Yes - I really appreciated that you were honest about the title of this video right away! Great video. It would've been cool to get comparisons of time-stretched audio between 48 & 96k, and how it would translate on TH-cam 😄 Thank you for your awesome work!
@@Liio.ChantelI could be wrong but as far as I can recall, certain DAWs (if not all) typically resamples audio to a higher sample rate when editing (ie. time stretching) to counteract artifacts.
I use 44100 for 25 years in electronic music production and recording. Clients never complained, Awards are here, support from Tiesto and Armin was in 201X's. Also mixed some Big Stars stuff - very often I have got 44100/48000 16/24 bit sessions. It doesn't influence a hit record. When I am listening to 60-70s records it is very often sounds like shit in terms of quality. But people don't care. From technical point of view, higher sample rate means more information for processing (pitching and stretching) but I never had a problem with 44.1 material. It is just sounding different when processing 44.1/96 material. Also I have done a blind test of 44.1 VS 96 of the electronic mix done entirely in the box. More people voted for 44.1 which sounded a little bit grittier. Now I track my hadware synths into 96K FLAC in case I will do hi res mixes for my fans. Many years ago I have mixed a song with 2 vocalists. One is Big World star recorded on SM58 and other is young vocalist recorded in top notch chain with M149. The sm58 recording was phenomenal and m149 was awful. I think big names just CAN afford analog or 192 or 96. But if CLA will mix in the box at 44.1 it will be CLA mix and no one will care about how he mix it.
Finally someone with a brain. 44.1Khz is sufficient to reproduce the original audio without any loss. 96KHz will allow you to sample up to 44.8KHz frequencies without any loss, as per Nyquist Theorem. The thing is that the human hearing frequency range is 20Hz to 20KHz. Unless you're making music for dogs it doesn't make any sense to use 96KHz.
Ultimately it comes down to performance. In descending order Performance Instrument Mic placement Room Mic pre / analog chain Bit depth Converters Bit rate
One of the easiest ways to demonstrate the difference in is video frame rate. The afternoon soaps were recorded and aired at 30 frames per second, while the film-originated prime time shows were recorded at 24 frames per second. Intuitively, one would think that the higher frame rate (sample frequency) would yield more detail and hence a better rendition. In practice, it turned out that it was far less pleasing. It felt cheaper because it removed part of the 'unrealistic coloring' that cinematic productions have by nature. I like your tape analogy as well. Odd as it may seem, the audience apparently doesn't want productions that are too accurate.
I like 96k because you can play with slowing things down a lot. Outside editing purposes I don’t hear the difference other than some aliasing artifacts sometimes.
I've always recorded at 96k/24bit with the idea of "future proofing" my stuff. I've also yet to come to a point where I've needed that "future proofing". I find 96k noticeably cleaner for time stretching events and sample work. Having "more" data to work with while manipulating. It's not the exact same, but I've always looked at it as 1080p vs 4k video (8k being 192k). With 96k and 48k there are advantages and disadvantages to both, I've always built my recording rigs with the plan to record and mix with 96k from front to back. But in the end as long as someone is happy with their mix/product, no matter how they got there, who cares! Cheers! Great video as usual Colt!
I’m realizing that since we don’t use compact disk anymore my method of recording at 88.2 so I can mix down to 44.1 is obsolete now. I believe everything has moved to 48 24 bit over 44.1 16 bit.
@@stevewoodyt 24 bit for sure. CD is still very popular globally actually. Whether it's audible or not, people do want hi-res audio from Apple, QoBuz, Spotify etc, so it's good to be able to deliver 96/24 and take advantage of that.
Actually, 96kh doesn't increase accuracy at all for each sample. It only doubles the amount of frequencies from about 24kh to 48kh. But, at 96kh, when processing you get increased quality when pitching and reduced aliasing as well as reduced need of added filters implemented in plugins that automatically oversample for certain stages.
Thank You! It's what I learned in audio school too. If you only record and don't do any digital processing, our ears shouldn't be able to hear a difference. So I feel a bit confused when I see this and read the comments. As long as aliasing and anti aliasing filters are not involved, why would you hear a more accurate recording?
@@magnusboder6680 I believe that it comes from the misconception that higher sample rate adds samples between 20 and 20Kh. It doesn't because two samples are enough to represent a perfect waveform when going back to analogue. Sorry for not being entire clear, I'm not an expert in digital theoremes but you could review Nyquist-Shannon's work to understand that better.
@@raphaelherzig3316 You are absolutely right. And I learned about this with Nyquist crossing and aliasing. But again, it only applies to digitally processed audio. And in the video he talked about a live drum recording and the accurateness of it's recorded sound. And those words made some people thinking about recording in a higher sample rate and than mix in a lower to save cpu load. And this is just... Craziness :) So that's why I decided to write, even though I'm not native in English :)
It sounds like when filmmakers such as Peter Jackson decided to double the frames per second on movies, and all of a sudden it looked “too real,“ and also “not like a movie.”
Sound is continuous as well as visual is continuous in the analog world. Sound is sampled at, for example, 48kHz while video is sampled at 30 or so fps. As I've said for years the illusion of sound is that it can present the illusion of stationary solidarity while the illusion of motion pictures do the exact opposite - present the illusion of continuous movement with a rapid succession of "frozen" images. How the brain interprets the signal-images from the human transducers ears and eyes is another topic. Compare the flicker rate in visual where the image is not seen as continuous to the much greater sampling rate in audio 48 K for example.This discussion can carry on past a bottle of Jameson easily.
Ideal test (in my mind, at least) - Track one song with mult splits of each channel, one going into a converter at 48khz, one at 96khz. Mix the 48khz fully in one DAW session, print mixdown, duplicate session, change project sample rate, swap audio files to the 96khz samples, print mixdown - compare.
Everyone talks about the end format when comparing. No one talks about how effects like reverbs especially, sound better at 96k where it can time slice the decay to a more natural sound. I started using 96k since PT latency is reduced per Avid's specs. HD disks are dirt cheap today so IMO 96k is the sweet spot.
Good vid, but.. I don't think it works to use a cross-sense example: the experience of eyes and ears aren't the same. Like, does it make sense to say "I like the smell of avocado, so therefore green must be my favorite color!". No! :) Also... many say that true analog is the best (like no digital in the process: tape, to say, vinyl)... that path is effectively infinite samples per second... (or what? The space between electrons?:)... a non-approximated capture of sound without "snapshots".... which is closer to 96k than 48k. It's mind if all taste at the end of the day. Do what you like!;) it's all fun to debate though... I switched to 96k a few years ago and noticed it feels more detailed... but it's really hard to know for sure what's "better"... I would rather add my own color shaping with saturation and other processing to achieve "vibe" than have it come baked into an underlying element like sample rate. But still; not sure. I don't mind a head start... if I'm oil painting a night scene, would I rather start with a black canvas than a white one? Maybe. Maybe not.
Very interesting take, Adam… thank you sir for sharing. Laz, I could play you anything ran through standard DAW reverb plugins at 44.1/16 and you’d never know the difference. Your comment about hard drive space not being an issue, when touting such a high and silly sample rate makes your overall statement difficult to take seriously. If you had a 1/4 of the sample library I had… and it was at 96, you would change your tune in a hurry.
I used to do a lot of rap production and they would have me do these chopped and screwed pitch shifts down super low and I found that 96khz was really good for that because you kept the high frequency information intact
But.... microphones don't actually pick up that information... a microphone capturing anything audible above like 16 or 18khz is honestly pretty rare.... you need like microphones made for scientific purposes for that usually. Do you mean like the synthesized instruments, because that might have some information that extends beyond nyquist.
@@Linguae_Music I mean vocals, on a microphone. I dont know the exact science behind it but there was a very noticeable difference so if you're skeptical try it yourself! It might have more to do with stretching the digital audio than the analog characteristics of the mic and the preamp
Sampling at 96KHz might provide more samples/second but, ultimately, it determines the highest frequency you can record - half the sampling rate (Nyquist). The human ear limit is ~20KHz. 44.1KHz/48KHz captures everything we can hear. My 2c: Use the frequency that's appropriate for the target medium: 48KHz if the target is video, 44.1KHz if the target is audio (CD, streaming audio).
I try to work in 96k as often as possible. When I was going back and forth with the pros and cons (some of which were discussed in this video) my ending idea was thinking that a chef dosen't prepare a Waygu steak with a steak knife because that is what the end user will experience. He has to work at a higher fidelity. So I do as well. I can always smear transients later if I want, but you can't un-toast bread.
So, you don't have any technical arguments for your choice? A steak doesn't get any better if cut with a knife that took Japanese mermaids 8 years to make..
@@hansemannluchter643 correct. Because that's in your ability as a professional to make whatever you get the best it can before it leaves your desk (or kitchen as per the analogy). My point was I want to work at the highest fidelity possible. I've found that any sonic benefits to a lower sample rate can be replicated with either hardware or plugins.
Confirmation bias is the tendency to search for, interpret, favor, and recall information in a way that confirms or supports one's prior beliefs or values.
I tend to agree with Dan Lavry's thoughts on conversion. It seems quite well thought out that the ideal sample rate would be 60khz or so. From his "white papers": "Good conversion requires attention to capturing and reproducing the range we hear while filtering and keeping out energy in the frequency range outside of our hearing. At 44.1 KHz sampling the flatness response may be an issue. If each of the elements (microphone, AD, DA and speaker) limit the audio bandwidth to 20 KHz (each causing a 3dB loss at 20 KHz), the combined impact is -12dB at 20 KHz. At 60 KHz sampling rate, the contribution of AD and DA to any attenuation in the audible range is negligible. Although 60 KHz would be closer to the ideal; given the existing standards, 88.2 KHz and 96 KHz are closest to the optimal sample rate." I tend to work at 88.2 and 96. I work on classical, jazz and acoustic-heavy music for the most part. It can bog down the system for sure. Though it does help when mixing to avoid aliasing buildup with certain plugins that I like that don't oversample. All of this is relatively small fries. Some of my favorite records from the analog days are noisy and messed up sounding anyway. Then the earbuds argument. I don't really do this for the earbud crowd, though. I do it because I love great audio and art.
"smearing" ain't always a bad thing. It's been part of the allure to analog since, forever. I've been really happy with the small move from 44.1 to 48 myself.
You might wanna try doing a blind test because that’s not how digital recording works. Nyquist states you only need two sample points to accurately reproduce a frequency. A larger sample rate just means you can record higher frequencies. It’s possible you had some audible aliasing but I highly doubt that. Aliasing usually sits way below the audible range unless you’re trying to create it. Most likely you were being tricked by either volume or your eyes or both. Set up and ABX test and correctly pick out the 96k recording from the 48k recording 10 out 10 times and then you can say you can hear the difference between 96k 48k. Haven’t seen anyone do that ever in my life though. Most people can’t even distinguish a 320 Kbps MP3 from a lossless file.
88.1 or 96k means you may not need oversampling as much IMO and latency is cut in half generally. For me the thing that makes me want to record at 96k is the more natural sound especially for acoustic music. UAD and others admit clocking is different between sample rates on some devices. My current interface (mt 48) and prior one (Antelope) sound quite different in the top end especially when you go between 48 and 96. Unfortunately it's not as clear cut as 1 is always right.
You hit the nail on the head for accuracy. The music I do requires it and at a full final mix, my CPU is always barely 50%. Because I don’t use many plugins anymore since I’ve been able to collect the 20+ year analog setup I now have. Even with pop country or doing rock, the converters I have and my mixing approach sounds better and more defined from the get go. So, I’ve lived there and haven’t changed for many years now. When I eventually change converters, as nothing lasts forever, I’ll listen again and decide then what I’ll do. Quality hard drives are crazy cheap now, so, storage isnt really any deciding factor for me. Great video and I appreciate the time it takes you to make them. I wish I had the time to do the same.
I literally clicked on this knowing it was click bait. I use 96k because latency is halved and reverbs seem to blend better with the source material. As someone who shoots music videos , book trailers and TH-cam stuff as well as runs a recording studio I can assure you that I run out of SSD space in a days of video work. On my Audio computer and external HARD DRIVE I haven’t come close to filling it yet.
