals noise floor and clipping, each on the opposite end of the scale, from which you want to keep clear of. If the noise floor is around 90dB, all the extra dynamic rang will give you is more detailed noise floor. The last couple of most significant bits are often unused to prevent clipping. You want to make sure the recording gain is set so that you just prevent clipping, and have the strongest signal vs noise floor. Least significant bits of 24 bits represents nanovolts, 32 bits in the picovolts and smaller. A single resistor produces way more noise due to thermal effects.
32 bit float can be useful for editing, but be aware that every operation you apply degrades the signal by a tiny amount due to rounding errors. Floats can be considered lossy. Same goes for integer calculations. Conversions often happen between the two in software, especially in domain conversions like amplitude/time to amplitude frequency domain such as an equalizer, or notch filters.
As usual, another interesting topic and discussion by people that are way more experienced and smarter than I. I am neither a recording engineer nor artist. I just enjoy listening to good music, both digital and analog; digital for convenience (streaming music stations) and analog (LPs) for a more emotional and enjoyable experience. That is just my personal experience and ears, please do not start the analog vs. digital debate. Any way, I have a Rega phono stage with a analog to digital converter and usb digital output that I have been using to digitize my LPs. I like having my favorite LPs on my DAP. The sound quality of those rips is pretty good. The ADC is limited to 16 bit 48khz. Does anyone think that moving up to a converter with 24 bits and 96khz might yield a better sound quality?? I have a decent Cayin N3 Pro DAP and some nice JVC IEMs. Thanks for any input.
'A distant story remembered by someone, but not me! ' haha quality!
If I understand this correctly, the 44.1 kHz sampling rate came to be a standard because this was what the readily available Sony PCM-1600 U-matic digital recorder used. This worked reasonably well back in the day. Specialized digital audio recorders (usually with diverse, proprietary encoding techniques) were rare on the ground, not to mention being really expensive. When the powers that be made things official with the Red Book for CDs, this became the standard.
@Douglas Blake that was part of it too no doubt
A vintage 2 bit system that sounds great is a bargain!
Kerosene powered for best effect.
One of the best channels around here!
It all depends on what they are talking about, if it's A/D and D/A, then 24 bits is it, However if you are talking about a DAW, or any digital audio processing system, don't limit it to 24bits, as you will be doing heaps maths on that audio (even just changing the levels) and you want to keep any artefacts as far away from the audio as possible, particularly as you will inevitably be doing transform after transform. Of course it will probably be cropped back to 24 or 16 bits in the end.
I don’t disagree with recording at that resolution today, but I always wonder what film and music would be like today, had producers had technology or resolution from the future. I think of classic film or music from the 20s and 30’s had use of cameras and mics and studios from a decade or two in their future. This is an opportunity to record at a level that could be utilized a decade from now.
88.2kHz vs 96kHz explained, thank you!
The CEO of Sony wanted a particular recording of Beethoven's 9th to fit on a single disc, so Sony engineers dropped the sample rate down to 44.1K to achieve this. They did not know they could have tightened the track pitch to achieve this goal yet, which came later to increase the capacity of CD-ROM.
By default, my (B.M.C.) DAC upsamples the bit rate of every PCM (16 or 24 bit) signal to 32 bit before it will be converted into analog. When a 32 bit (floating point) is more ‘CPU-friendly’, that would make sense. Less noise and jitter, maybe? Some audiophile recording labels offer 32 bit recordered DXD downloads nowadays in WAV format (as FLAC codec does not support floating point). It is proclaimed that these ultra big files might sound better depending on the quality of the DAC. In case the DAC has to work much harder to handle these big files, it could produce more internal noise. Then, it will work out in the opposite direction - resulting in a more distorted sound reproduction. For me, using a Mac-Mini (model late 2012) with Audirvāna software while bypassing Apple’s internal audio core to send these files to my DAC the differences between 32 and 24 bit of the same file at the same DXD sampling rate in the same (WAV) format, seems to be extremely small. At least at my home system. It is more like a “trade off”indeed. But it keeps interesting to check out all these new advantages and follow the discussions in the comment section.
