i think these excessive gain issues are from the conditioning that many of us received from recording on tape, where you actually benefitted from as hot a signal as you could get before clipping. The beauty of digital is; you can get away with much lower volumes and still retain the quality of a hot recording, thus leaving you plenty of headroom for signal processing....thanks again man, great stuff!
Another HUGE mistake many make. Don't keep adding treble to everything to clarify it. Remove the mud around 400-600 hz and you'll be surprised how that clears it up and does not wind up so brittle sounding in your mix. Often times what you remove is more important than what you add, not to mention, a lot of energy is in that region which leaves more overall headroom when you are gain staging. But most important of all, Listen! Removing TOO much is not good either because you will lose warmth. It is indeed a balancing act. Again, pay attention to what you are hearing.
@Aiden Macleod Right back at you dude, don't give up, keep pushing, you can get that worthless degree; in the mean time, keep believing internet myths like never boosting and such. The only douche bag here is you coming here with your nonsense and myths like if you are some kind of George Martin. Don't spread your bullshit pseudo-knowledge.
All you smart engineers technicians etc. Man said don't keep adding treble to EVERYTHING to clarify mix!!! Sometimes its beter to remove mud which is somewhere between 400 and 600 Hz. Even lower. It brings more dynamic to mix. And you have to decide what freqs to add to what and what to remove to what. If you have "boxy" vocal, you won't save it by adding trebble freqs. You have to find that "boxy " freqs to make vocal clearer. And so on for acoustic guitar and other instruments. Then you'll get mor headroom for mastering and finally get master much louder without loosing dynamics... When you stink you Won't get rid of it with perfume. You have to wash yourself and then put the perfume on
I don’t even have FL Studio but I love Michael’s content and explanation to recording which you don’t need FL to implement. Thank you for your contribution as well Michael! 👍
Who says it's not important? It's a must! You literally can't play a song through a DAW if things aren't gain staged. It will stop playback if there is to much sound pressure. Purestgain from Airwidows goes on every track, buss, folder, VCA and Master buss in every project. I rarely even use the track sliders in the DAW any more. I just put the Purestgain controls on the track and use those unless I'm lazy. The sound difference vs. a standard gain plugin is noticable after about 20 tracks.
It would be impossible to stress enough how important gain staging is. I used to make that mistake constantly. I get people to ask me to help them fix a mix who have a bunch of fuzzy black rectangles for files a ton. Their eyes invariably light up when they see how much that one tiny step did to improve the ability of a track to breathe properly.
For about 10 years I didn´t take care of gainstageing at all and always had critiques like " I hear a hiss on the vocals", " Something is causing distortion" etc.. These days, and in particular after joining Graham´s Mixing University, my mixes get a proper gainstageing and...BOOM... they sound better!!!
How could you not gain stage? Most of the stems I get will keep a DAW from playing them back for more than a few seconds before they stop playback to avoid overloading the system.
Hi Graham. Quick pointer on the the Waves SSL Channelstrip. The fader is output not the input. The switch is only for the metering. Also the input knob adds saturation to the sound (not a drastic amount but still) So I tend to use clip again for the input, the input knob if the compressor doesn't have much to grab on to, and the output fader to balance out the processing. Essentially what you said, but the fader thing I had to point out just in case.
The saturation is WHY people use an emulation of a channel strip like the SSL, though UAD's is the only emulation that sounds like the real thing. The Waves emulation just makes noise, and the Slate Digital is only slightly better. Still, they do as a touch of analog feel, but not the warm richness and truly harmonic analog distortion that the real hardware or the UAD adds. If you're not wanting the saturation, you're much better using FabFilter EQ and compression. Even if you do want the analog feel, many would argue you're better off using the FF plugins with the opto compression in Pro-Q2 and Saturn for saturation. In any case, there is no point in using an SSL emulation if you're avoiding the manufacturer"s attempt to sound like the real strip AFAICT. Especially the Waves version. We ought to all be boycotting Waves anyway, as they are an Israeli company who steals technology, often markets snake oil plugins with trendy names endorsing them just to grab a few bucks from naive young audio engineers who want a magic button, and support the racist, apartheid Israeli regime that commits more war crimes against humanity and war crimes in a week than any other country"s government's commit during the entirety of their existence. Also, their plugins rightly belong to the American tax payer, side they pay for whatever R&D they don't steal with billions in foreign aid that never gets repaid. We ought to not just be boycotting them, but freely distributing their entire library to the American tax payers to whom they belong.