Great thing about 96k is that you get half the latency while maintaining the same buffer size. This is great for real-time monitoring of kick triggers live, neural DSP or if a vocal is going through a daw and echoed/input monitored. :) for a live show 16bit at 96k is a great way to go if low latency and super reliable performance from the computer and interface is desired. It gives the computer much less to process since 24bit has a lot more data to process/write than 16 bit.
You seem to be talking about budget conscious sessions. Storage is so cheap now. I record everything at 96/24 on a quite affordable M1 MacBook Air, with multiple plug-ins and never have a problem with the processing, or slowing down. Accuracy? Yeah, I want my (rock) drums to be as accurate and high quality as possible. It's not just for classical. A lot of clients I work with demand 96/24 - which is the main reason I switched from 44.1/16. In the end, working at 96/24 does not impact my studio work one bit - there are no cons. So I do it because that's what I'm asked to deliver and also because I want to future proof my work.
Same as other people said, the intro was Super appreciated, No one actually likes clickbait it just gets more clicks due to the algorithm, but this vid was Great
Your sincerity and how you're expressing your prevalence towards 48k is really helpful. Because I've listened to how you mix in your videos, I trust that you're coming from a place of time-tested wisdom that works for your ears. And of course, you mentioned 96k for more dynamically sensitive music like bluegrass and classical, so thank you for pointing out that difference.
I would like to see more science on what you're hearing, whether conversion quality during downsampling, or ears being tuned/used to enjoying the buildup of more foldback aliasing distortion at 48khz (very little, but some and more than 96khz). I also prefer 48khz, and with the advancements happening with antialiasing algorithms, It's likely never going to be a problem. Another thing to consider is, it may be the same reason people love Soundtoys decapitator which has no oversampling and a TON of foldback aliasing distortion: Ears like what they like. If it sounds good, it is good.
Ngl, I almost didn't watch this because of the click baity title but I'm glad I did. I do a lot of sound design work and 96khz and 192khz are regular daily use sample rates for me, specifically because I do some much multi-octave pitch shifting and time stretching. Working with higher sample rate files really goes a long way in preventing artifacts and improving intelligibility, like you mentioned with Elastic Audio. Whenever I do music I pretty much exclusively work in 48khz. I think high sample rates are a specific use case and there's not enough education out there on why and when to use them. Same kind of argument with 32-bit depth. 90% of people do not need to record at 32-bit but if you have a job like mine and have to record things like fireworks or guns, having a 32-bit capable recorded can be a godsend.
+1 for stretching. I recorded myself blowing an Aztec Death Whistle on a Sanken ultrasonic mic at 192k and stretched it. Sounded like satan screaming and the quality was amazing!
I’ve been recording at 96k, only at the advice of my last producer, for about 10 years now, and there are some valid arguments around CPU usage, less so now though. I can only imagine how much faster computers will be in the next 10 years and beyond, as computers get faster, lossless 96k and beyond becomes easier and easier… As much as the “vibe” is concerned… that’s possible. It’s definitely not for “no reason” that I use tape emulation when mixing, I like the “vibe”, or “tapey” quality of how it sounds. I definitely might run my own experiments. It would be cool to choose between a 48k vs 96k song, but it would have the be the same audio in both examples, recorded on two systems simultaneously, or there’s too many variables when deciding what’s better. I won’t deny it’s possible though, there could be a “vibe” to 48k that I’m missing… There’s also a lot of research done around aliasing, which is caused often times by saturation plugins that don’t have oversampling features. I haven’t tested this extensively, but it is another reason why I tend to lean 96k…. It is interesting…
Really interesting - Thanks for making this, I was working in 96 and like you said 'the realism' is there. I'll experiment with 48. And the whole thing about 'no ones gonna hear it' argument - what i've always said is the consumer wont hear it but they will most likely FEEL it.
I used to capture my sessions which are typically comprised of live-tracked drums, guitar and bass through various interfaces, and vocals through preamps loaded into usb 500 series racks with hardware inserts on effects solely in 96k and I was very pleased with the clarity and accuracy; HOWEVER, the need for more and more storage and constantly running out out of drive-space became more than I could bear; but I kept trying to saunter on. It wasn’t until I acquired some vintage gear that I added to my hybrid system that topped out at 48K that I said what the heck, let’s see how working in 48K is and I have to say the negligible sonic differences and the ability to store a greater number of projects to one drive has me working totally in 48k now without feeling like I have compromised anything. Totally worth it.
Hello! I don't work as a mix engineer and my english is not very good. I have a degree in audio and acoustics though. And now I am a little confused. Both from some information in the video and from reading the comments. As long as a sound is just recorded, and no digital processing is involved; aliasing and anti-aliasing filters. There is as far as I know, from a scientific perspective, no way to hear a difference with human ears. Now I see in the comments that people get the idea to record in a higher sample rate, and down sample to 48k for mixing... Only the other way around could be benefitting if you mix with plugins. I mean according to my knowledge. If you heard a difference, it may be because of your converters. But this is individual for every kind of converter and has nothing to do with sample rate. I really don't try to be a rude idiot now. And I have been wrong about things before. So feel free to correct me everyone if I am wrong! I really love to understand how things works and wish to have acurate knowledge :)
You are right. Your confusion is justified. There are apparently great numbers of people in the music and recording industry who still do not understand digital audio, particularly resolution and bit depth. What you have learned is correct.
There is a very good chance the preference he feels is how his AD and DA converters handle the different sample rates. If he had different converters, he might feel differently.
The higher resolution will help when working with multiple tracks. When you combine your 96K master tracks this will help when recording the lower resolution final two-channel product. I'd just say to each his own. After all there are artists and producers who seek out vintage analog equipment (old mics or old amps for example) in search of a specific vintage sound. Although 96K would aid the mastering, in the end you have to pick what suits your ears.
I don't think this generalization holds, and it may be specific to the converters in your interface. I'm in the middle of recording an album with a quartet of stringed instruments. We've mostly been working at a studio that's been recording at 24/48, with some really botique preamps, and high-quality mics. Some of us have also been bringing our own (AKG & Telefunken) mics to sessions. There were a number of complaints about the sound quality we're getting at the studio, so I decided to give it a try myself. So I recorded a session of the group, using a MOTU 828es with a DAV BG-8 preamp, into Logic at 32/96. We used two pairs of C414s, a pair of Telefunken M60 FETs, and a pair of Beesneez Lulu FETs. Four of the 8 mics were exactly the same mics we had been using at the studio. The recording I made at 96k just sounds so alive and 3-dimensional compared to the 48k recordings from the studio. If you just do a direct A/B comparison, the 48k recording just sounds flat and lifeless in comparison to the 96k recording. I really think that while some converters in some interfaces say that they can do 96k or 192k just fine, that they don't actually do as good a job as others at those sample rates. For what it's worth, the ESS 32-bit converters in the MOTU seem to have no problem at all delivering consistently preferable results at 96k.
I work in 88.2 because it's double of 44.1 which is where a lot of masters end up at (CD Baby still only accepts 16 bit 44.1 kHz). And it's still high enough to do some elastic audio stuff.
Honestly I don't feel there will be a huge difference between 88.2k vs 96k because it's the difference between a hand grenade and a rocket. There might probably be a case out there because there is ALWAYS a damned edge case, but it'll be rare.
Colt, might I suggest a follow-up video about the Nyquist theorem and how resolution only determines how high an audio frequency can be recorded? I am surprised at how many people in the recording industry still don't understand digital audio. Higher resolutions are not more accurate, not cleaner or more detailed. Higher resolution merely makes possible the recording of higher frequencies. 40 kHz is capable of capturing 20 kHz sounds, the theoretical limit of human hearing. I'm simplifying, of course. There are technical reasons for using 48 kHz. Bit depth is another issue. It determines the number of amplitude values that can be assigned to each sample. Accuracy and clarity are, in fact, affected by bit depth. But 24-bit recording makes for such a low noise floor, it has been adopted as an industry standard.
FINALLY!!! someone talking about the math behind nyquist theorem in the comment section! I was staring to worry a lot , thank you so much for talking about that. I just can’t understand how everyone don’t understand a basic math theorem.
Yep. It's amazing to me how much this simple concept is overcomplicated. Sample rate controls max frequency you can store, and bit depth controls the quantization noise floor. That's it!
I started using 32/96k a few years ago because of time stretching (obviously better) and I felt that distorted guitars sounded slightly smoother and less harsh. I record 6K RED footage, so storage space on audio is minimal compared to that, but you work on way more songs than I do. The biggest downside to me is processing, but I’ve heard it said some plugins work better at 96. On my computer it “feels” about 50% more intensive on the computer (may not be the case though). I think its the kind of thing where find what works for you and go with it, like DAW choice. I seem to remember JJP saying he prefers 96, I wonder if some people just have different tastes and naturally prefer one over the other. I also wonder after being mastered from 96k to 48 for streaming, etc if some of the clinicalness would go away. Good topic to ponder. Thanks Colt!
Idea for the A/B test: Put a transformer isolated Split after all your preamps and send each into different converters as you track a song, and then listen to the summed mixes differences with the same plug ins and session settings on each
That’s exactly what you would have to do. But even just having two completely separate rigs that were identical is more trouble than I can convince myself to go through.
@@ColtCapperrune Oh for sure! I aint gonna do it either. That then opens complexities of different plug-ins's sample rate handling, non-linearity, dithering, and then mixing down to lower sample rates in mastering. Too many factors to isolate. Just having the convo is enough for me-- earnest and insightful as always my man!
use dual active split. a transformer isolated split will show differences from the transformer. sample rate differences doesn't affect the waveform, they will null in the audible range. 96k recording brings in higher frequencies that will cause more aliasing through processing compression and saturation. use 48k and OVERSAMPLE as much as possible in your plugins.
@@joemaymusic and not to mention: what it does to you as a musician/producer while the stage of recording, listening to different sounding stuff will inspire you to do different things.. that psychological aspect of the creative stage affected by that technical difference.. you will never be able to isolate that aspect…
Higher sampling rates than 44.1/48KHz are mostly pointless and its a myth that rates that are higher are more accurate, pretty much all you do when using these are recording sounds that the human ear cannot hear, 44.1/48KHz can perfectly describe sound in the range of human hearing. For timestretch there can maybe be an advantage and you can potentially get lower latency with your audio interface.
it seems no one noted here, but every converter could sound different from another one when changing their own the sample rates. I remember many years ago when did a comparison test using 2 separate rigs: mics> preamps> splitter> into 2 separated Rosetta800 each one connected to its Macbook rig, one session set at 48 the other at 96. We repeated the same experiment with 2 Motus 896, and well, me and my fellow both preferred the 96k session from the rosetta800 BUT the 48k session from the Motu 896. If you compare same recording at let's say 48k and 96k with a certain converter, if you do the same with another one the difference between 48 and 96 could be clearly insignificant or more evident, also considering the nature of audio signal you're going to acquire. As already said by many of you, most of the tangible perception of the sample rate impact it's into acoustic instruments and/or very complex interactions with their ambient. Then probably the music genre comes in play where also Colt says about the"vibe" he preferred. This when recording, then when mixing/editing etc and you modify the starting material with proessing plugins etc, it's another story. m2c.
Q: Do more realistically lifelike recordings sound “better” when doing music production? Or are we all reaching for vintage, high-coloration, high saturation analog devices or digital emulations of them? It’s the same as with filmmaking. Do filmmakers want their movies to look like high resolution “video”? Or do they go to great lengths to achieve an altered and artistic look to them? I think most of us already have the answers to these questions, eh?..
The only place where I've found high sample rates to have objective advantages is where you have to do extreme time stretching and pitch shifting (which, for me, was making creature sound effects during college exercises). 48k is more than enough for music.