Thank you old man 👍👍🏻👍🏼👍🏽👍🏾👍🏿
An argument like this comes up in photography and photo editing. The eventual consensus is that you want to capture as much information as you can (given storage and performance constraints) so that what you lose in editing isn't important. It's not about the raw stuff as a final product as much as how much you can munge it up before it gets ugly.
There is no 32-bit PCM. 24-bit already goes beyond the noise floor possible with electronics. "32-bit" in audio just means that the 24-bit PCM is represented in 32-bit floating point format (CPU friendly) - it is still just 24-bit resolution.
That might be true for playback, but not recording. For recording and sound editing, 32-bit is a real thing (but quite new, at least in pro-sumer tech). A microphone signal can go to 2 different pre-amps, then the signal can be combined by the recorder into a 32-bit PCM.
There is 32 bit PCM. Main use is in recording.
@@emeryththeman basically yes, that's my understanding that it goes through 2 ADCs and is then combined. It's true that microphones don't generally provide more dynamic range, though. Mine has a maximum SPL of 130dBSPL and self noise of 13dBA, that still makes it below the range possible to capture in 24-bit. But that doesn't mean that while recording to 24-bit I can just set the proper gain level once and be prepared for any kind of situation. In filmmaking and field recording one always has to set a proper gain level for the situation with 24-bit. To be honest I'm not sure why, the numbers being what they are. Maybe just that setting the gain level in the recorder provides better distance of signal from noise than setting it in post? I should actually try to understand this better...
And then with 32-bit recording, there is no gain setting at all.
@@emeryththeman maybe the need to set the right gain with 24-bit comes from the precision of integers at different volume levels. Recording at lower levels, when you then bring the volume up in post, the difference between samples will have less precision than if it would have been recorded at higher gain in the first place?
is it even possible to record the 1/2^32 difference in voltage?
Most of the 32-bit implementations are actually in floating point. Which for filtering CAN make sense.
Yes it can be recorded on digital disk or tape. Any bit length can be recorded. The question is can you capture it?
Many audiophiles conflate bits and sample rates used in the recording process with bits and sample rates in the playback files, this is a mistake. High sample rates and more bits are used in the recording process to enable editing but are pointless for playback.
If you actually had a 24 bit recording with a true 144dB dynamic range and the equipment able to handle it (which is impossible) and your quiet listening room had a 30dB background then a peak sound above background would be 174dB which would be lethal. No, I am not exaggerating, this sound level kills. Even a true 16 bit 96dB recording would be unplayable as the quiet passages would be inaudible or the loud passages screamingy loud. Most modern recordings can be delivered with 10 or 12 bits. A full symphony orchestra giving it all it's got (ƒƒƒƒ) peaks at about 104 dB SPL. Let's give the orchestra 105 dB, and 105 dB - 30 dB = only 75 dB real dynamic range. Most recordings are compressed dynamically to a sensible range so the listener is not constantly having to twiddle the volume control.
Higher sample rates on playback produces ultrasonic noise and does not improve audio in the audible range.
Fact
Only correction i would add is higher samplerates for playback do matter if you desire a smoother sound (less digital grain and grit) or if you have a great DAC that supersamples, that's fine too. DSD playback is my favorite though.
I wonder if there is any song in the world that has real sounds of such high (144dB+) dynamic range.
@@LunarLightLtd1 This is a common misunderstanding of how digital audio works. A 44.1KHz CD can deliver audibly perfect waveforms up to 20KHz, the upper limit of human hearing. No graininess or stair steps. Higher sample rates on playback increase bandwidth above audibility , this is all it does. This is why so called hi-res audio is marketing madness. On the other hand, DSD can produce stair steps and distortion and huge sample rates are required to make this effect inaudible.
@@SanceShaji Probably impossible and highly undesirable. The best electronics struggle to give 120dB (20 bits) dynamic range. A true 144dB dynamic range recording would be unplayable. If you set volume as loud as you could stand on the loud passages then the quiet sounds would be far too faint to hear.
I prefer 24 bits while tracking for the lower noise floor than 16.
