What is the best sample rate to record and mix your songs at? Is it 44.1k? 48k? 192k?!?!
Which sounds better?
Here’s another question: does sample rate even matter?
I want to break these questions down for you once and for all in today’s video so you can make a smart decision and move on with your musical life!
great video. second half of the content even better, and ultimately more important, than the first. love the way you keep it real.
GREAT info, simple,clear. Thanks!
Graham, agreed on everything you conveyed. I record at 24/48 and the main thing that I perceive is an extended dynamic range that I get with this configuration over the 44.1. Now whether I lose that when its taken down from 48 to 44.1, I can’t say. But for what I perceive as an improvement at 48 with minimal additional compute resource, I feel like its a great compromise.
You get the extended dynamic range because you record at 24 bits.
Don’t get confused here: Sample rate (44.1KHz or 48KHz) affects “quality” of sound in terms of fidelity. Depth (16 or 24 bits) affects dynamic range. So, of course, you get extended dynamic range recording at 24 bits over 16 bits.
Ok, understood, thanks for the clarification.
The first few seconds of the video is really misleading – Sample Rate is absolutely NOT analogous to frame rate – higher frame-rates playback with less change between individual frames and rely on highly imperfect vision to create the impression of moving pictures; audio samples are used to describe a curve, the sound is not a bunch of individual sounds played one after another, it is a mathematically re-created, unbroken waveform (imagine a straight line – you know where the start is, where the end is, you get the EXACT same line with those two points as you do with a million points. Audio sampling is the same with sine waves.
YES. This what’s important. Using a higher number of bits gives you more dynamic range within the DAW which gives you more room to manipulate levels using your plugins, etc. without going over “zero” and getting digital distortion. Then when all the editing is done and you have your levels correct, you can convert (with/without dither) back down to 16 bit. So, as mentioned, higher bits gives you more HEADROOM while mixing.
What about aliasing? For example, we have a frequency 9kHz which had 1st harmonic at 18kHz & 2nd harmonic at 27kHz. At 48 kHz plugins which generate harmonics (like ampsims & analog emulations & had no internal oversampling) will process 9kHz with 2nd harmonic at 21kHz (instead of 27kHz) because of aliasing. So the processing affects the sound differently at differrent sample rates. It may cause an undesirable effect on upper range of spectrum. Ehat you think about that?
Thanks. Shortcuts for the time-pressed 😉 – The important (reinforcing for me) message here starts at 7:46, and you have heard all you really need by 9:12.
Great explanation. I’ve always done 44.1 but honestly didn’t understand what the difference was. Your frame rate analogy and explanation was great. Thanks a lot G
Nice presentation Graham….. technology is expanding exponentially these days, from the software techs at Apple, Microsoft, Burroughs, Texas Instruments, IBM,etc…. I am hearing that very soon the hardware will be coming to match the current abilities of the software now available. But your explanation is excellent advice for all the people new to recording presently.
Having been in the recording industry since the 1950’s, it’s been a whirlwind of changes since then…. but I do miss the golden era of the 1950’s…. talent was talent or it was not…. no gimmicks (software plugins) to create a supposed talent (ability).
For the past 25 years my emphasis has been in the area of Advanced Forensic Audio Restoration for people and companies around the world.
Anyway, ALL of your presentations are ‘spot on’ and timely…. I have encouraged many people around the world to click on your You Tubes and to visit all that you have to offer and posted. Please continue offering and sharing your experiences and advice.
Well done….
I’ve tried higher sample rates and as much as I wanted there to be a big improvement in the final product, I can’t honestly say that was the case. And… running my system at 96k created a HUGE problem for me in terms of CPU performance. I’m back to recording at 44.1 or 48.
Perhaps too esoteric for your base audience, but it bears remembering that most DAWs and plugins oversample when performing DSP calculations so your 44.1khz files are actually being processed at higher sampling rates…hence some engineers’ possibly objective perception that higher bandwidth capture files do result in better sounding output, even after final downconversion.
I haven’t searched yet to see if you’ve answered this similar question regarding bit depth, but do you make the same “do it only if you feel it” recommendation? Knowing that PT is calculating at least 32-bits, would you defy application preferences and force files to be captured at 16-bits just because that’s the format your mix will end up distributed at in order to conserve cpu power and storage? Historically, I like to compare the workflow of using high bandwidth files for capture to the way analog tape engineers used to choose different bias settings and tape speed for tracking decks than for mixdown and mastering. It does make logical sense to assess your planned chain from start to finish and make informed decisions from the beginning of the process. And the tools we have available now are so much better at saving us from making bad technical decisions than only a decade ago that most listeners probably won’t hear the difference, but performers and producers must be more picky than “most listeners”! Just my $0.02, thanks for all your great work and excellent contributions to the world of audio education!