Thanks again Colt! I have been playing with sample rates as well. My experience seems to match yours. I recorded an album at 192K, a classical, choral work. My mentors sort of trailblazed that for me because they track at DSD rates, so around 2.8MHz or 5.6, I believe, 1-bit. They do this because the idea is that this is "as close as one can get to analogue" since the sample rate is so high, the digitization of the soundwaves is arbitrarily close to analogue. It works well enough, but I also have taken to doing more and more similar genre recordings at 48Khz. This may be totally subjective, because like your own post, I have read other people who are top sound engineers say that 48Khz is all one really needs and that it is the best rate. For myself, I found that my masters of my 192Khz CD sounded warmer at 44.1 (CD rates) than at 192, and I could hear certain issues much more clearly at 44.1 than at 192. I don't know that the sample rate changed anything but I noticed the problem at 44.1 playback and it seemed a bit smoother at 192. Since our print for CD is at CD rates, of course, this is why I paid attention to it. I have since gotten into the general pattern of recording everything at 48Khz / 24 bit. I appreciate your reference to 96 as "clinical" - and I agree. It feels more real at 48, warmer, actually, which is where I am trying to get with this choral work. 24 bits give me a lot of headroom and so it just works. There is one other practical reason for 48Khz rates too. Video recorders, especially those on iPhone or Android devices default to 48Khz for their recorded sound. While phone mics are certainly far inferior to my "real" mics (ranging from Zoom H6 to DPA 4006As), the identical sample rate does two great things for me: It eliminates video-audio drift if I have to edit Video and Audio together, and in one situation, the Zoom refused to play with its SD card and I lost the material but the phones got it so I could reconstruct something (not great, but workable) from the phone mic arrays). I think it comes down to "feel" over "math". I don't begrudge recording at DSD rates - but I cannot afford a HORUS / HAPI to do it with right now, and I CAN meet my field recording gigs really easily with my field kit - the aforementioned Zoom H6 with either its own mid-side mic and / or higher-grade outriggers, or if I feel miserly, some Karma Silver Bullets for outriggers. (btw, these Karma Silver Bullets - they are no longer in production but they appear on eBay from time to time. I have nine of them. They are tiny omnis - about $20.00 apiece when they were sold, but talk about bang for the buck... they are really, really good and they can give my higher-end mics some level of competition that would surprise a lot of folks.) All the best to you!
When 96k came to pro tools it felt like 30ips 24 track: the biggest difference was …. The vocal track. So clear so pretty. I am obviously very old and remember recording reggae and rock at 15 ips! The latency at 96 is so much lower and it is a joy to work with a pretty sound. I do everything at 96 except Dolby Atmos mixing.
I appreciate the honesty with the click bait stuff. Love your videos. If you're a musician who writes and records their own music without a producer, 96k opens lower latency settings on many interfaces like your Apollo. If you're a digital drummer or finger drummer who plays virtual drum instruments and is sensitive to latency, going from 64 samples to 32, opens up room for a few plug-ins while still playing in real-time. Go look at the latency figures for all of the UAD plug-ins for your x16, a large number of the plug-ins give you lower latency when in 96k vs 48khz and this is on top of the fact that 96k opens up 32 samples. The higher your sample rate, the lower your latency. If you're working mostly in the analog domain via a console, you want 96k. In analog land, your console provides the color, grit and vibe, so you want things to be accurate and transparent when converting to digital. You have 2 newtons, why not use SILK on the master buss to dirty it up rather than using 48khz? I personally want my recordings to sound exactly like what I was hearing while still in analog land.
I honestly cant hear a difference when real blintesting, i recomend anyone to do a blindtest. And you must guess the right option 5 times in a row, thats how you know you can tell a difference. The only thing higher sample rates do is raise the extention of the high freqs, it is not at all more accurate on the frequencies both share. This can be scientically proven. More sample points does not mean a more accurate reconstruction. Also idk about accurate sounding more clinical, doesn't make much sense, more accurate should just be more equal to straight wire to the speaker. But doing Time stretching and infrasonic recordings would have an advantage though
"This can be scientifically proven. More sample points does not mean a more accurate reconstruction." So as I noted above 4.8 samples per second of a 10 kHz wave will be as sonically accurate as 96 Khz sampling. Right?
4.8 samples per second? idk what you mean. But a samplerate of 16k per sec for example will perfectly reconstruct up to 8khz. just as good as 192k would for example. thats just how the physics work. main difference might be in the antialiasing filters in converters, which all have a lowpassfilter. Which may reach down to a lower frequency on a smaller samplerate. But I find it strange when ppl say lower samplerate like 44.1 will sound "granier". If there is a difference, it would more be like a tiny bit more muted on the topend/air. But unless you can hear 18-20k and over i doubt youll hear any difference that matters. Atleast thats what Ive heard, and when I did blindtests.@@raymota4515
I was able to get ahold of some stems for a few songs by The 1975 from their 4th Album, "Notes on a Conditional Form." They use 96khz, and it sounds really crispy!! But I truly don't think THAT is the reason for it. I still use 48khz, and with Dolby Atmos in the picture, thankfully 48khz is an option for it. I can't imagine the file size's for a 96khz Atmos session...
If you were to try another experiment in this vein, I would suggest getting a split snake so you can record the same mics and performance on two identical systems. The other thing to point out is that most high profile live sound consoles, that are digital, are sampling at 96khz. I have noticed a difference, in the fact that I’ve worked in one venue that went from an analog console to a Dlive. I did notice a difference there. But it would’ve been leagues ahead of the analog console compared to an m32.
Very interesting, thank you. I recall one fella who always wanted to record his drums at 7.5IPS and then lay them onto another machine at 15 or 30 because he so liked the sound at 7.5. Same kinda effect I guess.
Hey, we all have preferences, but it’s interesting that Avid, Allen & Heath, Midas, Yamaha, & all the other major live sound console manufacturers chose 96 kHz. Also, the plug-ins don’t all really scale like that. The first time you load a plug-in, it has to load the whole thing, but the next instances are only a lesser “cost” especially to RAM (aka memory).
96k definitely feels more "accurate". When I work at 96 it feels like I make mix decisions quicker, everything seems a bit easier. That being said, I almost never opt to start a new project at 96 unless it's a delivery requirement or somebody on the team wants it.
Great video ! As a live-sound engineer for over 25 yrs, I mixed on digital consoles at 48K for years and found it more then good enough for the purpose. Somewhat less than 2 years ago the company I mostly work for switched to a famous British brand 96K consoles. The sound of these are more “snappy” less “blurry” then the 48K consoles, and yess it’s a feeling and not proven science. Anyway, the 96K consoles helps me to “open up” a big channel count live project and I’m under the impression that there’s less latency issues. The rehearsals and shows are being recorded and every time I open a live session in my home studio they sound “clinical” and cold. Taken into account that live shows are performed in sometimes not so nice acoustical venue I definitely prefer 96K for live sound because of the clinical sound, in the studio ti’s just the opposite . Keep the good video’s coming!
I suspect it would have more to do with track mixing at 96khz. The reason I think that is because most of our input and output stages tend to roll off approaching 20khz.. so going inbound, most of the frequency res you gain from 96khz - which will handle up to the 48khz freq band @96khz clock - is probably rolled off by actual analog coupling. BUT! When you start mixing things together, I would definitely expect a case where, especially in high frequency content noise (eg the transients), you will really start to use that extra frequency resolution. On the way out, at least back to the speakers again, you get stuck with the roll offs once more. I would not be surprised that even with the squash back to the lower bandwidth, you end up with a different "feel" on the final result. Different stuff to work with.
One thing that could cause more of a slowdown than expected with 96k over 48k would be CPU "cache misses"; CPUs have their own cache of memory to store data for processing. Access to data outside of that cache is much, much slower (to the point where a computer program with really bad data management can see a big performance improvement by simply organizing the data better). Although 96k uses just twice as much data as 48k, that could potentially cause more than twice as many cache hits, leading to a significant performance difference.
Colt, I need to confess something... Back in 2009 or 2010, I´m working on a great studio at Brazil. And I it was my 1st time recording with ProTools HD and Digidesign 192´s... lots of analog gear...and someone told me Madonna´s album back in the day was recorded with "the same" HD rig running 24bit 192K...Man, I had the crazyest idea and recorded a full session running at 192k... I simple couldn´t mix the project until downsize to 48k...
A quick primer on sample rate. Your snapshot analogy is right but maybe a little more detail might be helpful. Sound is a variable pressure wave through air. When it strikes the diaphragm of a microphone it regulates a voltage exactly matching the pattern of the sound pressure wave (well, nearly exact, but that's another subject). This is a conversion of sound waves to voltage signals. The Analog to digital converter then takes measurements of the voltage signal level (your snapshots) usually expressed as some fraction of one volt positive and negative (IIRC). So, one can think of it as an X, Y grid. Measurements of time are X (snapshots at 48k or 96k Hz) and the numbers of grid lines of the Y axis is the bit depth, 16 bit, 24 bit, 32 bit etc.) So, I'd say the "Snapshot" is really of both the voltage measurement on the bit depth scale. Again, not saying you are wrong, just going into a bit (haha) more detail.
Haha. So no matter how rich your girl is you can always give her diamonds. No matter how much gear your producer has you can always giver her/him more hard drive storage.
The other thing to think about, not that I’m a proponent of either, is that neither really matters as long as the record and performance is excellent. A lot of it gets conflated in our world. I’ve made records at all sample rates from 44.1-96. And it’s always about the music, not the sample rate.
funny because i saw a video of JJP and he said he hated 48k and preferred 96k. clearly a taste thing. i personally do my own tracking at 96k but im an artist/producer and its specifically for vocals really(obviously the entire session is at 96k to support it). if the engineer wants to mix at 48 or 41, im fine with it.
IMO, 96k is the balance point of oversampling and ant-aliasing. Enable the plugin's oversampling cost smearing-transient. In 96k i can use less oversampling to prevent the transient and less aliasing distortion as well.
I do everything in analog. No tracking, straight to mastering chain, so nothing is lost. I prefer 96k. I will say that before I got mastering grade Lavry Gold converters, I might have agreed with this take but Lavry Gold converters make everything sound beautiful, 44 through 192.
Hi Colt, could you please make a video or a short on how you have been archiving your sessions and projects. Thank you kindly. Wishing you a great day!
I hopped on the 48 team maybe 6 years ago. I had always worked in 44. So I exported a beat at 44, then 48, 88, 96. I may have stopped there. I noticed that as soon as you get to 88, the sound of your digital instrument vsts changes very drastically. Not so much for pianos or other live sampled instruments, but especially with synth instruments like a warble bass, a saw, stuff like that. So I stuck with 48. But like you said, if you’re tuning a vocal or time stretching, 96 is easily the way to go. Just gotta think ahead before you start tracking vocals. Like is this guys voice gonna need a lot of work? If so, 96. If not, 48.
I equate it to how the 4k TVs look (on the stock settings)... more accurate for sure, but it also ruins the lighting and makes everything look like you are standing there in the room. The reality of it ruins your ability to immerse yourself in the show.
I wouldn't say its entirely an accurate description. consider how 4k resolution compared to 1440p or 1080p is a dramatic and clearly perceivable difference. whereas however when it comes to audio, 44 kilohertz and 48 kilohertz already covers all of our perceptible human hearing range. 96k contains more analog information than we can even experience as humans, and the auditory playback hardware you have might not even support frequencies that high anyways therefore a more accurate description would be, comparing something like 16k video, to 32k video. both are quite overkill and over what we're capable of seeing anyways. however the higher resolutions always provides more flexible options when it comes to zooming in and such during post production your analogy has a good idea that I recognize though. having the most "realistic" version of something isn't always what we want
Don't think it works to use a cross-sense example: the art experience of eyes and ears aren't the same. Like, does it make sense that to say lI like the smell of avocado, so therefore green must be my favorite color!". No! :) Also... many say that true analog is the best (like no digital in the process: tape, to say, vinyl)... that path is which is infinite samples per second (or what? The space between electrons?:)... which is closer to 96k than 48k. It's all taste at the end of the day. Do what you like.
This is kind of the same reason why I like Apollos. The vibe feels better than some converters that are more clinical but technically more accurate or have more clarity. Apollos sound more analog.
recently been experimenting with both, and i appreciate you making this video.. helped me make my decision without spending any more of my time, and storage, trying to figure it out.