Higher sample rates don't mean "smoother" audio representation! This is a common misunderstanding of digital audio which isn't always that intuitive. Sample rate dictates the highest frequency you can record. Humans can hear up to 20 kHz at best so we don't need to record frequencies higher than that. Sample rate must be at least twice as large and that's why 44.1 kHz is enough for human ears. Higher sample rates do not increase fidelity. The sampling theorem says you can reproduce the signal 100 % as long as the sampling rate is at least twice the highest frequency in your signal. DACs create "smooth" analog signal from the sample points regardless of the sample rate. Higher sample rate means redundance in this sense. 88.2 kHz sample rate is not 2 times "smoother" than 44.1 kHz, it is the same fidelity defined twice. Somebody telling you twice you won the lottery for $5 million doesn't mean you won $10 million! 44.1 kHz sampling rate already defines the sounds we humans can hear with 100 % accuracy. That's the power of digital audio and something a lot of people do not understand well because it goes against our intuition. Similarly bit depth doesn't define fidelity in the way many think. When dithering is used correctly, bit depth defines only the level of the noise floor. For every additional bit, the noise floor drops 6.0206 dB so that theoretically 24 bit allows about 48 dB lower noise floor compared to 16 bit. However, 16 bit already allows over 90 dB of dynamic range and when shaped dither is used, well over 100 dB of perceptual dynamic range. In consumer music no more than about 80 dB of dynamic range is need. This means 16 bit digital audio already gives more dynamic range than we need. In music listening 24 bit is total overkill, but very useful in music production because audio levels aren't optimized. So, 24 bits when producing music, and 16 bits when listening to it. So why does your DAC sound different with different sampling rates? The most important reason is expectation bias. You expect higher sample rate to sound better so that's how you hear it. Simple placebo effect. It's amazing how much our expectations affect how we hear things. Careful blind test removes this aspect. Differences in audio tend to disappear in blind listening tests. Another reason is that your DAC really creates different analog sound from different sample rates. The difference is real, but it doesn't mean the higher sample rate version has more fidelity. It means there is a difference, two interpretations of the signal. There are perhaps slight differencies in high frequency phase response, but that doesn't mean difference in fidelity. Your DAC giving the best sound at 192 kHz is not because 192 kHz has more fidelity than lower sample rates. It's because your DAC has the most pleasing imperfections to your ears at that sample rate, who knows maybe because of expectation bias? Summary: - digital audio allows 100 % fidelity within the frequency and dynamic range limits. - sample rate defines the highest frequency we can have in the recording. - bit depth defines the level of the noise floor (dynamic range). - 44.1 kHz sampling rate is enough for humans. Do you hear above 16 kHz? - 16 bits is more than enough in music listening. - 24 bits is beneficial in music production because levels aren't optimized. - be aware of expectation bias.
In regards to sample rate. I was always taught that you should be recording at 96kHz (or better if you choose to do so). The reasoning behind this is while working with outboard equipment (consoles, comps, eqs, etc.) you are not recieving a max frequency of 20khz. You are recieving whatever frequency that specific gear is outputting which is not a perfect sine wave. Just because you can't hear it does not mean it is not there. Recording at a higher sample rate creates redundancy (as I believe you stated, on mobile. Difficult to check lol) to help ensure you don't have any aliasing. For anyone who reads this and doesn't know, aliasing is when a recorded sound has a higher frequency than the sample rate can properly capture (divide the chosen sample rate by 2. This gives the highest possible frequency you can record with little to no problems). The effects are frequencies that playback incorrectly like a constant hum that is not there and not caused by your equipment rather caused because there happened to be a frequency higher then the sample rate can handle. IMO if you exclusively operate within your DAW I do find it a bit useless to record above 44.1k or 48 since you won't be using any outboard gear, or very minimal if you do. Aliasing is also another reason why many monitors say they can playback sound at much higher frequencies then you can hear.
Some plugins don't oversample. And that can be obvious. Not just very subtle, but very audible. So you need to test them on your low sampling rate and your high sample rate to make sure there isn't any difference.
"The sampling theorem says you can reproduce the signal 100 %...." -- Nyquist was speaking in the mathematical realm. At some point we must deal with a host of problems in different disciplines when entering the realm of physical implementation.
@Joseph Winett That dude uses strawman arguments and sine waves in his examples. Sine waves don't have harmonics, audio is more complicated than his tests.
Radio Sound is the best source for human hearing tho , try to listen ( flac ) has higher mb compared to mp3 ( smaller file ) has less quality sound compared to wav ( so obvious)
When it matters, many good plug-ins offer an option called 'oversampling' which effectively lets you work in a 44k session, but flip a switch when it's time to export/bounce/render the audio, which tells it to basically calculate at 96k under the hood when oversampling is on. This rarely matters unless you have a specific problem that you are trying to solve. For what it is worth, to my ears the best sounding guitar effects processor I have ever owned (Digidesign Eleven Rack) sounds like a buzzsaw at any setting other than 96k!
"The higher the sample rate the smoother the signal will be" This is a bit of a misconception! Based on the Nyquist-Shannon Theorem, when your Digital-Analog Converter (DAC) 'interpolates' or connects-the-dots between the samples being fed to it, it does so in a way that would be equivalent to the continuous, real-world electrical signal that existed below half the sample rate (bar aliasing, which is combated with the reconstruction filter being set a bit below half SR). The content below 22050 hz in a 44.1khz sample will be *identical* to the content below 22050 hz in a 192khz sample. One might go as high as 48khz to give the anti-aliasing/interpolation filter room to breathe and appease those who claim to have super-human hearing extending above 20khz, but going beyond that wastes processing power & storage space for no additional benefit to you or the listener. It can even be detrimental when accounting for factors like jitter. As you said, super high sample rates are unnecessary!
aceyage oversampling is useful for some specific DSP mathematics, especially frequency domain stuff, but it’s not going to add to the fidelity of your signal below the nyquist frequency of your overall project. It’s not particularly relevant to the issue at hand.