This is a debate that rages on through the ages, only the terms like “sample frequency” “bit depth” “rpm” “track width” “oxide formula” “bias level” or “ips” change with technological advances.
Nice explanation, Graham.
So, this leads to other questions…
If you think your song might be used in a video, should you mix at 48 instead of 44.1?
Also, since standard CD audio is 16 bit at 44.1KHz why do so many people recommend recording at 24 bit (or more) and dithering to 16 bit when doing a final rendering?
As I understand it, record labels are beginning to require that files must be submitted at 24bit 96kHz. However CD Baby and other indie distributors still require 16bit 44.1kHz files.
CD Baby takes file at 24 bit 96k now. So does ITunes and several other electronic distributors.
I record at 44.1Khz. It’s funny when I first bought the HD192s, I thought it was going to make a huge difference. It did but it wasn’t the sample rate. lol. I did test with my friends to see if the could hear a difference, did not tell them what was different about the test audio. Two, out of 10 people were able to perceive a difference between 44.1k and 96k. Eight were not able to tell. I couldn’t hear it myself. I felt a bit ripped off that 192k didn’t give me analog tape thickness. For most of my stuff I’m recording at 44.1k. Now that CD Baby, ITunes, and several of the online distributors, are asking for 24 bit 96k. I’ll probably master at that sample rate for album/Ep releases.
Thank you! If I mix at 44.1k, 24 bit——-will I still need to do dither and noise shape when rendering to a 44.1k, 16 bit file for CD? (I currently mix at 48/24—-and dither and noise shape when rendering for CD).
Bravo my friend…you nailed it! I record/mix at 44.1. With my hardware and software I have the ability to record/mix in a zillion different sample rates (including DSD) and at the end of the day…it’s always 44.1. I’ll switch to 48 for video…otherwise I’m at 44.1. It works just fine! Thanks Graham 😃.
Cheers
Hi Graham
I like to record at 48KHZ simply because the track is ready for video if required.
Why not record at 96k for a better capture and then down sample to 44.1 to mix?
I use this formula in both my audio restoration and mixing systems.
I definitely hear a difference on the restoration side when I compare audio recorded at 44.1k and at 96k. The audio is bigger, fatter, & has more detail in the 96k files than in the 44.1k files.
My 2 cents.
Nice video and well explained.
Honestly i never go below 96Khz unless its a live recording. For classical and Jazz i sometimes use 192 Khz. Yes, I have a 12 core mac pro but the point for me is that recording 32/96 will give me the flexibility and delivering files for Vinyl pressing they require 24/96 formats.
What do you use to record at 32 bits?
What about bit rate. (16,24)? What are your thoughts on that?
Thanks, Tim
As long as you’re careful to record at proper levels, the bit depth doesn’t matter as much as sample frequency.
Higher bit depth gives you more head room to avoid clipping and can help rescue a recording that wasn’t recorded with enough gain as long as you’re recording on equipment with a low noise floor or have denoising software.
Sample rate affects higher pitches much more than lower pitches. Wobble is present but the severity depends in part on how well your digital signal is converted and smoothed back out to analog sound. A lot of people will not be able to tell the difference between high and low quality because they are listening with bad digital to analog conversion, poor speakers, poor amps, or lossy formatting like mp3.
A lot of of the opportunity to maintain quality of sound is in the hands of the listeners. Do you give up quality from the start for those few people that currently have a good setup or might have a better setup in the future?
According to the Producers and Engineers Wing @ Grammy
https://www.grammy.com/sites/com/files/delivery_recommendations_for_recorded_music_projects_final_09_27_18.pdf
Graham – the case I would make for high sample rate recording, 96/24 for example, is the archival aspect of it. In the old analog world, the original magnetic tape master was always much higher fidelity than the distribution media; vinyl or cassette or consumer reel to reel. So thank God we have those old tapes, the only definitive records of those performances we have. Aside from the lack of tape hiss, digital recording has LESS fidelity than tape, so since we are not archiving performances on tape anymore I would argue for the original recording to be the highest bit rate possible, in order to approach the fidelity that of tape.