I am personally suspicious of this "clinical" feeling of 96khz. 48khz audio perfectly "captures" the audio, as far as how we perceive it. Humans can't hear that extra octave of bandwidth that 96k provides. Even in terms of envelope of the audio (transient response), I don't think that the differences could be loud enough to be heard. Plus, humans are very bad at hearing the true envelope of sound, which is why static phase shifts are usually inaudible. I think the biggest difference would be in the lowpass filtering on input and output stages. 96khz audio can afford to use more gradual filters with higher cutoffs, meaning the top end of the audio is better preserved. There's less filter ringing, less phase shift, and less treble rolloff. Still, these differences are mostly above 10khz. So at best, this is a very subtle difference, comparable to two different transformers. Or an extra couple dB on a high shelf. And the filter ringing is literally at 20khz. I would wager that 20khz ringing is inaudible to most people. In the box, more differences can arise. EQs cramp at lower sample rates. Most people will just adjust the bandwidth and gain to compensate. So if anything, this is a workflow difference. Distortion, obviously, acts differently at lower sample rates. If anything, the aliasing of 48khz is the source of the grungy character. Still, oversampling is an easy fix for aliasing, and it induces minimal side effects. The main differences between sample rates occur when pitch shifting or time shifting signals. Also, the CPU and storage are notable variables. The sound is the least important factor when choosing a sample rate, so this should be a non-issue IMO. Not a single listener will be able to tell the difference. Record in whichever makes sense. For example, kick drums can be stretched minimal side effects, since they have a small bandwidth and mostly low frequencies. So 48khz is okay. But drum overheads might benefit from 96khz if you're planning on mercilessly stretching them to fit the grid. Anyways, that's my two bits. I should do tests to back up my claims though. Maybe the input/output filtering is more different than I think.
Given human hearing, the only audible differences for tracking (not time-based editing) at 96kz for mic’ed performances are likely due to antialiasing filters. The audio quality for tracking shouldn’t improve from 44.1 to 192Kz (at least for humans). People are entitled to do what they want and have their preferences and rationales. Yet, most mics, preamps, etc. aren’t capturing or producing frequencies at or above 20Kz for a converter to convert. If a person likes the filtering differences of a higher sampling rate, go for it. But 44.1 and 48 are used to make masterful recordings on a daily basis.
I typically use 44.1 or 48k, but have used 96 or even 192 on occasion when I waned to slow a track way down for certain types of effects. You get more fidelity (like you mentioned wrt elastic audio).
I still record at 48K for compatibility, since I move from place to place and some places are not ready to work at 96, and agree that for some reason I prefer recording on 48. However I’ve been trying up sampling for mix or mastering and it is an impressive thing cause when you’re dealing with more complex waveforms and more harmonics that’s when 96k shines also if you have any analog processing like a bus compressor you can clearly hear the difference specially in how the harmonics are better preserved and you don’t have to deal with aliasing.
There must be a way to run the same performance into two separate systems, one set at 48K and one at 96K. That's usually how A/B tests are done. You would need two computers and two interfaces and a whole lot of signal splitters, I suppose.
That’s the only way you could really do it, and then how do you record the same performances to multiple systems at the same time and have it line up with everything else that was previously recorded on each system? You would have to slave one pro to rake to another so they functioned as one even though they were two completely separate rigs. And I’m Too dumb to even figure out if it’s possible, or how to do it lol
I'm not a PT user, but I can't see why you couldn't lock the two systems using SMPTE or MIDI time code. There is a lot of extra "button"- pushing for record enabling and such but hey, this is science. It isn't supposed to be easy. Engineers have synced multiple analog tape recorders for 50 years, and digital is suposed to be better ;-) Anyway, interesting video.
I hear you for ITB mixes but I really have to disagree for OTB mixes. High voltage summing and slamming mix bus transformers makes it sound like a record and sounds better at 96 ime.
This pretty well covers why I'm comfortable with 48, and to be honest, why I was comfortable with 44.1 before that. I don't do any fakery in my music; if something I record isn't right, I re-record it. Autotune and time stretching aren't part of my process. And even if they were, I suspect many people focus on the numbers because it's something you can solve just by spending money, while actually making art of substance is harder and where the more important decisions are. If someone is just making soundalike filler, the fine points of bit rate are academic because you're just dumping noise into the void. It's pretty common to always want to be using the most of anything. But the idea of future-proofing seems futile. 5 years from now the standard will probably have doubled, yet all those pro albums recorded in 44.1 won't sound any more out of date than they do now. When I first upgraded to a system that could handle 96, I tried it and was underwhelmed; it seemed not really worth it.
Digital Reverbs sound really lush in 96K. The record head is slightly more accurate meaning there is less latency when recording. In order for y'all to hear how good 96K you need a high end DAC. Being that I monitor with Grace M905, I can attest that 96k sounds amazing! Now that music can be released in higher quality formats other than the classic 16 bit 44.1; It's worth trying for higher. I am inclined to mix at 96k, but the PROCESSING involved is no joke! Everyone will opt out of 96K really easy when they don't have the CPU power. If a session is 44.1K I'm fine with that too, because it is super nuance to the average ear. The part that no one cares is actually very true. It is more so an audiophile preference to dabble in 96k.
In the late 90's I started out in Digital Performer, recorded at 96K and never looked back. In Western PA we don't have a lot of Nashville musicians and much manipulation is needed, maybe that's why I chose 96k.
I think recording at 48k instead of 44.1k defintely is the biggest audible difference. Yes, 96k/88.1k can sound a little clinical. But as a guy not using an external clock, doing final SRC to 44.1K and 320mp3 or AAC's, 48K works fine to record in. And I use a lot of virtual instruments, so recording with higher than 48K isn't usually going to make a big difference. I have read a number of accounts from people with larger setups using external clocking that going to 96k can produce more "depth" and "width" in the sound field.
on my own method, I just record at 96khz then go back to 48khz when composing. my main genre is EDM but I also record vocals and instruments beyond synths because I include them in my tracks. my reasoning to mainly record stuff at 96khz as you already said when doing pitch correction or stretch it does helps. so having it at higher sampling rate as the original recorded stuff pretty much solves a lot of problem, but I keep the main editing and processing at 48khz. it's also double edge anyway since if the client wants at 96khz I already have all of the recordings at highest rate. To me higher sampling rate is basically like high speed framerate in video. it's like 48khz = 60fps and 96khz = 120fps, so it does sound more natural at higherframe rate since you can see more frames, smooth motion!
I'll have to test with guitars but recording bass, I notice that there was so much of a difference between how the bass sounds analog coming through my monitor and how it sounds recorded that I do max out the the sample rate and downsample it AFTER
As an ‘audiophile’ I store most of my music as 96K Flac and upsample to 768k in Roon before streaming to my DAC for playback, and yes it is super clean and super ‘clinical’ and every bad mix and bad master out there stands out like a sore thumb. But I am the minority.
I've switched from 88.1k to 48k, since im using more digital hardware synths and sample based drum plugins that are pointless to go higher. Just use reaper oversampling where it might sound better (some saturation)
This concept reminds me of the differences between filming 24 frames a second and 60 frames a second. 60 frames per second looks real but 24 frames per second looks cooler
Personally I do prefer 96k, but the difference is SOOOO subtle. For now I'll take the extra tracks/voices/plugins etc of 44.1/48 while computers catch up. The file size thing is real though... I worked on a Three Days Grace record that we tracked at 96k. The entire master backup at the end had to be spread across 3 2tb drives. Drives are cheap now, but the logistics of making sure those backups were correct and the end of every day was a bit challenging.
i did a record at 88,2 in the years of TDM and PTHD10, never did one again, the amount of disk space, the halved plugins on TDM i could use, with no sonic benefits to justify all that "cost". my luck was that being in a well furnished studio i could use almost only hardware compressors and 50% of the eqs where analog too. otherwise i'd have to had converted all the tracks to 48, and with those computers that woul'd have took ages
Do you think that maybe the reason the 96k transferred down to 48k sounded like it was almost there might be that the plugins handled the 96k different than the 48k. When I get back to the US I will try a few experiments with this idea. Kind of like I did when we went from analog to digital. I would record all the drums and bass on analog tape and transfer it to digital. That was when I was on Sonic Solutions as my DAW. At that time Sonic led the pack with the best sounding DAW, because it was 24bit, and the others were not. It worked well transferring the Drums and Bass to the DAW. It became more of a question of time then, so I axed that idea pretty fast.
I have a feeling the reason why converting the session to 48K was only halfway there is because it’s just cutting out half of the information. Rather than relying on the converters to choose that information if that makes sense.
This must be like the difference between 60fps and 24fps. 60fps looks like reality tv. 24fps looks like cinema and generally is nicer for our brain to process during action sequences especially. If you look at Gemini man in 60fps 4K, it actually looks like actors acting rather than witnessing something special.
Unfortunately, this perpetuates some myths (as does the 44.1 vs 48kHz video you did a while back). That's not how sample rate works. If that was how it worked, 20Hz content would be many times more accurately represented than 1kHz. That's just not how it works once you exceed two samples per cycle. Six points on the same circle don't make the circle any more round than three. Sample rate only dictates the highest recordable frequency. Having done tests myself with recording to two identical systems simultaneously using custom splitters, and monitoring through a sample rate agnostic converter (Benchmark), as well as exporting each to its opposite sample rate and doing double blind tests, I can tell you that my experience and measurements in the controlled testing do not match your subjective results in an uncontrolled testing situation with multiple variables.
Colt I like what you have on 96k verse 48k, I tried both but because in the beginning 48k was widely used in video forever. I found from 90s on that working 48k works for audio and video clients so staying that for economics really. Backing up 96k requires double the storage space. Keep moving forward Colt the list will get smaller someday!
It's all about aliasing in plugins and not forcing realtime oversampling. With 4 plugins in a chain all upsampling downsampling etc.... it can get a little mushy. I work with Analog modulars a lot and 96k (well I use 88k if I'm going higher sample rates) and they can make plugins alias a lot at 44k. Can I hear a 96k or 88k recording wav vs 44...nope. It's all about plugins :)
Every engineer I respect records at 48k 24bit. That is what I do as well. I don't tend to do lots of time stretching or tuning as I am of the opinion that getting it right at the source works far better than fixing it in post.
![If you can't hear this then you're not an audiophile [See description for link to followup video]](/img/n.gif)
That honesty at the beginning goes a long way
It truly does annoy me… Thank you for watching
Yes - I really appreciated that you were honest about the title of this video right away! Great video.
It would've been cool to get comparisons of time-stretched audio between 48 & 96k, and how it would translate on TH-cam 😄
Thank you for your awesome work!
Second this.
@@Liio.ChantelI could be wrong but as far as I can recall, certain DAWs (if not all) typically resamples audio to a higher sample rate when editing (ie. time stretching) to counteract artifacts.
This is why I love this dude, upfront
I use 44100 for 25 years in electronic music production and recording. Clients never complained, Awards are here, support from Tiesto and Armin was in 201X's.
Also mixed some Big Stars stuff - very often I have got 44100/48000 16/24 bit sessions. It doesn't influence a hit record.
When I am listening to 60-70s records it is very often sounds like shit in terms of quality. But people don't care.
From technical point of view, higher sample rate means more information for processing (pitching and stretching) but I never had a problem with 44.1 material. It is just sounding different when processing 44.1/96 material.
Also I have done a blind test of 44.1 VS 96 of the electronic mix done entirely in the box. More people voted for 44.1 which sounded a little bit grittier.
Now I track my hadware synths into 96K FLAC in case I will do hi res mixes for my fans.
Many years ago I have mixed a song with 2 vocalists. One is Big World star recorded on SM58 and other is young vocalist recorded in top notch chain with M149. The sm58 recording was phenomenal and m149 was awful.
I think big names just CAN afford analog or 192 or 96. But if CLA will mix in the box at 44.1 it will be CLA mix and no one will care about how he mix it.
tangentially: I prefer 44.1/24-bit audio over 16-bit audio at any sample rate.