Tristana Wild There is some really audible aliasing wiith some, usually older, plugins or those that don't oversample. That’s where higher sample rate can help as those mirroring frequencies are now out of the hearing range. I won’t use plugins that don’t sound exactly the same at 48k or 96k. Working at 96k is just unnecessary CPU wastiing otherwise, agreed.
went back to my first real practice mix (I'm a noob, learning by fucking up to be honest!) and went through EVERYTHING to gain stage properly. one track, one plugin at a time... what a world of difference, really brought out a lot more clarity from everything. I was able to knock a lot of the different points of compression down a bit because of it, and really opened up the sound of the song. Thanks for this!
Ohhhh.. so when you time stretch should you resample at a higher rate 1st?..I understand that it's ideal to run the project at 96k but if I'm just making a track or scoring for a video. How would I handle that situation.
@@rendilly No. You acquire at a higher sample rate. When you time stretch or repitch, you're starting with audio with a maximum frequency around 24kHz. Slow down by half and you're now at 12kHz. If you acquired at 96kHz, then slowing down by half still leaves you with audio at 24kHz and plenty of high end. If you're acquiring at higher sample rates, you need a quality clock source as the timing of your converters becomes more crucial.
Just a correction regarding the sample rate. It is not like higher resolution in a photograph. That's a big misconception. Whatever sample rate you are using to record, it is always 2 samples per frequency. So, 96k doesn't provide more accurate sampling, but rather expands the frequency sampling range to 48KHz, which is way beyond human hearing. There are other things that go into play when recording at higher sample rates, but that's another discussion. I agree with you regarding the benefit/ drawback of higher sample rates. And you should always mix in 32bit floating point environment at least, never 16bit.
Aside from the arguable benefits that you mentioned of higher sample rates (not opening that can of worms here, but for what it's worth I track @96/32float and have 60/70 tracks on a 4 core i7 imac and I haven't maxed it out yet.) I do believe many synth plugins actually sound noticably better at higher sample rates, and I've also heard some performance related issues reported because sometimes the synth plugs will be processed internally @96k, with the DAW @41/48k. While I imagine quality and updated plugins will make this a less frequent issue, it is something to consider.
@@vooshmoozik6185 I'll see if I can find a link to the performance issue in that regard but here's one angle of the conversation. www.keyboardmag.com/lessons/should-you-record-at-96khz
I agree with everything regarding the Sample Rate with keeping in mind that most people are doing their stuff in a homestudio environment or just working at a small professional level. If I'd go in a bigger studio with a major project I'd want at least the 96khz Sample Rate. There is a (small) audiophile market that wants to be served, too. Personally I work with 48 khz and I'm totally fine with that. I'm even considering getting back to 44.1 khz just for the performance gain.
i haven't made music in years...finally have enough to buy a new imac...I've spent the last 6-12 months watching videos like this to figure out what I was doing wrong years before, so when I finally can start, I'll have more knowledge right away.. Thanks for making this video....
One thing I think was missing from point #3 (recording/mixing at higher sample rates) is when recording at 44.1kHz the anti-aliasing filter (low-pass filter) that prevents frequencies above 22.05kHz from being recorded can also roll-off some of the high frequencies in the range that you want to record (below 20kHz). This is essentially rolling off high frequencies with a high-shelf EQ on every track you record. By recording at higher sample rates, the anti-aliasing filter can be designed better because it needs to reach 0dB before 48kHz (if recording at 96kHz sample rate) instead of 22.05kHz, so it has plenty of time to reach 0dB without affecting frequencies that you care about under 20kHz. It's explained better here: www.indiana.edu/~emusic/etext/digital_audio/chapter5_rate.shtml That is the strongest argument I've heard for recording at higher sample rates. Maybe recording at higher sample rates to eliminate anti-aliasing filter problems and then batch converting to 44.1kHz for mixing to save CPU power would be the best of both worlds?
I'm a beginner, but having this knowledge up front will help me to avoid mistakes going forward. Thanks for putting out these videos, subbed. I call myself a beginner because even though I started many years ago with Sonar 8.5 studio, then upgraded to Sonar X3 studio, I never put in the time and effort to really learn the software. I'd bring it up and basically fly in the dark, just figuring out the very basics of laying down a few midi tracks and vocals with some reverb. I recently downloaded Cakewalk by Bandlab and now I'm making a concerted effort to learn it. I had given it up for years because obviously if I didn't know what I was doing, the creative process would drag to a crawl. Instead of making music, I was struggling with the software to the point where it wasn't fun anymore. It's a ton to learn but I'm being patient and not trying to learn it all overnight. But every day I try to learn something new and build my knowledge base.
This video is pretty spot on based on many home recording projects I've heard as well. Understanding dynamic range and gain structure are basics anyone getting into recording should understand if they want their recordings to sound good. I went from analog to digital when DAW's were first being offered around 1990. I had Turtle Beach 56K and Cakewalk by Twelve Tone. There was no forgiving if you tried to record too hot of a digital signal; it just overloaded and broke apart. Most don't seem to have a grasp of EQ or the fact they compress the hell out of the signal as well these days. Most music being recorded not just in home studios but commercial ones lack range and nuance and most performances are mediocre relying on studio magic. Don't get me going!! Good video for beginners though.
I argue about all 3 of these all the time. Some people just think louder is better and higher numbers are cooler. It is what it is. Great video as always!!