Audiophiles be like: “My ears and my opinions have empirically determined that there is no logical scientific reason for the artists or engineers to improving the resolution of this recording” ..Pure nonsense… Many ppl thought like this at every advent throughout the many evolutions of recorded audio… The weird bigotry of saying “it’s good enough for *MY* ears” while simultaneously romanticizing “Audio Nirvana” truly confounds me.. One half of every audiophile is obsessed with resolution while the other half seems to shun it like a vampire in full sun. 😂🤦🏼♂️🤷🏼♂️
i don't agree, during the mixing part the headroom is very nice to have. it make's tracks better to mannage and give space. for the final mix 24 bit is the way to go. after some good dithering
I wish he answered this question by telling me what his produced gear can do.
Are his amps preamps ……. Capable of 24 or 32 bit resolution.
Generally a system capable of 120 db range is off the chart with design and expense. Most people are happy with 105.
I thought the bit depth is not so much about dynamic range but about fine nuances, or greater resolution.
Nope, it's only about dynamic range/signal to noise ratio. Since 16-bit is transparent, more bit doesn't do any good for consumer audio formats. Studios might need 24 bits for headroom when recording and mixing though. There's no more nuance or resolution to be heard by upping the bits from 16 to 24 or 32 bits. Check out Monty Montgomery's explainers, he's the best: th-cam.com/video/cIQ9IXSUzuM/w-d-xo.html and th-cam.com/video/FG9jemV1T7I/w-d-xo.html .
Thanks 🥰
114dB is massive. Also that is very small mV signal, also resistors in DACs become very difficult to be both accurate and stable with temperature at the level of 32bit LSB (we're talking 0.01% or below tolerance and
There is only one dac in the world at the moment that can playback almost full 24 bits, 130db no distortion, it's so good you can "listen" to actual data stream signal. Topping d90se
Is there a comparable ADC?
48K is a natural multiple when no external reference nor compatibility was required. But 44.1K was chosen because it allowed 3 x 2 channel samples into each video line as video recorders were first used. The original Digital Converters like the Sony PCM-100 did not have storage, pro-video decks were used. Often 3/4" U-matic. Or naturally with Sony, later Beta.
And technically electronics can not resolve voltage differences past 22 bits. Even 20 bits is stretching the abilities of a ADC to convert. Nor a DAC to resolve. Even the best ladder DACs.
This space intentionally left blank.
Nice explanation
If someone is listening at even 100 db they will never hear anything under about 10 db unless total silenced 100 db
@Douglas Blake Really. 10 db is not much all in all. hahaha, at least you made my point very clear. Rock on👍
This is one of those subjects you could argue about all day, everyday. Most say that there's nothing beyond 96/24 in PCM, after that go DSD. When they record in 192 it's so they can zoom in as much as possible to cut and edit. So 24bit and 96k is all PCM can do as far as quality. Most say - Don't argue with me personally
People forget that sample rate directly influences latency. It's going to need more CPU for crackle-free playback, but cranking it up will mean better latency, as unintuitive as this may sound. Almost no applications in the box require anything abive 44.1 or 48, but if you're trying to record e-drums into your vst and monitor yourself too, that is going to matter. As far as dynamic range is concerned... ya no, it really doesn't matter. Oh and SR matters when slowing down samples too, for obvious reasons.
Most say 24bit/96k for playback, everything above that is for editing. Every format has a limit, and pressing X2 on a calculator doesn't improve sound quality
And even DSD 256 is the limit, I don't believe there's any more sound quality above that, until they invent a new format above DSD
Correct me I am wrong (and there is a good chance I might be!) , from what you are saying a bit gives you a fixed amount of dynamics ie 1 bit will give you x Dbs while 10 bits will give 10x ? That is to say your max and min levels can be dynamic but you need sufficient bits to express that. My Maths lecturer said to me many years ago that any model that does not work at its limits is probably wrong. So using that premise : 1 bit resolution will only define the lowest part of the sound you are representing using binary, everything else is not defined and there for lost. So also working at 32bit means most of the high end bits are also redundant as there is no sound at that level to set the bits to 1? This would automatically set the limit of resolution you can go to regardless of sample rate. Real music (as I am sure you know) is more complicated then a sine wave. In a real environment sound (and the recorded wave form) will be the superposition of all the instruments ( each of their fundamentals and harmonics), the reflections from the environment. In-bedded in all of this will be phasual (cant spell!) information. If you look at the wave form at a particular point in space and a particular point in time- you will have a voltage detected by you sensor. If you increase you time base from an instant to a couple of seconds you will have info. This at the present is analogue. The fine detail on that wave form will tell the listener about the instruments and where it is ( with reference to another sensor position). If you want to use digital representation of this you need to drill down to the point where it beyond our ability to detract any difference.By increasing the sample rate you can get closer to pulling these nuances out. But is this not true of the amplitude of the signal as well ? If we have a fixed bit to amplitude Db ratio then any voltage fluctuation with in this ratio will not be seen. This will translate into a harmonic/phase difference. Where am I going wrong with this observation ??