Finally someone with a brain.
44.1Khz is sufficient to reproduce the original audio without any loss.
96KHz will allow you to sample up to 44.8KHz frequencies without any loss, as per Nyquist Theorem. The thing is that the human hearing frequency range is 20Hz to 20KHz.
Unless you're making music for dogs it doesn't make any sense to use 96KHz.
Blind-test?
There are HiFi forums that will ban you for just thinkig it..
Ultimately it comes down to performance. In descending order
Performance
Instrument
Mic placement
Room
Mic pre / analog chain
Bit depth
Converters
Bit rate
Some of my favorite songs have "bad mixes" and that's what makes them special.
They're bad in just the right way.
One of the easiest ways to demonstrate the difference in is video frame rate. The afternoon soaps were recorded and aired at 30 frames per second, while the film-originated prime time shows were recorded at 24 frames per second. Intuitively, one would think that the higher frame rate (sample frequency) would yield more detail and hence a better rendition. In practice, it turned out that it was far less pleasing. It felt cheaper because it removed part of the 'unrealistic coloring' that cinematic productions have by nature. I like your tape analogy as well. Odd as it may seem, the audience apparently doesn't want productions that are too accurate.
I like 96k because you can play with slowing things down a lot. Outside editing purposes I don’t hear the difference other than some aliasing artifacts sometimes.
I've always recorded at 96k/24bit with the idea of "future proofing" my stuff. I've also yet to come to a point where I've needed that "future proofing". I find 96k noticeably cleaner for time stretching events and sample work. Having "more" data to work with while manipulating. It's not the exact same, but I've always looked at it as 1080p vs 4k video (8k being 192k). With 96k and 48k there are advantages and disadvantages to both, I've always built my recording rigs with the plan to record and mix with 96k from front to back. But in the end as long as someone is happy with their mix/product, no matter how they got there, who cares! Cheers! Great video as usual Colt!
There are already streaming services that offer a premium 'hi-res' service. So 96khz IS future proofing you.
96k is 4k real deal
I’m realizing that since we don’t use compact disk anymore my method of recording at 88.2 so I can mix down to 44.1 is obsolete now. I believe everything has moved to 48 24 bit over 44.1 16 bit.
@@stevewoodyt 24 bit for sure. CD is still very popular globally actually. Whether it's audible or not, people do want hi-res audio from Apple, QoBuz, Spotify etc, so it's good to be able to deliver 96/24 and take advantage of that.
Actually, 96kh doesn't increase accuracy at all for each sample. It only doubles the amount of frequencies from about 24kh to 48kh. But, at 96kh, when processing you get increased quality when pitching and reduced aliasing as well as reduced need of added filters implemented in plugins that automatically oversample for certain stages.
Thank You! It's what I learned in audio school too. If you only record and don't do any digital processing, our ears shouldn't be able to hear a difference. So I feel a bit confused when I see this and read the comments. As long as aliasing and anti aliasing filters are not involved, why would you hear a more accurate recording?
@@magnusboder6680 I believe that it comes from the misconception that higher sample rate adds samples between 20 and 20Kh. It doesn't because two samples are enough to represent a perfect waveform when going back to analogue. Sorry for not being entire clear, I'm not an expert in digital theoremes but you could review Nyquist-Shannon's work to understand that better.
@@raphaelherzig3316 You are absolutely right. And I learned about this with Nyquist crossing and aliasing. But again, it only applies to digitally processed audio. And in the video he talked about a live drum recording and the accurateness of it's recorded sound. And those words made some people thinking about recording in a higher sample rate and than mix in a lower to save cpu load. And this is just... Craziness :) So that's why I decided to write, even though I'm not native in English :)
Correct. Finally someone understands!
So this means you record at 48,thank you
It sounds like when filmmakers such as Peter Jackson decided to double the frames per second on movies, and all of a sudden it looked “too real,“ and also “not like a movie.”
The enalogy is completely wrong. Sound is not frames, sound is a continious signal. DAC makes it contunious from digital data.
Sound is continuous as well as visual is continuous in the analog world. Sound is sampled at, for example, 48kHz while video is sampled at 30 or so fps. As I've said for years the illusion of sound is that it can present the illusion of stationary solidarity while the illusion of motion pictures do the exact opposite - present the illusion of continuous movement with a rapid succession of "frozen" images. How the brain interprets the signal-images from the human transducers ears and eyes is another topic. Compare the flicker rate in visual where the image is not seen as continuous to the much greater sampling rate in audio 48 K for example.This discussion can carry on past a bottle of Jameson easily.
@@CuriousPassenger he just said saound is frames...
Ideal test (in my mind, at least) - Track one song with mult splits of each channel, one going into a converter at 48khz, one at 96khz. Mix the 48khz fully in one DAW session, print mixdown, duplicate session, change project sample rate, swap audio files to the 96khz samples, print mixdown - compare.
Obviously, if you can’t perform a true A/B test, then what’s the point?
Everyone talks about the end format when comparing. No one talks about how effects like reverbs especially, sound better at 96k where it can time slice the decay to a more natural sound. I started using 96k since PT latency is reduced per Avid's specs. HD disks are dirt cheap today so IMO 96k is the sweet spot.
Pfff… 768k is where it’s at. Harmonics scream and the lows are like butta.
For time based effects there is exactly zero difference mate.
@@CuriousPassengerunless... it is badly coded or adds saturation. Or... both. Lol.
Good vid, but.. I don't think it works to use a cross-sense example: the experience of eyes and ears aren't the same. Like, does it make sense to say "I like the smell of avocado, so therefore green must be my favorite color!". No! :)
Also... many say that true analog is the best (like no digital in the process: tape, to say, vinyl)... that path is effectively infinite samples per second... (or what? The space between electrons?:)... a non-approximated capture of sound without "snapshots".... which is closer to 96k than 48k. It's mind if all taste at the end of the day. Do what you like!;) it's all fun to debate though... I switched to 96k a few years ago and noticed it feels more detailed... but it's really hard to know for sure what's "better"... I would rather add my own color shaping with saturation and other processing to achieve "vibe" than have it come baked into an underlying element like sample rate. But still; not sure. I don't mind a head start... if I'm oil painting a night scene, would I rather start with a black canvas than a white one? Maybe. Maybe not.
Very interesting take, Adam… thank you sir for sharing.
Laz, I could play you anything ran through standard DAW reverb plugins at 44.1/16 and you’d never know the difference.
Your comment about hard drive space not being an issue, when touting such a high and silly sample rate makes your overall statement difficult to take seriously.
If you had a 1/4 of the sample library I had… and it was at 96, you would change your tune in a hurry.
I used to do a lot of rap production and they would have me do these chopped and screwed pitch shifts down super low and I found that 96khz was really good for that because you kept the high frequency information intact
But.... microphones don't actually pick up that information... a microphone capturing anything audible above like 16 or 18khz is honestly pretty rare.... you need like microphones made for scientific purposes for that usually.
Do you mean like the synthesized instruments, because that might have some information that extends beyond nyquist.
"you kept the high frequency information intact " Hence, they all strive for that "lofi" vibe. Right ?
@@Linguae_Music I mean vocals, on a microphone. I dont know the exact science behind it but there was a very noticeable difference so if you're skeptical try it yourself! It might have more to do with stretching the digital audio than the analog characteristics of the mic and the preamp
@@raymota4515 right I think I was the only one trying to keep the fidelity high hahaha
Sampling at 96KHz might provide more samples/second but, ultimately, it determines the highest frequency you can record - half the sampling rate (Nyquist). The human ear limit is ~20KHz. 44.1KHz/48KHz captures everything we can hear. My 2c: Use the frequency that's appropriate for the target medium: 48KHz if the target is video, 44.1KHz if the target is audio (CD, streaming audio).
I try to work in 96k as often as possible. When I was going back and forth with the pros and cons (some of which were discussed in this video) my ending idea was thinking that a chef dosen't prepare a Waygu steak with a steak knife because that is what the end user will experience. He has to work at a higher fidelity. So I do as well.
I can always smear transients later if I want, but you can't un-toast bread.
Exactly right. Don’t burn the bread or the nuts.
So, you don't have any technical arguments for your choice?
A steak doesn't get any better if cut with a knife that took Japanese mermaids 8 years to make..
@@hansemannluchter643 correct. Because that's in your ability as a professional to make whatever you get the best it can before it leaves your desk (or kitchen as per the analogy).
My point was I want to work at the highest fidelity possible. I've found that any sonic benefits to a lower sample rate can be replicated with either hardware or plugins.
Confirmation bias is the tendency to search for, interpret, favor, and recall information in a way that confirms or supports one's prior beliefs or values.
Ain’t it interesting…
I tend to agree with Dan Lavry's thoughts on conversion. It seems quite well thought out that the ideal sample rate would be 60khz or so. From his "white papers":
"Good conversion requires attention to capturing and reproducing the range we hear while filtering and keeping out energy in the frequency range outside of our hearing. At 44.1 KHz sampling the flatness response may be an issue. If each of the elements (microphone, AD, DA and speaker) limit the audio bandwidth to 20 KHz (each causing a 3dB loss at 20 KHz), the combined impact is -12dB at 20 KHz. At 60 KHz sampling rate, the contribution of AD and DA to any attenuation in the audible range is negligible. Although 60 KHz would be closer to the ideal; given the existing standards, 88.2 KHz and 96 KHz are closest to the optimal sample rate."
I tend to work at 88.2 and 96. I work on classical, jazz and acoustic-heavy music for the most part. It can bog down the system for sure. Though it does help when mixing to avoid aliasing buildup with certain plugins that I like that don't oversample. All of this is relatively small fries. Some of my favorite records from the analog days are noisy and messed up sounding anyway. Then the earbuds argument. I don't really do this for the earbud crowd, though. I do it because I love great audio and art.
"smearing" ain't always a bad thing. It's been part of the allure to analog since, forever. I've been really happy with the small move from 44.1 to 48 myself.
You might wanna try doing a blind test because that’s not how digital recording works. Nyquist states you only need two sample points to accurately reproduce a frequency. A larger sample rate just means you can record higher frequencies. It’s possible you had some audible aliasing but I highly doubt that. Aliasing usually sits way below the audible range unless you’re trying to create it. Most likely you were being tricked by either volume or your eyes or both. Set up and ABX test and correctly pick out the 96k recording from the 48k recording 10 out 10 times and then you can say you can hear the difference between 96k 48k. Haven’t seen anyone do that ever in my life though. Most people can’t even distinguish a 320 Kbps MP3 from a lossless file.
Heresy like that gets you banned on many HiFi forums!
This reminds me of how watching high framerate videos can be weirdly off-putting. Like there's a certain level of fidelity that's too uncanny
88.1 or 96k means you may not need oversampling as much IMO and latency is cut in half generally. For me the thing that makes me want to record at 96k is the more natural sound especially for acoustic music. UAD and others admit clocking is different between sample rates on some devices. My current interface (mt 48) and prior one (Antelope) sound quite different in the top end especially when you go between 48 and 96. Unfortunately it's not as clear cut as 1 is always right.
You hit the nail on the head for accuracy. The music I do requires it and at a full final mix, my CPU is always barely 50%. Because I don’t use many plugins anymore since I’ve been able to collect the 20+ year analog setup I now have. Even with pop country or doing rock, the converters I have and my mixing approach sounds better and more defined from the get go. So, I’ve lived there and haven’t changed for many years now. When I eventually change converters, as nothing lasts forever, I’ll listen again and decide then what I’ll do. Quality hard drives are crazy cheap now, so, storage isnt really any deciding factor for me. Great video and I appreciate the time it takes you to make them. I wish I had the time to do the same.
I literally clicked on this knowing it was click bait. I use 96k because latency is halved and reverbs seem to blend better with the source material. As someone who shoots music videos , book trailers and TH-cam stuff as well as runs a recording studio I can assure you that I run out of SSD space in a days of video work. On my Audio computer and external HARD DRIVE I haven’t come close to filling it yet.