Not sure if anyone's mentioned this yet, but you can also get your level by CMD+clicking in the volume section of the track to toggle through to see peak level (you'll toggle past delay compensation value to see this), which some might find useful. Cheers.
When recording high frequency instruments like violins and cymbals, you might still want to consider a higher sample rate, so there is less risk of aliasing, i.e., the ultrasonic frequencies traveling back down into the audible range as inharmonic distortion. Especially true (and clearly audible) when doing any pitching up of hi frequency audio. You want to do that in a separate session running at a high sample rate, then low pass your sound somewhere around 20k, and then import it back into your session at 44k. The difference can be massive.
good tips. I do my own gain staging from the sound design, I always stay in the green on the mixdown and give 8 db headroom for the master, and I do use 48k/24 bit when exporting.
Lots of good information here. However, I like to mix at 48k because my focus is on sync licensing, which requires 48k 24 bit mixes. 44.1k is good for commercial releases. CDs are only compatible with 44.1k 16 bit, but every music library and sync licensing opportunity that I have seen requires 48k 24 bit. That said, I use 32 bit floating point for recording, mixing and mastering and dither my masters down to 24 bit for sync licensing and resample and dither to 44.1k 16 bit for mainstream releases. Other than CPU conservation, is there an advantage to recording at 24 bit instead of 32 bit FP?
This is the first time I've posted here, but I've been watching your videos for several years now. This is one of the top sources for very concise instruction for mixers, whether they are experienced or just learning the skill. I consider myself closer to the latter. I would like to address the last issue, Sample rate. In my limited experience (I mean that sincerely), I have the understanding that sample rate is like pixels in digital photograph. So the more 'pictures' that you take of one second of music, the more accurate the digital representation of the analogue source. For mixing, 44.1k is plenty and you will probably never hear (some say that they can) a difference. But... One of the issues that comes up in my little recording world is dealing with some singers who don't sing accurately; so pitch correction is needed. This, I believe, is where 'more samples' benefit. If you have a higher sample rate, you can more accurately correct pitches. This could also apply to alignment. And a big one is creating harmonies from a melody vocal. I do this a lot. I use Cubase (newest full version), which has an amazing pitch correction feature. But for building harmonies, I always use Melodyne Studio. I find when using higher sample rates, there are less artifacts; there's almost never none; but for harmony, they aren't noticeable. So, I'm of the mind that one of the purposes for higher sample rates are for things like this. The actual mixing doesn't require the higher sample rate, but the editing beforehand does. So recording in a higher sample rate is, for me, to allow for more accurate editing. We don't all have singers that can do the job, especially if they are going to home studios owned by newbies like me. So everything is recorded, edited, mixed and mastered in the high sample rate and only after mastering is it re-sampled that dithered. Right now, I have been recording at 88.2k and have been for over ten years, but my reason had more to do with the computer's math co-processors only having to divide by two to lessen the load on the cpu. I don't know anything about this stuff, but it sounded logical. But those days are long gone; computer power and OS,s are a none issue now, at least for audio recording. We now have super fast computers and inexpensive large storage. I'm going to move up to 96k in the new year if only to be able to cut it in half for video tracks, which are, as mentioned in the comments, the sample rate for video. All of this is just my accumulation of knowledge put into my recording practices. If I have misunderstood any of this, please let me know and I look forward to hearing about it. So Merry Christmas to you and yours, since it's been a year since you posted this and it's that time again. Holiday wishes to all.
@recordingrevolution, you have a very good point at 10:00 and I go beyond that: the same applies for bitrate! If you let your whole project in 16 bits, the plugins still providing 24, 32 and even 64 bits in their outputs. I agree that in mixing phase you will be listening at 16bits, but when you generate a mixdown file, you probably will have the option to set to 96KHz and 24bits for mastering! (at least, Sonar has) That way you still having as 24bits "like" files. That's why I still recording in 44.1KHz/16bits (as you said, the "wins" are too low), but I can count on "upsample" and "upbit" from plugins.
I know nothing about Studio One. I was using a crack of Cubase 5 when I started and didn't understand anything. The first thing I did in Reaper was to send my vocals to another track for an EQ'ed Hall....WITHOUT READING ANY INSTRUCTIONS........I love Reaper
@@RichieHmusic It is intuitive. The other thing I love about Reaper is value for money. The Reaper folks haven't bought into that subscription model, and have decided to ask a fair price. It will pay off for them in the long run, big style.
@@RobertLinthicum totally agree. I refuse to use anything with a subscription model or iLok on principle. People who say REAPER isn't intuitive escape me. It's easier than anything to start with, and every single thing (except getting rid of the big clock) is customizable. Plus, with the ability to create custom actions, or macros, you can cut your editing and mixing time by 75%. People who say REAPER is to hard are the kind of people who say reading is to hard because they don't want to learn the alphabet. Very little that commenter says makes sense. Studio One is a nightmare compared to REAPER. Even Logic is agonizing after you get used to REAPER. There's no other DAW that will let you set it up so that everything you do is a key command or mouse click away, and everything you don't need is out of your way.
You should be checking your track levels before you start mixing, when adjusting your gain stage make sure the track fader is set to >0< then adjust for gains to -18 or -12 whatever,,drums, bass , lead vocals on the louder side.