24bits is more than enough for playback, but more bit depth is useful as headroom when doing post processing.
I read something about 48khz easier to use for film (video) purposes. Something about easier syncing 48khz stream with the video or tv use. But can't remember the exact details.
44.1Khz/16bit is kinda good enough for hifi consuming..
double or quadruple the rates for authoring or mastering...
yep i agree that DSD is great as a source format. DSD64 is like 192khz source in storage size.
a 4-minute song takes 170MB of space
It could be an improvement for those die hard audio critical listening people. I am not one of those.
Excellent explanation in less than five minutes!
It's off topic, I know, but what is the brand and model of the turntable next to you ?
Thanks in advance, best regards Jan from The Netherlands !
It's a VPI direct drive turntable with the 'Fatboy' gimbal tonearm, possibly the HW-40 Anniversary model, No idea what phono cartridge Paul McG uses, but I read somewhere that Harry Weisfeld (the HW in HW-40) was partial to the Kiseki Purple Heart moving coil.
VPI HW-40 Anniversary
Hi Jan, it’s a VPI turntable
@@stephensmith3111 Thanks for your feedback, much appreciated !
@@louisperlman8030 Thanks for your feedback, much appreciated !
I record in 64bit 🎤💬 & the audio is amazing
32 bit doesn't sound better than regular 16-bit audio for consumer formats. First off, about 90 percent of all available digital music is 16-bit. Upsampling to 32 bit doesn't change anything, it just wastes computer space. More bit doesn't mean you get a higher or better resolution. It means you get a better signal-to-noise ratio. Theoretically 32-bit would mean a SNR of 192dB (not possible today). But what good does that do?
You can't listen to music at 192dB, you'd die. 24 bit in theory means a SNR of 144dB, but in practice closer to 118-120. That is also way more than you need. Everything above 90-95 is transparent (assuming a flat frequency response and low distortion which all competent dacs have acheived at a reasonable price since the late 80's), you can't hear that noise floor if you don't kill the music and max out the volume on your amp. Also, most music we listen to have a dynamic range of 30db tops (lower than that for most pop music). So it's all overkill and doesn't make any audible differences to have more bits. In a professional setting like a studio it's preferable to work in 24 or 32 bits so you'll have more headroom when mixing and so on though.
I feel really big different between 24 and 32, too much ,so I cant back in 24
Another reason why 32 bit float can be better than 24 bit for recording and post-production : with PCM integer audio like 24 bit, not all the bits are equal, and the higher bits are more significant than the lower ones. You need to record louder for getting better quality, but you have a hard ceiling. For example, in the early days, when recording in 8 bit, the loudest signals in the maximum 6dB range had only 127 possible values. The quieter signals had even less possible values. Needless to say, the quality was very bad and full of quantization noise and errors. With 24 bit audio, even when above quantization noise, the subtle but important signals like natural reverberation are badly digitized if the recording level is too low, and brought up later. Let's say, theoritically, that no important signal should be quieter than the 9th bit. The 9th bit can have 255 different values for that low 6 dB range, and it may be enough for recording the subtle, natural reverberation. If you record properly at 24 bit, with peak levels under -12dBfs, then you already don't make any use of the two most significant bits, the 23rd and 24th, which make for 3/4 of the datas in the file, and the usable range left is around 13 bits. That's 78 dB of usable dynamic range, which is quite high, but may not be enough. With 32 bit float audio all the bits are equal. Please listen to the demonstration here : www.sounddevices.com/low-signal-32-bit-float/
i hear the difference, but hey, i happen to own the Air Reference 2 from Millon
If I understand pcm correctly, 24 bit is 12 bits per channel. 16 is 8 bits per channel. I think the confusion is everyone thinks more is better. Not necessarily. Look at DSD. It’s only 1 bit but the sampling rate is about 2 or 3 MHz! Of course DSD is PWM digital, pulse width modulation. Not PCM - pulse code modulation.