Great thing about 96k is that you get half the latency while maintaining the same buffer size. This is great for real-time monitoring of kick triggers live, neural DSP or if a vocal is going through a daw and echoed/input monitored. :)
for a live show 16bit at 96k is a great way to go if low latency and super reliable performance from the computer and interface is desired. It gives the computer much less to process since 24bit has a lot more data to process/write than 16 bit.
You seem to be talking about budget conscious sessions. Storage is so cheap now. I record everything at 96/24 on a quite affordable M1 MacBook Air, with multiple plug-ins and never have a problem with the processing, or slowing down.
Accuracy? Yeah, I want my (rock) drums to be as accurate and high quality as possible. It's not just for classical.
A lot of clients I work with demand 96/24 - which is the main reason I switched from 44.1/16.
In the end, working at 96/24 does not impact my studio work one bit - there are no cons. So I do it because that's what I'm asked to deliver and also because I want to future proof my work.
Same as other people said, the intro was Super appreciated, No one actually likes clickbait it just gets more clicks due to the algorithm, but this vid was Great
Your sincerity and how you're expressing your prevalence towards 48k is really helpful. Because I've listened to how you mix in your videos, I trust that you're coming from a place of time-tested wisdom that works for your ears. And of course, you mentioned 96k for more dynamically sensitive music like bluegrass and classical, so thank you for pointing out that difference.
I would like to see more science on what you're hearing, whether conversion quality during downsampling, or ears being tuned/used to enjoying the buildup of more foldback aliasing distortion at 48khz (very little, but some and more than 96khz). I also prefer 48khz, and with the advancements happening with antialiasing algorithms, It's likely never going to be a problem. Another thing to consider is, it may be the same reason people love Soundtoys decapitator which has no oversampling and a TON of foldback aliasing distortion: Ears like what they like. If it sounds good, it is good.
I like to use 96k, but I agree with you, Ears like that they like and if it sounds good in the end that's all that matters!
It can be that because you are now used to listening to 48k gtr and drums which has become a refrence point so the 96k sounds more clinical
Ngl, I almost didn't watch this because of the click baity title but I'm glad I did. I do a lot of sound design work and 96khz and 192khz are regular daily use sample rates for me, specifically because I do some much multi-octave pitch shifting and time stretching. Working with higher sample rate files really goes a long way in preventing artifacts and improving intelligibility, like you mentioned with Elastic Audio.
Whenever I do music I pretty much exclusively work in 48khz. I think high sample rates are a specific use case and there's not enough education out there on why and when to use them. Same kind of argument with 32-bit depth. 90% of people do not need to record at 32-bit but if you have a job like mine and have to record things like fireworks or guns, having a 32-bit capable recorded can be a godsend.
+1 for stretching. I recorded myself blowing an Aztec Death Whistle on a Sanken ultrasonic mic at 192k and stretched it. Sounded like satan screaming and the quality was amazing!
I guess you can always record stuff that you’re going to time stretch in 96K but keep your regular tracks at 48K
I’ve been recording at 96k, only at the advice of my last producer, for about 10 years now, and there are some valid arguments around CPU usage, less so now though. I can only imagine how much faster computers will be in the next 10 years and beyond, as computers get faster, lossless 96k and beyond becomes easier and easier…
As much as the “vibe” is concerned… that’s possible. It’s definitely not for “no reason” that I use tape emulation when mixing, I like the “vibe”, or “tapey” quality of how it sounds. I definitely might run my own experiments. It would be cool to choose between a 48k vs 96k song, but it would have the be the same audio in both examples, recorded on two systems simultaneously, or there’s too many variables when deciding what’s better. I won’t deny it’s possible though, there could be a “vibe” to 48k that I’m missing…
There’s also a lot of research done around aliasing, which is caused often times by saturation plugins that don’t have oversampling features. I haven’t tested this extensively, but it is another reason why I tend to lean 96k….
It is interesting…
Really interesting - Thanks for making this, I was working in 96 and like you said 'the realism' is there. I'll experiment with 48.
And the whole thing about 'no ones gonna hear it' argument - what i've always said is the consumer wont hear it but they will most likely FEEL it.
I used to capture my sessions which are typically comprised of live-tracked drums, guitar and bass through various interfaces, and vocals through preamps loaded into usb 500 series racks with hardware inserts on effects solely in 96k and I was very pleased with the clarity and accuracy; HOWEVER, the need for more and more storage and constantly running out out of drive-space became more than I could bear; but I kept trying to saunter on. It wasn’t until I acquired some vintage gear that I added to my hybrid system that topped out at 48K that I said what the heck, let’s see how working in 48K is and I have to say the negligible sonic differences and the ability to store a greater number of projects to one drive has me working totally in 48k now without feeling like I have compromised anything. Totally worth it.
Hello! I don't work as a mix engineer and my english is not very good. I have a degree in audio and acoustics though. And now I am a little confused. Both from some information in the video and from reading the comments. As long as a sound is just recorded, and no digital processing is involved; aliasing and anti-aliasing filters. There is as far as I know, from a scientific perspective, no way to hear a difference with human ears. Now I see in the comments that people get the idea to record in a higher sample rate, and down sample to 48k for mixing... Only the other way around could be benefitting if you mix with plugins. I mean according to my knowledge. If you heard a difference, it may be because of your converters. But this is individual for every kind of converter and has nothing to do with sample rate. I really don't try to be a rude idiot now. And I have been wrong about things before. So feel free to correct me everyone if I am wrong! I really love to understand how things works and wish to have acurate knowledge :)
You are right. Your confusion is justified. There are apparently great numbers of people in the music and recording industry who still do not understand digital audio, particularly resolution and bit depth. What you have learned is correct.
There is a very good chance the preference he feels is how his AD and DA converters handle the different sample rates. If he had different converters, he might feel differently.
The higher resolution will help when working with multiple tracks. When you combine your 96K master tracks this will help when recording the lower resolution final two-channel product. I'd just say to each his own. After all there are artists and producers who seek out vintage analog equipment (old mics or old amps for example) in search of a specific vintage sound. Although 96K would aid the mastering, in the end you have to pick what suits your ears.
I don't think this generalization holds, and it may be specific to the converters in your interface. I'm in the middle of recording an album with a quartet of stringed instruments. We've mostly been working at a studio that's been recording at 24/48, with some really botique preamps, and high-quality mics. Some of us have also been bringing our own (AKG & Telefunken) mics to sessions. There were a number of complaints about the sound quality we're getting at the studio, so I decided to give it a try myself. So I recorded a session of the group, using a MOTU 828es with a DAV BG-8 preamp, into Logic at 32/96. We used two pairs of C414s, a pair of Telefunken M60 FETs, and a pair of Beesneez Lulu FETs. Four of the 8 mics were exactly the same mics we had been using at the studio. The recording I made at 96k just sounds so alive and 3-dimensional compared to the 48k recordings from the studio. If you just do a direct A/B comparison, the 48k recording just sounds flat and lifeless in comparison to the 96k recording. I really think that while some converters in some interfaces say that they can do 96k or 192k just fine, that they don't actually do as good a job as others at those sample rates. For what it's worth, the ESS 32-bit converters in the MOTU seem to have no problem at all delivering consistently preferable results at 96k.
IF 96K is too much etc, HOW would you handle a 192K recording?
I work in 88.2 because it's double of 44.1 which is where a lot of masters end up at (CD Baby still only accepts 16 bit 44.1 kHz). And it's still high enough to do some elastic audio stuff.
Smart move.
Same
Honestly I don't feel there will be a huge difference between 88.2k vs 96k because it's the difference between a hand grenade and a rocket. There might probably be a case out there because there is ALWAYS a damned edge case, but it'll be rare.
Colt, might I suggest a follow-up video about the Nyquist theorem and how resolution only determines how high an audio frequency can be recorded? I am surprised at how many people in the recording industry still don't understand digital audio.
Higher resolutions are not more accurate, not cleaner or more detailed. Higher resolution merely makes possible the recording of higher frequencies. 40 kHz is capable of capturing 20 kHz sounds, the theoretical limit of human hearing. I'm simplifying, of course. There are technical reasons for using 48 kHz.
Bit depth is another issue. It determines the number of amplitude values that can be assigned to each sample. Accuracy and clarity are, in fact, affected by bit depth. But 24-bit recording makes for such a low noise floor, it has been adopted as an industry standard.
As an engineer, I definitely appreciate seeing this information in the comments.
FINALLY!!! someone talking about the math behind nyquist theorem in the comment section! I was staring to worry a lot , thank you so much for talking about that. I just can’t understand how everyone don’t understand a basic math theorem.
Yep. It's amazing to me how much this simple concept is overcomplicated. Sample rate controls max frequency you can store, and bit depth controls the quantization noise floor. That's it!
I've been using 48k for 10 years because Warren Huart uses 48k. I've learned to be completely content on 48k
Yup. Been on 48K for at least 10 years as well.
nothing musical about 48
I started using 32/96k a few years ago because of time stretching (obviously better) and I felt that distorted guitars sounded slightly smoother and less harsh. I record 6K RED footage, so storage space on audio is minimal compared to that, but you work on way more songs than I do. The biggest downside to me is processing, but I’ve heard it said some plugins work better at 96. On my computer it “feels” about 50% more intensive on the computer (may not be the case though). I think its the kind of thing where find what works for you and go with it, like DAW choice. I seem to remember JJP saying he prefers 96, I wonder if some people just have different tastes and naturally prefer one over the other. I also wonder after being mastered from 96k to 48 for streaming, etc if some of the clinicalness would go away. Good topic to ponder. Thanks Colt!
Idea for the A/B test: Put a transformer isolated Split after all your preamps and send each into different converters as you track a song, and then listen to the summed mixes differences with the same plug ins and session settings on each
That’s exactly what you would have to do. But even just having two completely separate rigs that were identical is more trouble than I can convince myself to go through.
@@ColtCapperrune Oh for sure! I aint gonna do it either. That then opens complexities of different plug-ins's sample rate handling, non-linearity, dithering, and then mixing down to lower sample rates in mastering. Too many factors to isolate. Just having the convo is enough for me-- earnest and insightful as always my man!
use dual active split. a transformer isolated split will show differences from the transformer. sample rate differences doesn't affect the waveform, they will null in the audible range. 96k recording brings in higher frequencies that will cause more aliasing through processing compression and saturation. use 48k and OVERSAMPLE as much as possible in your plugins.
@@joemaymusic and not to mention: what it does to you as a musician/producer while the stage of recording, listening to different sounding stuff will inspire you to do different things.. that psychological aspect of the creative stage affected by that technical difference.. you will never be able to isolate that aspect…
What about skipping all plugins? Just listening to the raw recordings
Higher sampling rates than 44.1/48KHz are mostly pointless and its a myth that rates that are higher are more accurate, pretty much all you do when using these are recording sounds that the human ear cannot hear, 44.1/48KHz can perfectly describe sound in the range of human hearing. For timestretch there can maybe be an advantage and you can potentially get lower latency with your audio interface.
it seems no one noted here, but every converter could sound different from another one when changing their own the sample rates. I remember many years ago when did a comparison test using 2 separate rigs: mics> preamps> splitter> into 2 separated Rosetta800 each one connected to its Macbook rig, one session set at 48 the other at 96. We repeated the same experiment with 2 Motus 896, and well, me and my fellow both preferred the 96k session from the rosetta800 BUT the 48k session from the Motu 896. If you compare same recording at let's say 48k and 96k with a certain converter, if you do the same with another one the difference between 48 and 96 could be clearly insignificant or more evident, also considering the nature of audio signal you're going to acquire. As already said by many of you, most of the tangible perception of the sample rate impact it's into acoustic instruments and/or very complex interactions with their ambient.
Then probably the music genre comes in play where also Colt says about the"vibe" he preferred. This when recording, then when mixing/editing etc and you modify the starting material with proessing plugins etc, it's another story. m2c.
Q: Do more realistically lifelike recordings sound “better” when doing music production?
Or are we all reaching for vintage, high-coloration, high saturation analog devices or digital emulations of them?