About #3 of no high sample rate because it kills your CPU? Here are a little two somethings that everyone seems to either ignore or forget: 1) It's only a MISTAKE when you are recording at high sample rates in your DAW when your actual input does not SUPPORT anything higher. If your audio interface doesn't go above 48, then there is no point in setting your DAW higher either. 2) You can very well use high sample rates without killing your CPU. You record the raw signal in high sample rate, then COPY the original file, CONVERT to a lower sample rate, open different project with the lower sample rate, do all your edits. When you are done with everything, you duplicate the project then set to high sample rates, replace the source files with your high sample rate files, because all edits done have been done from that one specifific original file. Bounce out, phase-shift, compare to the lower sample rate export. Do minor adjust. Done! You just have futureproofed tracks without straining your processor. You're welcome :) There are some benefits to high sample rate recordings other than just higher sound fidelity though but this has to be judged by everyone individually. If you have to do a lot of processing and time stretching, then you go as high as you possibly can. Fewest tracks require that though. But for people doing sound design, that's a very different ballgame. Edit: btw those large, clunky files people complain about? Fewest people CLEAN UP their project to only have the files remaining that are actually necessary, alongside a copy of the main source files. They keep every scrapped piece in there. Please clean your bin, fellow recording engineers. Also: Those files can be COMPRESSED for archiving :) Just saying. Multiple backups notwithstanding.
Makes sense to leave headroom for mastering. I'm going to try that. I always went for maximum volume before mastering. Although mastering didn't make it clip, it did limit how much mastering I could add to it.
From a technical standpoint, I don't have anything to offer to this discussion but, from a usability perspective, I think that a lot of "mistakes" can be circumvented. For example, beginning users in ProTools or Cubase will start their first projects based off the templates. The initial focus when starting out is getting a clean recording from their sound interfaces and making sure it's not clipping on the meter on the hardware device. Eventually, with more practice, it becomes fluid practice to observe hardware meter and track meter and master fader meter. It's a balancing act for the eyes initially. Most tutorials state to adjust volume at the track level and leave the master fader alone during recording and mixing. Things can get confusing for those composing their songs in sequencers. I remember seeing one tutorial in a hard copy guide to record your instruments as high as you can without clipping (sequencer recording on some midi capable keyboards). There's definitely a lot to learn that it can become overwhelming for first time home users. Which is why, after watching your video, I had a brainstorm that, in addition to project templates that DAWs provide, they should include a finished project that shows it's evolution from conception (all tracks recorded), mixing stage, mastering stage, and what it might look like when targeted for various venues, i.e. streaming, CD burning, radio broadcast. This way, beginning home recording enthusiasts have a standard of measurement to use when developing their projects. The analogy I have for this is sculpting museum figures. These people had the benefit of seeing what a finished sculpted figure looks like as its displayed in a museum. So they have a working visual standard of a product worthy of inclusion in a museum. Home recording people would benefit from the same. There's nothing in the purchased product that shows and/or adequately describes what a project should look like for bouncing to disk for the various venues.
Hey great tips, especially gain staging. But I must say that while it's hard to hear the difference between 44.1kHz and 96kHz sample rates, the processing that's done in your DAW definitely benefits from the highest sample rate you can manage. Depending on what your 're doing with your plugins, high sample rate can sound noticeably better. For instance if you are doing any kind of time stretching the higher the sample rate the better. In general, higher sample rates result in cleaner processing, because there is more data per second. It has to do with the math involved, but suffice it to say that the higher the resolution of your audio, the better it will sound after processing. Also when you down convert to 44.1kHz it sounds better the higher it is before conversion. Again it's the math your computer is crunching that makes the difference. Then again, you still may not actually hear the difference, but someone will be able to. Great tips for beginners though Graham.
@Beeblebrox One Sometimes, "digital" can be a source of misunderstanding, even for pros. you might find that video interesting: th-cam.com/video/cIQ9IXSUzuM/w-d-xo.html Please watch it to the end. This is the best explanation worldwide imho, and I say that as a person who has a degree in audio engineering. I love that guy. Highly recommended!
@Beeblebrox One By the way: This is for the readers- dithering is not used for down-sampling, it is used for the process of bit reduction. (24 down to 16)
@Beeblebrox One Oh really? Almost every "top" engineer I've read says use higher sample rates, for the obvious reasons. Dithering does not cause problems as far as I know. In fact, when dithering adds that small amount of noise to the signal, it actually improves audio resolution, especially in quiet passages. And, as Johannes Mazur pointed out, dithering is used when reducing the bit rate, not the sampling rate. A little bit of knowledge can be dangerous. You can mix and record at low sample rates if you want to, but I'll continue to record and mix at high sample rates, thank you very much.