16bit PCM audio recording means 16bit audio per channel, regular audio CD is 16 bits x 44100Hz x 2 channel = 1411200 bps
DSD64 has a data rate 4 times of CD, 5644800 bps
For reference to Paul:
# bits SNR
8 48 dB
16 96 dB
24 144 dB
32 192 dB
Speaking about not knowing what to do with dynamic range:
Don't you love how the automatic dynamic range limiter starts to fill the silence with background noise (like at 2:58)?
i can't hear it
Curious as to the lowest reliable frequency on can get online? I am thinking about some GOOD Towers that go down to 30 HZ + to - 3db. Had a pair before and miss the bottom end on some music. Thanks
You need a subwoofer to augment your mains.
@@AnimusInvidious Not. Why would anyone want to hear much of anything less than 30 Hz, mind u +/- 3 db so likely down to a realistic and solid 25 Hz. ??
It's not an on/off thing. There are slopes, and most speakers slope off before the bottom-most reaches of human hearing. The sub fills in that gap.
@@AnimusInvidious Well high quality speakers that are flat down to 30 Hz don't need a Sub. The speakers I am talking about are Tower Studio Monitors that in todays money would retail for $5,300 US with a shipping weight of 100 lbs each
Would still be optimized with a sub to fill in the < 30Hz region, imo. But you do you.
So. For wild recording, if you have a true 32, or 64 bit, or 32/64 floating bit recorder, you can never ever clip on the recording. In post, you can make a whisper, and an atomic bomb, be normalized.
Here we go again with not really understanding how things work.
The bit depth determines the resolution of waveform capture at an instantaneous point on the time domain. Therefore, the higher the bit depth the greater the waveform resolution. So yes 32bit contains more detail. And the faster it collects them the better. Because everything in between get electrically averaged out.
Bit depth determines total dynamic range (as well as resolution). The higher the bit depth, the higher the possible dynamic range as well (and here's where you're correct) the resolution as well.
@@Paulmcgowanpsaudio Digital photography bit depth analogy: dynamic range=#of colors&saturation (ie intensity) For audio, dynamic range is frequency&litude thats' it. When desiring and dealing with dynamic range captures this microscopically detailed, reproduction equipment slew rates are critical.
We go to further and further lengths to extract more from our sources, when it's the sources that are lacking. If it wasn't captured, it's not there. The best microphones in the world are still lacking in a major and obvious way (at least to me). I am working on a couple new concept microphones to fix that if you're interested to listen under NDA but willing to share. :)
32bit are intended for production not for playback.
Oh, and 32-bit is not completely useless - it can be useful for situations where the required dynamic range is not known in advance, such as documentary filmmaking, or even non-documentary filmmaking in a scene where you don't know how loud the actors or other production sounds are going to get.
Dolby tracks are 48khz
and now prove that you are actually resolving 32 bits of data with either an ADC or DAC at audio sampling rates
Id expect the dynamic range of the bit would be more accurate.
In order to maintain compatibility with 16 bit, no can do. The gradation of each bit has to be the same, so only the noise floor gets lowered.
The best thing about 32 bit is working with it. No digital clipping!
You can still clip 32 bit, just turn up the gain on the recorder. The improvement in S/N is on the quiet end, not the loud end. Beside the noise floor of a wire is higher than what 32 bit will yield, even 24 bit is at the limit of what can be achieved in noise.
@@Baygul318 I'm talking about clipping inside the DAW, not the analog signal.
Theres also lots of babble out there on "noise floor" for those who dare to research. Longer DWORD length = lower noise floor.... in theory.
you should also look at nyquist and sampling.. faster rate means you could shift the noise up then filter out.