It’s the same as with filmmaking. Do filmmakers want their movies to look like high resolution “video”?
Or do they go to great lengths to achieve an altered and artistic look to them?
I think most of us already have the answers to these questions, eh?..
The only place where I've found high sample rates to have objective advantages is where you have to do extreme time stretching and pitch shifting (which, for me, was making creature sound effects during college exercises). 48k is more than enough for music.
Thanks again Colt! I have been playing with sample rates as well. My experience seems to match yours. I recorded an album at 192K, a classical, choral work. My mentors sort of trailblazed that for me because they track at DSD rates, so around 2.8MHz or 5.6, I believe, 1-bit. They do this because the idea is that this is "as close as one can get to analogue" since the sample rate is so high, the digitization of the soundwaves is arbitrarily close to analogue. It works well enough, but I also have taken to doing more and more similar genre recordings at 48Khz. This may be totally subjective, because like your own post, I have read other people who are top sound engineers say that 48Khz is all one really needs and that it is the best rate. For myself, I found that my masters of my 192Khz CD sounded warmer at 44.1 (CD rates) than at 192, and I could hear certain issues much more clearly at 44.1 than at 192. I don't know that the sample rate changed anything but I noticed the problem at 44.1 playback and it seemed a bit smoother at 192. Since our print for CD is at CD rates, of course, this is why I paid attention to it.
I have since gotten into the general pattern of recording everything at 48Khz / 24 bit. I appreciate your reference to 96 as "clinical" - and I agree. It feels more real at 48, warmer, actually, which is where I am trying to get with this choral work. 24 bits give me a lot of headroom and so it just works.
There is one other practical reason for 48Khz rates too. Video recorders, especially those on iPhone or Android devices default to 48Khz for their recorded sound. While phone mics are certainly far inferior to my "real" mics (ranging from Zoom H6 to DPA 4006As), the identical sample rate does two great things for me: It eliminates video-audio drift if I have to edit Video and Audio together, and in one situation, the Zoom refused to play with its SD card and I lost the material but the phones got it so I could reconstruct something (not great, but workable) from the phone mic arrays).
I think it comes down to "feel" over "math". I don't begrudge recording at DSD rates - but I cannot afford a HORUS / HAPI to do it with right now, and I CAN meet my field recording gigs really easily with my field kit - the aforementioned Zoom H6 with either its own mid-side mic and / or higher-grade outriggers, or if I feel miserly, some Karma Silver Bullets for outriggers. (btw, these Karma Silver Bullets - they are no longer in production but they appear on eBay from time to time. I have nine of them. They are tiny omnis - about $20.00 apiece when they were sold, but talk about bang for the buck... they are really, really good and they can give my higher-end mics some level of competition that would surprise a lot of folks.)
All the best to you!
When 96k came to pro tools it felt like 30ips 24 track: the biggest difference was …. The vocal track. So clear so pretty. I am obviously very old and remember recording reggae and rock at 15 ips! The latency at 96 is so much lower and it is a joy to work with a pretty sound. I do everything at 96 except Dolby Atmos mixing.
I appreciate the honesty with the click bait stuff. Love your videos.
If you're a musician who writes and records their own music without a producer, 96k opens lower latency settings on many interfaces like your Apollo. If you're a digital drummer or finger drummer who plays virtual drum instruments and is sensitive to latency, going from 64 samples to 32, opens up room for a few plug-ins while still playing in real-time. Go look at the latency figures for all of the UAD plug-ins for your x16, a large number of the plug-ins give you lower latency when in 96k vs 48khz and this is on top of the fact that 96k opens up 32 samples. The higher your sample rate, the lower your latency.
If you're working mostly in the analog domain via a console, you want 96k. In analog land, your console provides the color, grit and vibe, so you want things to be accurate and transparent when converting to digital. You have 2 newtons, why not use SILK on the master buss to dirty it up rather than using 48khz? I personally want my recordings to sound exactly like what I was hearing while still in analog land.
I honestly cant hear a difference when real blintesting, i recomend anyone to do a blindtest. And you must guess the right option 5 times in a row, thats how you know you can tell a difference. The only thing higher sample rates do is raise the extention of the high freqs, it is not at all more accurate on the frequencies both share. This can be scientically proven. More sample points does not mean a more accurate reconstruction. Also idk about accurate sounding more clinical, doesn't make much sense, more accurate should just be more equal to straight wire to the speaker. But doing Time stretching and infrasonic recordings would have an advantage though
"This can be scientifically proven. More sample points does not mean a more accurate reconstruction."
So as I noted above 4.8 samples per second of a 10 kHz wave will be as sonically accurate as 96 Khz sampling. Right?
4.8 samples per second? idk what you mean. But a samplerate of 16k per sec for example will perfectly reconstruct up to 8khz. just as good as 192k would for example. thats just how the physics work.
main difference might be in the antialiasing filters in converters, which all have a lowpassfilter. Which may reach down to a lower frequency on a smaller samplerate. But I find it strange when ppl say lower samplerate like 44.1 will sound "granier". If there is a difference, it would more be like a tiny bit more muted on the topend/air. But unless you can hear 18-20k and over i doubt youll hear any difference that matters. Atleast thats what Ive heard, and when I did blindtests.@@raymota4515
I was able to get ahold of some stems for a few songs by The 1975 from their 4th Album, "Notes on a Conditional Form." They use 96khz, and it sounds really crispy!! But I truly don't think THAT is the reason for it. I still use 48khz, and with Dolby Atmos in the picture, thankfully 48khz is an option for it. I can't imagine the file size's for a 96khz Atmos session...
I mix Atmos as well, can't imagine a 96k session 🤣
If you were to try another experiment in this vein, I would suggest getting a split snake so you can record the same mics and performance on two identical systems.
The other thing to point out is that most high profile live sound consoles, that are digital, are sampling at 96khz.
I have noticed a difference, in the fact that I’ve worked in one venue that went from an analog console to a Dlive. I did notice a difference there. But it would’ve been leagues ahead of the analog console compared to an m32.
At least you were honest in saying this won't give us anything. It didn't.
Very interesting, thank you. I recall one fella who always wanted to record his drums at 7.5IPS and then lay them onto another machine at 15 or 30 because he so liked the sound at 7.5. Same kinda effect I guess.
Justin at Sonic Scoop explains technically why 48k is the better choice.
Well, if Justin says so then the matter must be settled then. QED
Hey, we all have preferences, but it’s interesting that Avid, Allen & Heath, Midas, Yamaha, & all the other major live sound console manufacturers chose 96 kHz. Also, the plug-ins don’t all really scale like that. The first time you load a plug-in, it has to load the whole thing, but the next instances are only a lesser “cost” especially to RAM (aka memory).
96k definitely feels more "accurate". When I work at 96 it feels like I make mix decisions quicker, everything seems a bit easier. That being said, I almost never opt to start a new project at 96 unless it's a delivery requirement or somebody on the team wants it.
Great video ! As a live-sound engineer for over 25 yrs, I mixed on digital consoles at 48K for years and
found it more then good enough for the purpose.
Somewhat less than 2 years ago the company I mostly work for switched to a famous British brand 96K consoles. The sound of these are more “snappy” less “blurry” then the 48K consoles, and yess it’s a
feeling and not proven science. Anyway, the 96K consoles helps me to “open up” a big channel count live project and I’m under the impression that there’s less latency issues.
The rehearsals and shows are being recorded and every time I open a live session in my home studio they sound “clinical” and cold. Taken into account that live shows are performed in sometimes not so nice acoustical venue I definitely prefer 96K for live sound because of the clinical sound, in the studio ti’s just the opposite . Keep the good video’s coming!
I suspect it would have more to do with track mixing at 96khz. The reason I think that is because most of our input and output stages tend to roll off approaching 20khz.. so going inbound, most of the frequency res you gain from 96khz - which will handle up to the 48khz freq band @96khz clock - is probably rolled off by actual analog coupling. BUT! When you start mixing things together, I would definitely expect a case where, especially in high frequency content noise (eg the transients), you will really start to use that extra frequency resolution. On the way out, at least back to the speakers again, you get stuck with the roll offs once more. I would not be surprised that even with the squash back to the lower bandwidth, you end up with a different "feel" on the final result. Different stuff to work with.
What do you think of recording and editing at 96K, then convert to 48K when mixing?
I'd use a multi mic splitter and two different converters to be able to compare the same signals.
96k has quite less latency.
One thing that could cause more of a slowdown than expected with 96k over 48k would be CPU "cache misses"; CPUs have their own cache of memory to store data for processing. Access to data outside of that cache is much, much slower (to the point where a computer program with really bad data management can see a big performance improvement by simply organizing the data better). Although 96k uses just twice as much data as 48k, that could potentially cause more than twice as many cache hits, leading to a significant performance difference.
Colt, I need to confess something... Back in 2009 or 2010, I´m working on a great studio at Brazil. And I it was my 1st time recording with ProTools HD and Digidesign 192´s... lots of analog gear...and someone told me Madonna´s album back in the day was recorded with "the same" HD rig running 24bit 192K...Man, I had the crazyest idea and recorded a full session running at 192k... I simple couldn´t mix the project until downsize to 48k...
A quick primer on sample rate. Your snapshot analogy is right but maybe a little more detail might be helpful. Sound is a variable pressure wave through air. When it strikes the diaphragm of a microphone it regulates a voltage exactly matching the pattern of the sound pressure wave (well, nearly exact, but that's another subject). This is a conversion of sound waves to voltage signals. The Analog to digital converter then takes measurements of the voltage signal level (your snapshots) usually expressed as some fraction of one volt positive and negative (IIRC). So, one can think of it as an X, Y grid. Measurements of time are X (snapshots at 48k or 96k Hz) and the numbers of grid lines of the Y axis is the bit depth, 16 bit, 24 bit, 32 bit etc.) So, I'd say the "Snapshot" is really of both the voltage measurement on the bit depth scale.
Again, not saying you are wrong, just going into a bit (haha) more detail.
Haha. So no matter how rich your girl is you can always give her diamonds. No matter how much gear your producer has you can always giver her/him more hard drive storage.
The other thing to think about, not that I’m a proponent of either, is that neither really matters as long as the record and performance is excellent.
A lot of it gets conflated in our world.
I’ve made records at all sample rates from 44.1-96. And it’s always about the music, not the sample rate.
funny because i saw a video of JJP and he said he hated 48k and preferred 96k. clearly a taste thing. i personally do my own tracking at 96k but im an artist/producer and its specifically for vocals really(obviously the entire session is at 96k to support it). if the engineer wants to mix at 48 or 41, im fine with it.
IMO, 96k is the balance point of oversampling and ant-aliasing. Enable the plugin's oversampling cost smearing-transient. In 96k i can use less oversampling to prevent the transient and less aliasing distortion as well.
I do everything in analog. No tracking, straight to mastering chain, so nothing is lost. I prefer 96k. I will say that before I got mastering grade Lavry Gold converters, I might have agreed with this take but Lavry Gold converters make everything sound beautiful, 44 through 192.
Hi Colt, could you please make a video or a short on how you have been archiving your sessions and projects. Thank you kindly. Wishing you a great day!
I hopped on the 48 team maybe 6 years ago. I had always worked in 44. So I exported a beat at 44, then 48, 88, 96. I may have stopped there. I noticed that as soon as you get to 88, the sound of your digital instrument vsts changes very drastically. Not so much for pianos or other live sampled instruments, but especially with synth instruments like a warble bass, a saw, stuff like that. So I stuck with 48.
But like you said, if you’re tuning a vocal or time stretching, 96 is easily the way to go. Just gotta think ahead before you start tracking vocals. Like is this guys voice gonna need a lot of work? If so, 96. If not, 48.
I equate it to how the 4k TVs look (on the stock settings)... more accurate for sure, but it also ruins the lighting and makes everything look like you are standing there in the room. The reality of it ruins your ability to immerse yourself in the show.