The issue of high sample rates is actually quite nuanced. With audio signals entering through an interface, you likely won't hear any difference by recording at 44.1 kHz or 96 kHz. However, with sounds generated inside the box, it can make a big difference because, for example, an instrument recorded at 96 kHz may not have aliasing that would occur if it was recorded at 44.1 kHz. This is not a "wine tasting" difference, but something that sounds obvious even to someone who's mixed FOH all their life :). So...instruments that oversample internally should solve this, right? Well...mostly, but the problem there is an algorithm optimized for real-time oversampling may cut corners compared to offline, non-real-time upsampling. In most cases I record at 44.1 kHz, but for virtual instruments, save the MIDI file, open it up in a 96 or 192 kHz project, render at the higher sample rate (which means there will be no aliasing in the audio range), then downsample back to 44.1 kHz. Because I'm downsampling audio that has no aliasing, there won't be any at 44.1 kHz whereas the virtual instrument output would contain aliasing. This is a very basic example, matters are much more complex with plug-ins like amp sims, some dynamics processors, etc. The bottom line, unfortunately, is that there is no "one size fits all" solution. The people who say "recording at 96 kHz is ridiculous" and those who say "recording at 96 kHz sounds so much better" can both be right, depending on the project, the software being used, the plug-ins, when the plug-ins were coded, who coded them, etc. etc. etc.
I’ve been following you for little while, but I just want to say I really love your videos. You go into so much depth without getting off topic and rambling. Thanks for all the info you share
Great gain-staging points btw. Anyone using Ableton Live, if you want to Gain-Stage, you can use the "Utility" plugin. Don't use the "Limiter", it colors the sound a tiny bit. But the Utility is equivalent to lowering clip level, 100% transparent.
What a difference!!! I'm used to live mixing on a big (Allen & Heath GL2800 56 channel) analog board. Their manual has a section on gain staging. They want every channel, every subgroup, all the mains... up in the yellow with peaks at +3 db. That, they say, gives you the best signal-to-noise ratio possible. For live applications, that holds water. Apparently that is NOT the case for digital recording. Guess I better get to learning if I want to get into using ANY computer software, huh??? Already clicked on the link for your FREE Radio Ready Guide.
This is GREAT!! Recently got back into music/production, these are things I didn't know that save time instead of a lot of trial and error. I was working to hard on the end process trying to get the "right sound", when I could have corrected it at the beginning. Good to know. I knew nothing about gain staging before today!!
Thank you Graham! Just applied those basic rules to my latest project and it sounded infinitely better! Never really took the time to understand it all before. Great video, clear and straight to the point.
Speaking as a softsynth/DSP developer. Common misconception: sample rate is NOT a stepped resolution. Most audio engineers think that a digital signal is defined as tiny blocks with each having an associated amplitude. Some imagine it as a staircase. That's simply not true. What sample rate means is that at a specific interval (44100 times per second for standard cd's) an amplitude is defined. They are points in time, not blocks. When this data is converted by the DAC to an analog signal, a smooth line is 'drawn' between these points. If your digital signal represents a 20 kHz sine, that sine is perfectly reconstructed by the DAC. Or any sinusoid below the Nyquist-frequency. If you run this DAC-converted, digital signal through an oscilloscope, you won't see any 'staircasing'. Your sinusoid will be perfectly replicated. That's why having audio at higher than 44.1 kHz sample rates is ridiculous. We can't hear beyond 20 kHz. And if you claim you can, congratulations; you are a dog. Think of it like this; is having a screen that emits a spectrum of light that extends beyond our visual spectrum a good idea? No. We can't see ultraviolet or infrared light, so there is no point in adding this extra data in a visual signal. It would even be harmful in the case of uv-light. Then why would we need an audio signal that contains data that's imperceptible to our ears? We don't. It's pointless.
On an actual console you start with the channel faders at 0db then you adjust the pregain so it isn't clipping while EQing the sound and then once you get it set, pull the channel fader back and do next channel, once all set then start mixing... That way you know once you hit unity gain 0db on the channel fader anything above that will start clipping.
Thanks for this one. Gain staging is not a step I was aware of until now. I did a quick hunt on how to do it in my DAW and think this will make a massive difference. Much appreciated
1. You should do the gain stage with a vumeter aligned to -18 rms, thats the optimal signal/noise level 2. Each track is different, the drums transients are different in each song, so u can't focus on peak level, there's an old system created by bob katz, the K level, for comercial music u should set the master to 0 vu calibrated to k-12, that's like -12 rms aprox. Not only for avoid clipping, if you always mix at the same perceived volumen and yours monitors/headphones knob volumen is always in the same position your brain will remember how sound everything and you will take better decisions. Also with this system you always mix on the same point of isophonic curves (Fletcher-Munson) 3. I mix at 44.1khz because my cpu can't run all the plug-ins at 88.2khz, but if I u can, mixing at 88.2khz its the optimal between quality/AD-DA errors and its the exact double of 44.1khz so the downsample is a very simple mathematic operation for the SRC (just divide by 2), also at 88.2 kHz all the plug-ins goes automatically oversampled (even plugins without oversampling option) so avoid aliasing and all the plugins works x2 quality.
I agree with the first two mistakes. The last one, well... I think the best way to explain it is this: If you only plan on using a picture for Instagram or a Facebook profile picture, or even an album cover... Will you take it in a 1mp camera, or will you try and get the best possible image quality? Exactly the same goes for audio. Just because it will end up in a 44/16 mp3 doesn't mean you don't have to record the best possible way.
I have to agree with you on the high sample rates limiting how many tracks you can work with without pushing the CPU. I capture my individual audio tracks at 48 khz /24 bit and couldn't be happier with the quality I get. Rather than doing a two track mixdown on the multi-track station I pass this signal via analog to a stereo tube preamp where I can slightly saturate the tube preamp for a warmer feeling tone that is then recaptured at the same 48 khz/24 bit sample rate on the mastering station. Once the 2-Track stereo audio file has been mastered, I then reconvert down to the standard 44 khz/16 bit sample rate during the mastering process.