Yes – I mentioned this as I see our industry moving to higher quality recordings for playback eventually so I can see the case for moving in that direction.
If you are using a fairly modern audio interface that utilizes oversampling, you shouldn’t really hear a difference between sample rates. I think all modern interfaces from very cheap ones and up utilize oversampling. A guy did a test using an old cheap interface from the early 2000s which captured audio perfectly. Unless you can hear over 20khz, it makes very little difference. Most people cant even hear 20khz especially if you are over 25 years of age, and/or listen to loud music constantly, and perform live. Also higher rates can cause more problems as well as ultrasonic waves can create distortions in the audible range if using speakers not designed for ultrasonic audio. That said there are benefits, lower latency, potentially reduced aliasing on plugins which also can be achieved selectively at lower sampling rates with plugins that utilize oversampling. Many plugins (including some stock plugins in certain DAWS) include oversampling though and some even enable by default. IMO you are better off at lower rates like 44.1 or 48 and selectively using oversampling when needed as you will have more CPU power for more plugins or just using a better plugin. If you need lower latency you can try higher rates or you could also try an interface that offer lower latency.
Okay Graham, I love your channel, and it could be even better if you get your sample rate theory correct once and for all. Okay…the only real thing you get when you mix at super high sample rates are higher frequency ranges (see Nyquist theorem), and some added distortion caused by ultra sonic frequencies in the lower audible frequencies. So, if you are doing scientific work, or will be slowing down the audio substantially, nobody will hear the difference. The output is not in stair steps, the DA converter converts the digital into a smooth and accurate wave form. If you hear a difference, it’s most likely because your I/O interface may be optimized sonically for that rate…a little bit of possible trickery by the manufacturer. By the way, your buddy Dave Pensado and a lot of the other engineers that are well past 40 years old, as good of mixers as they are, can’t hear the upper frequencies everybody says are there. Most of the upper frequencies that are there are most likely a type of distortion. Have you watched Monty’s video yet?…it doesn’t sound like it to me. https://www.youtube.com/watch?v=cIQ9IXSUzuM
In the end…it won’t be the super high sample rates (88.2, 96, and 192 Khz) that wins…it’s the talent and skill (you know this). Warren Huart mixes in 48 Khz, but says they will be moving to higher sample rates because of future proofing. This is necessary IMO for big timers to stay with the super rich boy snobbery. Okay…enough…thanks much for what you do!
Tim, Graham’s discussion about stairsteps is his way to visually communicate the difference in the waveforms produced and is an accurate description which this diagram illustrates as well. Each plateau effectively represents a snapshot of data about the real waveform in digital terms. The more snaps the more accurate the representation on the original curve.
2.bp.blogspot.com/-VhAQjHl_Kd8/U3dK-vhlwKI/AAAAAAAADfg/MFcADbcfFU4/s1600/Analog-Digital+frequency+examples.png
For the last sentence in your comment….This is generally only true if you need to record higher frequencies. The DA converter creates a smooth and accurate mathematical analog solution between the points. That is why you only need two data points at 22050 Hz to get a full wave form. After this, the frequencies have to be low pass filtered below the top frequency in order to not have aliasing problems (ambiguous solutions due to lack of data). In reality, the samples are more like lollipops and not stair steps. A stair step implies that the sample is held for a period of time between samples….which is not the case. However….having just said that, I am looking for a great argument to show me otherwise. I looked at the graphic you linked to, and this is considered by top experts to actually be an inaccurate representation. People are making different arguments about psycho-acoustics and that the upper frequencies are felt and not heard. Interesting idea, but won’t make or break a great song. see the link to Monty’s video above in my first comment. Keep in mind that I am not trying to insult Graham. Graham has a really good thing going and I respect that a lot! He is also, probably, a better mixer and musician than I am. This is a fun topic for me…just like how most people understand aerodynamic lift in the incorrect way. But…I too will be found wrong on a number topics too….I hope I can gracefully accept it when it comes…as it already has many times! THX! Sorry soooooo long….
My computer has a quad-core processor and 6 gigs of RAM, but Audacity still crashes with 24/48, 32/48, or even 24 or 32/44.1. I don’t know if it’s a RAM issue or a DAW issue.
It might be an OS issue. Sometimes installing a fresh OS will help. It might also be a river, CPU, RAM, or sound card issue. Do non audio programs crash as well?
Solid, common sense advice. That’s why I keep you on my short list of audio advice gurus.