Blimey Paul - you don't mention DSD once in this video!
96khz is new audio standard because 48khz is broadcast video standard
Repping San Diego 👏
Headroom!!!
My house has one of these rooms too 😆
Please do not let your hand rest inside that component! I was scared the whole video you would do that while explaining and electrocute yourself!
24 bits is already more than enough for capture , transport, and playout. Just as 96khz is also a practical limit. The three LSBs are just noise. 32bits or more are needed in DSP systems due to the multiplication needed to alter gain and frequency response.
Very good expectations Paul. Finally someone to talk straight, and real engineering, for the audiophile tribes. Congratulations 🎊
Much appreciate for the expandable answer
it's word length
yes.
I'd like to see 32 bit with a 3 db per bit codec
Good video, Paul. I like the VPI 40th Anniversary turntable you have there...............forget the bits!
MASSIVE MISUNDERSTANDING OF THE DIFFERENCE BETWEEN 24 BIT FIXED VERSUS 32 BIT FLOAT! Everyone needs to do their homework here. The float is an extra 8 bit internal exponent, that has nothing to do with ProTools converters and only applies to the internal mix bus. All converters for ProTools are 24 bit and automatically converted to 32 for the internal mix bus; it is pointless to record 32 bit float files. However bouncing 32 bit float files for mastering provides the mastering engineer more headroom, but all converters are fixed converters and ONLY decode 24 bit. (There is an exception given that Steinberg has recently released 32 bit fixed point converters, but these do not apply to ProTools and are not float files).
32 bits is totally unnecessary. 16 bit gives all the dynamic range you could ever use. The dynamic range of 32 bit would be like from one molecule hitting the eardrum up to the power of the big bang.
Record in 24, process in 32, playback in 16.
Today's music has no dynamic range so who cares?
I just hope that James does not end up equalizing and compressing his creations, like nearly all engineers do (it is like a drug that makes nearly all recording engineers climax to vandalizing sonic works of art).
If James does end up equalizing and compressing his creations, then his question is pointless, and he might as well stick to the .mp3 format while dining at McDonalds.
Ahhhhh bit wars Sega does what nintendont
oh,
how tf did i end up here....can some one dumb it down?
💪💪💪👋🙏
even the noise on a battery - which is very stable - is louder than -144dB.
that means, nobody has a power supply that gives you noise free power for you amp circuits to reach -144dB accurate signals. so, 24 bits is accurate enough
no
32 bit is better but industry hasn't developed much technology doesn't have 32 bit audio card (max 384 khz/24 bit)
Is driving 248 Mph better the 186 Mph? What purpose would it make?
@@obsprisma
Yes, when your being passed at 186 mph. 😁
@@obsprisma Poor analogy.
please record in DSD......
Since DACs cant resolve past 21 bits, 24 vs 32 bit is moot. Besides, the rest of the equipment in the recording and playback chains all have S/N ratios higher (worse than) the S/N ratio for 24 bit (144dB), so still pointless. 16 vs 24 is worth it, but 32? Nope.
The entire range of human hearing does not require anything wider than 16 bit sampling.
Bit depth is about volume, not frequency. Are you able to identify which is louder, A car horn & a jet engine? If so that means 16 bit is not beyond your range..
Gregf, you know nothing. You should investigate why they prefer to publish 24 bit in classical music recordings where multiple microphones are used.
Just ask yourself: „would I rather have a 24 year old girl or a 32 year old. So you’re able to easily remember the essence of what Paul really well explained 😎
Copy Paul
First.
Congratulations
@@graxjpg
Congratulations on your congratulations. 😁
I disagree. I oversample my system to 768 bits and definitely hear a difference.
Ofcourse you do 🤣🤣🤣
I prefer 666 bits 😁
No reason to go to 32 bit FOR WHO. If i buy a 32 bit song or mix with 32 bit. Just because you dont need it in theory doesnt mean the 32 bit file isnt the better music file. This is never discussed.
Actually, standing next to a jet engine is more like 160+ db Spl. Ask any aviation mechanic that tunes them. Need kidney belts. Just saying…