Hey Graham, I guess you won't read that, but there is something that makes mixing at 96k not that pointless at all. Due to Nyquist, the ordinary stock EQ gets an asynchronous bell shape at high frequencies. This is what they (umm..we) call non-analog behaviour, as it is a digital EQ side effect. The most popular developer-work-around is a switchable built-in oversampling but that introduces latency and eats some CPU-Power. When you mix with higher sampling rates you don't need oversampling. The curve of the high EQ-band behaves like its analog counterpart and sounds more natural as the Nyquist frequency slides up. That is the true benefit of mixing with higher sampling rates. You don't need to switch on internal oversampling (where it is available within a plugin) and therefore actually safe DSP-power.
Great advice/pointers fellow human ! I appreciated especially your "controversial" thought about sample rates ... some of the best records in history were made on analogue machines that dont come close to as intricate and multifaceted as we have with our DAWS and amazing array of 21st century recording tools/options ... like good coaches in Basketball keep saying 'keep it simple' ... simplicity executed with great merit always goes further than gimmicks and shortcuts/complications.
I like mixing at 96 kHz because I feel like it definitely makes it easier to hear your eq moves, especially when it's subtle, or a narrow cut. I can really hear the frequencies better. Really depends on your hardware though. I can only manage a full mix at 96 kHz on my Mac Pro, whereas my MacBook Pro can only manage a full mix at 48 kHz. Tracking at 96 kHz has lower latency too which helps with amp sims, but it's very taxing on the cpu. Best advice, try a few different sample rates out and see what your computer can reasonably handle and just go with that.
There is a noticable difference with the initial jump from 44.1 to 48 which is easy to hear. Beyond that there are minor differences. Many people mix with 44.1 and it sounds absolutely fine. It's worth noting, a mastering engineer once said to me, if you want a high resolution, but still need to convert down to 44.1, then 88khz is the way to go because it mathematically divides neatly. It makes sense. Oddly enough he said 96khz is not always the best for bouncing down to 44.1. in terms of rounding up or down with the truncated slices and peaks.
I though this myth was laid to rest years ago but it keeps popping up! It's only easier to downsample from double the frequency if your downsampling method is just averaging adjacent samples. Maybe back in the 90s some sketchy companies did that in their hardware but nobody does it like that now because the maths to convert between common sample rates is both simple and lossless. It makes no difference at all if you throw away 60% of your original signal rather than 50%, none at all.
@Beeblebrox One Dithering affects the bit conversion (e.g. from 24 to 16) and it has to be done just ONCE (preferably you leave it for the master engineer) and not sample rates.
Hey Graham! I just want to point out 2 things for adding to the debate. One is that Clipping isn't a bad thing, we use it so much and because of the keep-repeating that it's bad, people take years to understand that it is actually a very important tool, specially in aggressive music. For the beginners: please take this one wisely, it's not clipping in whatever way, there's almost infinite possibilities here and you must know how and why you're doing it. And the other one is that there's some other aspects with that last topic (as you said, controversial), I read Al Schmitt, Geoff Emerick and some tech guys talking about Nyquist filters and suppression of subharmonics of those upper frequencies. Everyone with horse power is working at 98khz and up, but as you mentioned CPU must run alongside the intent. Thank you and keep the good work!
You are referring to analog clipping, right? Digital is the opposite. Plus you can always add more volume and punch using limiters, EQ, and compression on the master, and in the digital world, it's best with more headroom to add frequencies/volume to your final product.
actually no. any clipping is tool. all tool can do more than it's supposed to do and very often they do. however, some basic understanding is lacking here. clipping is a different thing that also can happen inside compressors and eqs (there's a famous trick of clipping API eqs). clipping is about reaping peak information hoping to have some pleasing distortion back and this effect have very little to do with the exact common understanding of compression and eq. let's get more visual: guitar amps, specially obvious in high gain preamps. so you can clip whatever you want (tubes, transformers, ICs, plugins, etc.) as long as the sound you're hearing is what you're aiming. just a warning for kids: clipping can or cannot hurt your gear, sometimes it is worth.
great advise! I've been doing this for a very long time but your simple advice still ring as a gold standard. It gave me a confirmation of what I suspected all these time. Keep doing the good things!
As I understand it mixing at 48k mean you don't have to up sample for video and for sync licensing. It's just a tiny bit more CPU power and fits more formats for licensing. IIRC.
It's useful to hear what you say about the sample rate when mixing, for just getting on with arranging and mixing, as well as clarifying points on track levels. Nice one.
I can't thank you enough for your excellent information I have in fact made all of these mistakes and wonder why all my mixes sound terrible thank you so much keep up the good work you're awesome
I'm not sure why, but when I was mixing in 44.1 I was getting a weird artifact on one my kick samples. When I went to 48k the noise went away. So the only reason I went to 48k was because I thought I was avoiding artifacts. But now when I think about it, if the standard and industry mostly use 44.1, I might as well go back down and try to use another sample or try and treat the artifact. Great advice!