The difference is real, not placebo. Your brain can perceive the data gaps, even if you don’t consciously register it as sonic flutter, and it affects your perception of the fidelity. The only way to surpass the brain’s ability to process data gaps that fast is to record at 192 kHz, and retain that sample rate as much as possible. With that much data density, the gaps become imperceivable, and you have something that is indistinguishable from a live performance. The problem is with retaining that level of data density.
You need to consider generation loss. If you record at 44.1 and then resample it at 44.1, you will have less data than the original file, because it’s a sample of a sample. More missing data means wider time gaps. You lose a lot more data than if you record at 192 and resample down to 44.1. What’s worse is that with each resample, you lose more data. This is true regardless of the resample rate, and whether you upsample or downsample. That’s why 44.1 recordings resampled at 192 kHz may sound higher resolution, but they still sound like recordings of recordings.
The amount of data you can capture depends on how much is already there; more data in the sampled file means less generation loss. That’s why people can hear a difference between 44.1 files sampled from 44.1 and 192 files, and even the difference between 44.1 files sampled from 96 and 192 files.
You also have to consider that editing is always destructive; it always results in data loss. So you have to consider both editing and resampling generation loss. This is why eventually, 384 kHz will become the standard. It’s the only way to actually get an end product with a sample rate density of 192 kHz or better.
Hi Graham,
Nice information, but in order to underline some of the previous comments, I will take the time and re-phrase some important aspects:
Physics is not an opinion. It is true or false. The Nyquist theorem shows, that you need twice the sample rate to re-create a frequency exactly. Every additional sample does not help to improve the quality. So going higher than 44kHz means that you targeting bats as an audience.
Higher sample rates introduce additional noise. If you are checking the specs of a sound card like RME, the noise floor gets worse.
The bit rate is a completely different beast. Going from 16 to 24 bits means that you are going from 65k levels to 16m levels. As a result, even low amplitude signals getting a nice resolution – what you will hear.
Following your mixing tips it is important to understand, that 24 bit will allow you to mix with very low signal levels while having a audio signal resolution in amplitude that is way higher than a human ear can differentiate.
Last not least: recording in 48/96k is a bad idea if you are going down to 44.1k, because you get floating point results. So you introduce an error which no gear can catch. This is why dithering is used: the additon of noise covers the artifacts of that convertion.
Hi Graham,
thank you very much for the video. It is an important topic, so I feel the need to do some additions, even if some points were said in the former comments, as well.
Physic laws are not an opinion. With that in mind, the Nyquist theorem shows without any debate, that you need twice the sample rate of the frequency you want to recreate. Not a little bit less, not a little bit more. The question is: why 44.1kHz? It is 20K plus an overhead for the HPF before/after the ad/da converters. This means if your target audience is human, you will be just fine. If you are targeting bats, 96K will be the right one.
The real difference is the bit rate: it defines the resolution of how well differences in level within a signal can be represented. With 8 bit there are 256 steps to recreate your analog waveform level, at 16 bit we talking about 65 thousand steps and at 24 bit more than 16 million steps.
The human ear can differentiate roughly steps of level.
Mixing means a lot of additions and subtractions – which is not a problem. Even multiplications are fine. But you always get errors, if a signal will have a longer mantissa than the number format allows. In order to reduce this effect, the internal mixing processor is working with 32 or even 64 bit resolution. But this has nothing to do with the sample rate.
In addition, there are some issues with sample rates like 96K: you will add high frequency noise, which will add up. This is no sudo science. Check out spec sheets of soundcards like of RME: the noise floor will get worse, if you switch to a higher sample rate. Secondly, if you convert a 96K track down to 44.1K, it won’t be an integer division. It is a division by 2.177 which will introduce converting errors. Dithering will reduce the effect by adding noise, so you will not hear it as bad. But why recording for bats and adding noise to get a result even not as exact as at 44.1K?
Long story short: If somebody thinks that the Nyquist theorem is wrong, he should prove that in order to get the Nobel prize and working with 88.2K. Than they can down to 44.1K easily without dithering and the only issue would be additional efforts for the CPU and storage.
Amen!
Nyquist theorem was not and is not wrong. Your interpretation is wrong.
Nyquist et al. Have stated the “minimum” frequency capable of recreating a wave. If you shift the frequency from that minimum slightly you get low frequency wobble.