I wouldn't say its entirely an accurate description. consider how 4k resolution compared to 1440p or 1080p is a dramatic and clearly perceivable difference. whereas however when it comes to audio, 44 kilohertz and 48 kilohertz already covers all of our perceptible human hearing range. 96k contains more analog information than we can even experience as humans, and the auditory playback hardware you have might not even support frequencies that high anyways
therefore a more accurate description would be, comparing something like 16k video, to 32k video. both are quite overkill and over what we're capable of seeing anyways. however the higher resolutions always provides more flexible options when it comes to zooming in and such during post production
your analogy has a good idea that I recognize though. having the most "realistic" version of something isn't always what we want
Turn off tv processing that "enhances" images and you'll be shocked at how good 4K can be. Just a suggestion.
Don't think it works to use a cross-sense example: the art experience of eyes and ears aren't the same. Like, does it make sense that to say lI like the smell of avocado, so therefore green must be my favorite color!". No! :)
Also... many say that true analog is the best (like no digital in the process: tape, to say, vinyl)... that path is which is infinite samples per second (or what? The space between electrons?:)... which is closer to 96k than 48k. It's all taste at the end of the day. Do what you like.
Like a Ultra HD camera - it's more "accurate" but it usually doesn't look better. A softer resolution often looks nicer.
This is kind of the same reason why I like Apollos. The vibe feels better than some converters that are more clinical but technically more accurate or have more clarity. Apollos sound more analog.
recently been experimenting with both, and i appreciate you making this video.. helped me make my decision without spending any more of my time, and storage, trying to figure it out.
I am personally suspicious of this "clinical" feeling of 96khz.
48khz audio perfectly "captures" the audio, as far as how we perceive it.
Humans can't hear that extra octave of bandwidth that 96k provides.
Even in terms of envelope of the audio (transient response), I don't think that the differences could be loud enough to be heard.
Plus, humans are very bad at hearing the true envelope of sound, which is why static phase shifts are usually inaudible.
I think the biggest difference would be in the lowpass filtering on input and output stages.
96khz audio can afford to use more gradual filters with higher cutoffs, meaning the top end of the audio is better preserved.
There's less filter ringing, less phase shift, and less treble rolloff.
Still, these differences are mostly above 10khz. So at best, this is a very subtle difference, comparable to two different transformers. Or an extra couple dB on a high shelf.
And the filter ringing is literally at 20khz. I would wager that 20khz ringing is inaudible to most people.
In the box, more differences can arise.
EQs cramp at lower sample rates. Most people will just adjust the bandwidth and gain to compensate. So if anything, this is a workflow difference.
Distortion, obviously, acts differently at lower sample rates. If anything, the aliasing of 48khz is the source of the grungy character. Still, oversampling is an easy fix for aliasing, and it induces minimal side effects.
The main differences between sample rates occur when pitch shifting or time shifting signals. Also, the CPU and storage are notable variables. The sound is the least important factor when choosing a sample rate, so this should be a non-issue IMO. Not a single listener will be able to tell the difference.
Record in whichever makes sense.
For example, kick drums can be stretched minimal side effects, since they have a small bandwidth and mostly low frequencies. So 48khz is okay. But drum overheads might benefit from 96khz if you're planning on mercilessly stretching them to fit the grid.
Anyways, that's my two bits. I should do tests to back up my claims though. Maybe the input/output filtering is more different than I think.
Given human hearing, the only audible differences for tracking (not time-based editing) at 96kz for mic’ed performances are likely due to antialiasing filters. The audio quality for tracking shouldn’t improve from 44.1 to 192Kz (at least for humans). People are entitled to do what they want and have their preferences and rationales. Yet, most mics, preamps, etc. aren’t capturing or producing frequencies at or above 20Kz for a converter to convert. If a person likes the filtering differences of a higher sampling rate, go for it. But 44.1 and 48 are used to make masterful recordings on a daily basis.
I typically use 44.1 or 48k, but have used 96 or even 192 on occasion when I waned to slow a track way down for certain types of effects. You get more fidelity (like you mentioned wrt elastic audio).
I still record at 48K for compatibility, since I move from place to place and some places are not ready to work at 96, and agree that for some reason I prefer recording on 48. However I’ve been trying up sampling for mix or mastering and it is an impressive thing cause when you’re dealing with more complex waveforms and more harmonics that’s when 96k shines also if you have any analog processing like a bus compressor you can clearly hear the difference specially in how the harmonics are better preserved and you don’t have to deal with aliasing.
That crunchiness you're hearing is probably a very small amount of aliasing noise which can be interpreted as color or extra harmonics.
Can you split a microphone into two signals and have two PC's recording at the same time with different sample rates?
There must be a way to run the same performance into two separate systems, one set at 48K and one at 96K. That's usually how A/B tests are done. You would need two computers and two interfaces and a whole lot of signal splitters, I suppose.
That’s the only way you could really do it, and then how do you record the same performances to multiple systems at the same time and have it line up with everything else that was previously recorded on each system? You would have to slave one pro to rake to another so they functioned as one even though they were two completely separate rigs. And I’m Too dumb to even figure out if it’s possible, or how to do it lol
I'm not a PT user, but I can't see why you couldn't lock the two systems using SMPTE or MIDI time code.
There is a lot of extra "button"- pushing for record enabling and such but hey, this is science.
It isn't supposed to be easy.
Engineers have synced multiple analog tape recorders for 50 years, and digital is suposed to be better ;-)
Anyway, interesting video.
I hear you for ITB mixes but I really have to disagree for OTB mixes. High voltage summing and slamming mix bus transformers makes it sound like a record and sounds better at 96 ime.
This pretty well covers why I'm comfortable with 48, and to be honest, why I was comfortable with 44.1 before that. I don't do any fakery in my music; if something I record isn't right, I re-record it. Autotune and time stretching aren't part of my process. And even if they were, I suspect many people focus on the numbers because it's something you can solve just by spending money, while actually making art of substance is harder and where the more important decisions are. If someone is just making soundalike filler, the fine points of bit rate are academic because you're just dumping noise into the void. It's pretty common to always want to be using the most of anything. But the idea of future-proofing seems futile. 5 years from now the standard will probably have doubled, yet all those pro albums recorded in 44.1 won't sound any more out of date than they do now. When I first upgraded to a system that could handle 96, I tried it and was underwhelmed; it seemed not really worth it.
Digital Reverbs sound really lush in 96K. The record head is slightly more accurate meaning there is less latency when recording. In order for y'all to hear how good 96K you need a high end DAC. Being that I monitor with Grace M905, I can attest that 96k sounds amazing! Now that music can be released in higher quality formats other than the classic 16 bit 44.1; It's worth trying for higher. I am inclined to mix at 96k, but the PROCESSING involved is no joke! Everyone will opt out of 96K really easy when they don't have the CPU power. If a session is 44.1K I'm fine with that too, because it is super nuance to the average ear. The part that no one cares is actually very true. It is more so an audiophile preference to dabble in 96k.
What about recording at 88.2 so when you render down to 44.1 it's an even division of numbers...?
In the late 90's I started out in Digital Performer, recorded at 96K and never looked back. In Western PA we don't have a lot of Nashville musicians and much manipulation is needed, maybe that's why I chose 96k.
I think recording at 48k instead of 44.1k defintely is the biggest audible difference. Yes, 96k/88.1k can sound a little clinical. But as a guy not using an external clock, doing final SRC to 44.1K and 320mp3 or AAC's, 48K works fine to record in. And I use a lot of virtual instruments, so recording with higher than 48K isn't usually going to make a big difference.
I have read a number of accounts from people with larger setups using external clocking that going to 96k can produce more "depth" and "width" in the sound field.
on my own method, I just record at 96khz then go back to 48khz when composing. my main genre is EDM but I also record vocals and instruments beyond synths because I include them in my tracks. my reasoning to mainly record stuff at 96khz as you already said when doing pitch correction or stretch it does helps. so having it at higher sampling rate as the original recorded stuff pretty much solves a lot of problem, but I keep the main editing and processing at 48khz. it's also double edge anyway since if the client wants at 96khz I already have all of the recordings at highest rate.
To me higher sampling rate is basically like high speed framerate in video. it's like 48khz = 60fps and 96khz = 120fps, so it does sound more natural at higherframe rate since you can see more frames, smooth motion!
I'll have to test with guitars but recording bass, I notice that there was so much of a difference between how the bass sounds analog coming through my monitor and how it sounds recorded that I do max out the the sample rate and downsample it AFTER
As an ‘audiophile’ I store most of my music as 96K Flac and upsample to 768k in Roon before streaming to my DAC for playback, and yes it is super clean and super ‘clinical’ and every bad mix and bad master out there stands out like a sore thumb. But I am the minority.
I've switched from 88.1k to 48k, since im using more digital hardware synths and sample based drum plugins that are pointless to go higher.
Just use reaper oversampling where it might sound better (some saturation)
This concept reminds me of the differences between filming 24 frames a second and 60 frames a second. 60 frames per second looks real but 24 frames per second looks cooler
Personally I do prefer 96k, but the difference is SOOOO subtle. For now I'll take the extra tracks/voices/plugins etc of 44.1/48 while computers catch up. The file size thing is real though... I worked on a Three Days Grace record that we tracked at 96k. The entire master backup at the end had to be spread across 3 2tb drives. Drives are cheap now, but the logistics of making sure those backups were correct and the end of every day was a bit challenging.
i did a record at 88,2 in the years of TDM and PTHD10, never did one again, the amount of disk space, the halved plugins on TDM i could use, with no sonic benefits to justify all that "cost". my luck was that being in a well furnished studio i could use almost only hardware compressors and 50% of the eqs where analog too. otherwise i'd have to had converted all the tracks to 48, and with those computers that woul'd have took ages
Subscribed because of your continuing honesty.
Do you think that maybe the reason the 96k transferred down to 48k sounded like it was almost there might be that the plugins handled the 96k different than the 48k. When I get back to the US I will try a few experiments with this idea. Kind of like I did when we went from analog to digital. I would record all the drums and bass on analog tape and transfer it to digital. That was when I was on Sonic Solutions as my DAW. At that time Sonic led the pack with the best sounding DAW, because it was 24bit, and the others were not. It worked well transferring the Drums and Bass to the DAW. It became more of a question of time then, so I axed that idea pretty fast.
I have a feeling the reason why converting the session to 48K was only halfway there is because it’s just cutting out half of the information. Rather than relying on the converters to choose that information if that makes sense.
This must be like the difference between 60fps and 24fps. 60fps looks like reality tv. 24fps looks like cinema and generally is nicer for our brain to process during action sequences especially. If you look at Gemini man in 60fps 4K, it actually looks like actors acting rather than witnessing something special.
Unfortunately, this perpetuates some myths (as does the 44.1 vs 48kHz video you did a while back). That's not how sample rate works. If that was how it worked, 20Hz content would be many times more accurately represented than 1kHz. That's just not how it works once you exceed two samples per cycle. Six points on the same circle don't make the circle any more round than three. Sample rate only dictates the highest recordable frequency. Having done tests myself with recording to two identical systems simultaneously using custom splitters, and monitoring through a sample rate agnostic converter (Benchmark), as well as exporting each to its opposite sample rate and doing double blind tests, I can tell you that my experience and measurements in the controlled testing do not match your subjective results in an uncontrolled testing situation with multiple variables.
Colt I like what you have on 96k verse 48k, I tried both but because in the beginning 48k was widely used in video forever. I found from 90s on that working 48k works for audio and video clients so staying that for economics really. Backing up 96k requires double the storage space. Keep moving forward Colt the list will get smaller someday!
It's all about aliasing in plugins and not forcing realtime oversampling. With 4 plugins in a chain all upsampling downsampling etc.... it can get a little mushy. I work with Analog modulars a lot and 96k (well I use 88k if I'm going higher sample rates) and they can make plugins alias a lot at 44k. Can I hear a 96k or 88k recording wav vs 44...nope. It's all about plugins :)
had the same experience, i think its probably the filter in the interface at the nyquist frequency
Every engineer I respect records at 48k 24bit. That is what I do as well.
I don't tend to do lots of time stretching or tuning as I am of the opinion that getting it right at the source works far better than fixing it in post.