According to Nyquist-Shannon theorem, there is a theoretical way to digitize all information in an analog signal at an average Nyquist rate of sampling if you can also choose at what time to the your samples. In other words, to record a signal up to 24 kHz you need a variable sampling rate that averages out to 48 kHz AND you need to pick the ideal points in time when those samples are taken. Not practical…
There are artifacts that are more noticable at higher frequencies and measurable but not audible at lower frequencies. It’s one thing to record, yet there are also issues with how the digital signal is converted back to analog which can be measured but I’m thinking is probably more prone to placebo effect.
How does time-stretching interact with sample rate? For example, if I record at a sample rate of 44.1k, and I stretch a 1-second clip out to 2 seconds, then is my result playing at 22k?
Sometimes I’ll notice artifacts in stretched clips. Could I prevent artifacts by recording at 88k, so the stretched clip ends up at 44k?
Great vid Graham. I was asking this question only in regard to youtube. I mix at 44.1, throw my mix into spotcut and load it up at 48….does spotcut do a good job?????
Much more important though is does it get converted to mp3. I watched a video showing how converting to mp3 adds gain to the audio. If my mixdown is “Hot”, it may have overs after the conversion. So….unless I’m a super pro, I should stay down at -3db. I also record at 16bit. I just have to make sure I stay well above my sound floor. If I compress with my threshold too low, I’ll bring up the noise. I sometimes gate it at -35db before I do anything….noise gone, set the input level on my first FX a little lower and MAGIC…..I have head-room.
I have another question. ISRC codes. Do you install them yourself? I know we can purchase them ourselves but, how are they attached to the uploaded sound files?
Hi Gary,
Your online distributor can assign and embed the codes for you before they distribute to the platforms.
If you get them yourself, they can be embedded during the mastering process using a program such as HOFA DDP.
Cheers
Thanks Graham
One thing that you did not talk about is dithering. I am a beginner but don’t you have to apply dithering to o forma higher sample rate to a lower sample rate ? and doesn’t it degrade the audio to apply dithering? I have always thought to avoid dithering it is best to record at the sample rate your intended audience will listen to music at which for streaming and CD’s and mp3’s is 44.1
As always, Great video Graham. It’s best to go with the standard 44.1kHz
The way I see it: If in the future sample rated go up, you may not gain anything in audio quality; but you definitely won’t lose anything either. On the other hand, recording at a higher sample rate and having to reduce it will cause a loss in audio quality. It’s kinda like “Lossless Audio Formats” which converts your file from something such as WAV. to MP3.
The sound will lose slight dynamics.
Best commentary I’ve heard in a long time. Hang in, keep it up. Appreciate your time and effort.
Awesome video! I’m at 44.1 for the same reasons. Any chance I can think less and simplify, I take it. It allows me to focus more on the creativity, which is a bigger win!
The human ear is capable of hearing 20kHz and those are mostly kids. An old war horse rocker like myself will be doing great to hear 15kHz on my best day. I’d bet 10-12kHz would be more like the upper limit of my hearing at best. Sampling at 44.1 is more than three times my ability to hear those highest frequencies. If sampling at 96,000/sec will improve my hearing I’d be all for it. But, unfortunately that’s just not realistic. As musicians and mixers, paying greater attention to lifelong ear protection is considerably more important than considering sample rates that exceed the ear’s ability to hear. Great video as always Graham. Thank you for all you do!
Great content as usual Graham and thanks again for all you do.
I also record at 44.1 People can argue all they want about what sounds better but in the end, like you always say, it’s the mix and the recording of instruments, and the song itself that actually count. If your EQ is off or you have too much compression it doesn’t matter if you record at the highest sampling rate, the song will sound bad. I released an EP through Tune Core last year distributed to all the digital stores/sites and all songs were accepted, as is. Not saying that they’re the best mixes but certainly good enough to be on the streaming sites and I have Graham to thank for that. For the new comers out there, watch ALL of Graham’s videos, you’ll learn so much! I’ve also taken several of his mixing courses… all worth it. It’s because he doesn’t get lost in the minutia. He gives you a process, a work stream based on logic, common sense and experience.
I’m a fan Graham, but a lot of your information here is simply wrong.
This movie that often gets shared in sample rate discussions will dispel all that’s not right here.
i was about to say it….digital is just as smooth as analog
Well John, that depends on the output filter on your DAC. I design electronics with ADCs & DACs (that sample up to the MHz ranges) and, yes, depending on how it is designed, you can get steps. If you designed the output stage of a DAC to smooth a 192 KHz sample rate and then you output a “sine wave” only feeding it at 10 Hz sample rate, yep, your getting some pretty nasty steps.
Now I understand that more is generally done to dynamically smooth audio DACs, but I have my doubts that typical audio DACs are smoothing enough to make data with only 10 samples per second into a pure sine wave as that video suggests (regardless of the bit depth).
By the way, “human hearing” doesn’t stop at 20KHz for everyone (mine extended over 30K when I was young – ouch). That’s an oversimplification and Nyquist has also been shown to be an oversimplification when used to justify limiting audio sample rates to 44.1HKz, but that’s a whole other type of blog subject.
…So after my rant, the moral of my story is that no, most typical sample rates are not going to make true stair steps on most audio DACs, but Graham is simply explaining to the average person who is not an electrical engineer the basics of converter sample rates so that they can make more informed, common sense decisions without spending years learning how the insides of their audio interface works. I think his explanation does that just fine…
Graham,
I really appreciate you videos. I’ve learned quite a lot. This has nothing to do with mixing, it’s kind of personal. I think you need to change your lighting. Watch your video and look closely at your eyes. There’s something spooky about the border between the iris and the pupil.
That’s proper lighting. I working in television and film, and the “halo” in the eyes means it’s lit properly. Check movies and television, and you’ll see it. It makes the eyes pop. Once you see it, you’ll never un-see it.
I don’t know how proper it might be but oh yes it looks creepy.
Hi Graham.
Great explanation audio v video / frame v sample rate.
I started with tape and when I switched to 16 bit 44.1 recordings and I was happy with most of the initial projects except for digi glitches/retakes when headroom was exceeded. However when 24bit 44.1 was available, I really noticed changes in some areas. In quiet sections of recordings, anything with HF content seemed to be noticably clearer and not as harsh. However, I found the difference less discernible in LF content – eg: kick, toms, bass etc. I used the same interface when I changed, but with a newer computer, could run 24bit. It may be a placebo effect, but I have found that I noticed a clear change when I listened back to older projects and then new. Thank you for your posts and videos and encouragement to many people.
Regards, Greg
There is a difference in switching from 44.1k to 48k that you can easily hear:
I noticed that editing vocals (pitch and time correction) with the Cubase variaudio (Melodyne like tool) is way better and gives me less “digital artifacts” when working in 48k. Referring to the explanations in the video it’s logical.
Regards, Heiko
Strangely, nobody referenced the fact why it is 44.1. According to Niquist least restorable frequency is twice lower than sampling rate so 22.05kHz. This is slightly above upper limit of human hearing. Now to more specific technical things. You actually cannot restore 22.05Khz accurately when sampling 44.1, it will be recorded with attenuation as to avoid aliasing your input must be filtered at 22.05 max, but usually filter is set to slighly lower frequency. Filters are not perfect brick walls, we are talking analog stuff here. And that is why recording analog to digital with high sample rates matters. Initial oversampling required for quality in high end of spectrum. You can probably compensate this using eqs after recording with lower sample rate anyway. By the way after recording with higher rate it is quite possible to downsample without loosing much in high end as digital filtering is much better than analog and you can have very “perfect” cut. That about recording.
A little on processing.
A lot of plugins actually upsample your sounds using clever interpolation techniques and downsample at the output. This can be vital for what these plugins do. For example, your mastering limiter will almost certainly do this trick for what they call ‘intersample peak detection’ and better quality limiting. Having higher sample rate inputs will mean better quality interpolation inside the plugin.
There is no good reason to go crazy and record & mix at 192K, but there is quite a good reason to go slightly above 44.1 standard. Personally, 48k sounds good to me and I do not see any reasons to go higher at the moment.
I use 48KHz because 48 was my old address when I was a kid.
I use 44.1 KHz 24 bit and it works really well for me.
Unless I mix for film where I normally record and mix in 24bit/48kHz, I use 24bit/44,1kHz. I get an extended dynamic range and less fuzz for the dither plugin I use.
In my excperience the dither plugin is the critical point when converting to CD or MP3 quality.
The common man is anyway not able to hear a difference. They usually don´t listen on high end equipment and do not have a “trained ear” to hear the difference anyway.
Most people today consume music, they don´t listen to music.
Hi Graham,
Unfortunately what you said about digital waveforms being “stair steps” is incorrect, common misconception.
When the signal is converted back to analogue (DAC), then it will once again be a continuous (smooth) waveform as what went in.
More info: https://www.youtube.com/playlist?list=PL-z5SavzzUq9UmrNdc8fFroHL1JzVCWEL
Right – when it’s converted back to analog. But what about when it’s still digital?
You’re missing the point, what you are hearing is always smooth. If you can hear it, it’s been converted to analogue via DAC. What you said gives off the impression that digital is inferior because of “stairsteps”, which just isn’t true.
How strange a question. For any digital processing the music has to be converted to digital. That is the fundamental reason we are having this conversation in the first place!
I thought so too!
Hi Graham,
There was one other issue you might have mentioned.
Without dithering your are leaving the quality of the changed sample right up to Apple.
I usually let ProTools do the sample rate change and dither down to 44.1K.
I actually, having done a lot of experimenting believe you are better off recording at the lower sample rates that are closer or the same as 44.1k
I actually usually record at 48k as a standard unless the client wants something higher.
The trick is to MASTER at a higher bit rate to get the maximum benefit from the plugins you use working at a higher resolution and then dither down.
This is not a sample rate issue but rather procuring the best sound from the plug in at that bit rate.
If people don’t understand dithering this may be helpful.
Thanks for sharing Michael
As always, thank you Graham, you are wise beyond your years. I love your down to earth approach!
Roger
Yep, as usual, Graham has a way of getting to the common sense of a subject.
I have always recorded at 96/24 because I want things as pristine as I can reasonably get during the recording process and my system has never missed a bit recording at that rate even maxed out on channels on the interface. But – it has occasionally made things more work (and a bit annoying) to mix when importing more channels and adding more processing – enough to bog down the system and make me change how I handle the mixing.
I am sometimes a bit disappointed hearing the 44.1/16 rendered version after listening to it extensively at 96/24, but that’s what is currently available for distribution except for those who are into HD Tracks or the like. (I won’t go off on compressed audio in this thread. 🙂 )
You’ve got me thinking about giving in to moving down to 48/24 just to make life easier – at least on higher track counts and – yes – many great recordings were recorded at 44.1, so that doesn’t mean a recording can’t sound great recorded at just 48…
Thanks again Graham for keeping it real and getting us thinking about the things that matter!
Thanks Walt – and kudos to you for focusing on what matters also!
There are no ‘steps’. The steps are a misconception because of the way graphs are drawn. Where there is a sample there is a value, where there is no sample there is no value. There are no values between the samples, there are no steps.
The graphs people draw are before your DAC converts, it’s an incomplete picture, and the steps are drawn in. They are not really there. A higher sample rate will give you greater frequency response AFTER reconstruction. And we only hear sound after reconstruction, there is no sound before reconstruction there are only 1’s and 0’s. Graphs, for audio linked to sample rate are best drawn as lolipop graphs not hold-over graphs where we just draw in the steps. This explains it a little better: https://paulsmithblog.wordpress.com/2014/11/27/24-bit-recording/
You might want to look up the sampling theory concept called the Nyquist Criterion. It is the foundation of sampled audio technology. It implies that in the presence of an ideal output filter (to turn the D/A signal into audio), the required sample rate is 2x the maximum frequency to be reproduced (yes, you need anti-aliasing filters,…). This means that 40K should be sufficient to capture all of the information in an audio signal. The major reason for the audible differences at higher sample rates is the performance of the output filter. Perfect filters take an infinite amount of processing power. Cheaper audio cards have poorer filter algorithms and don’t sound as good as better cards with more processing power. This is one of the reasons why some CD players cost thousands of dollars while others are cheap. Compare the same material recorded at different sample rates on multiple audio interfaces. The same 44kHz file will sound much better on an Apollo than on a cheap interface and the difference between 44.1 and higher sample rates will not be evident.
Great video – thanks! This all makes sense. Now, what if you’re a newb (me) and just realized that the last 8 sessions you did (and like) were all recorded at 48k 24bit and you want to get it back to 44.1k? (Ha!) What’s the process, what are the risks of lowering the rate / file size, etc? Alternatively, should I simply finish the process as-is in 48k and let Spotify, Apple, etc. dumb it down? If that’s the case, should I expect any sonic variation beyond the normal changes I might hear? Thanks for entertaining my rookie question! 🙂
What are your thoughts on a project that contains a mixture of 44.1 and 48k files. What would be the optimum sample rate to set your project to – 44.1 or 48k?