Mixing Mistakes I apologize in advance if this has already been addressed... I am not sure if this is common knowledge but your DAW's, Ableton does, audio strips can be used to enhance, colour, your recordings. Similar to an analog desk! I discovered it over time. Then another musician heard it too! Very cool, if it is/was intended. I have heard of some mixers that use the analog desk as a compressor... Max out the gain for each audio track then add the audio. Then you can add distortion, too... Let me know it you try it out? And then other mixers rarely do anything as the Mastering process will dictate the final sound, anyways!! This leads me to contest the dB level reliance on -20 dBs. A little saturation is very nice... In addition, I am willing to bet that all the stages are like the above - groups, master, returns... Best of Luck all! @mrvllous4 Marquis de J.
Thanks for this, particularly the first two. I deal with audio from multiple sources often and struggle with levels all the time. So, hearing you state the obvious gave me some guidelines. Cheers.
Great video, thanks for this information! It's funny how the same information, provided at just the right time, makes all the difference to how you understand a task.
Two point here: 1. For my workflow 48 KHz is a must. I do not know if the audio will be used in a CD, DVD, BR or on the web. Broadcast standard is 48 KHz. 2. I do not necessarily agree with the original statement that high bit rate is bad/taxing for the plug-ins. A lot of plug-ins have x2, x4 or even x8 oversampling (higher bitrates allows DSP to be more precise). Meaning that the plug-in before it processes the incoming sound, up-samples the data X times the incoming data rate. That is taxing the CPU. Imagine 100 instances of that plug-in. On the other hand if you already have high enough data rate, that process of up-sampling and down-sampling on each plug-in’s in and out does not have to happen (saving a lot of CPU processing power). If I understand it correctly, modern computers use DMA to move the data and CPU to process it. The CPU issues a command to a DMA chip to move a chunk of data from point A to point B, while the CPU is free to do other things. So the amount of data being moved is not that significant to CPU performance.
I want to point out that the anti aliasing filter probably causes the most trouble on low end gear. With a 44.1 sample rate the filter is set around 22.05 kHz and eats into the the audible range. Using 48 kHz and higher sample rate could help that shortcoming by having the filter set higher while recording. Your mileage may very depending on your gear. You could use REW to do a frequency sweep on your ADC.
I make the first mistake sometimes. I think that Trimming some of my hotter recorded signals would be beneficial. I should use the output knob on my avalon more because there is a very clear sweet spot on the gain. Thanks for this clip!
Hi dude and thanks for taking to time to educate the wonderful people of TH-cam. I have some thoughts on the content and the message though. 1) The plugin sweet spot thing is a myth. What your talking about has no basis in DSP. The math for samples -0.1 dBFS is the same for -18dBFS. The software does not do louder, it increases or decreases a range of bits.Hence also the 'drive' knob on many modelled plugins. What you can do is overload the plugin and run out of bits or reduce the bit-rate artificially and create digital distortion. Which brings me to the next one. 2) You cannot run a digital signal 'hot'. Any signal over 0 dBFS will clip and distort, any thing under will not. Say you run a 24-bit session and have a cheaper version of pro tools, your mixer is still native fixed 48-bit. This is why you can send an overloaded signal (hence why it's yellow and not red at the top of the meter) out of channel to a buss and perform compression and not have it clip or distort coming out your master channel. It will come out as mangled as you want by your cool and calculating compressor, but your mixer won't care since you basically have 288 dBFS dynamic range. Getting crazy hear and translating this to human perceivable dBSPL, hearing a signal that loud would probably burst some parts of you and would be akin to hearing the voice of God (strictly a Dogma reference). but back to math. akmedia.digidesign.com/support/docs/48_Bit_Mixer_26688.pdf 3) Recording a higher bit rate is very beneficial for any digital signal. Look into Oversampling and why you need it and Digital Aliasing. Shit, here, I'll make it easy for you: theproaudiofiles.com/digital-audio-aliasing/. Look at is this way. You're making more snapshots of audio per second. Your right with your statement about it sounding good, and that should be your main concern- but it is not a mistake to record at higher sample rates. Say you want to pitch your voice down 12 semitones or more extreme to double or triple that. Your voice will lose all it's upper harmonics. Because of the Nyquist frequency of a higher SR you can now have more! Same thing if you wanna stretch audio. More samples to stretch and therefore a higher fidelity for a longer time with less inter sample distortion. If someone wants to take a crack at what Ive said here, please do. Beware though that a copy-pasta from manufacturers stating they modeled "precisely after the original hardware, so it is preferable to provide input levels similar to those expected by an analogue master bus compressor", will find the word PREFERABLE in there useful food for thought and if not can expect a reply containing the words "Marketing", "Signal to Noise Ratio" and "Impulse Response".
I get your point. I worked for 25 years as an audio engineer and producer from the late 70's and saw the end of analog and the beginning of digital recording. Recording analog 'hot' signals use to mean banging the gain to saturate the master Ampex 456 multi-track tape to reduce the SNR and splat the signal onto tape as a pseudo compression effect. Worked a treat. I soon learned this was not possible working in the digital domain. Regarding recording at a high bit rate, I think the general main point being made is valid. Notwithstanding your point regarding high oversampling rates, digital aliasing and examples given the benefits are marginal compared with hammering the CPU. But as always - people have different experiences and reasons for engineering signals to achieve their aims. The key thing to remember, the whole thing is a about the music which is greater that the sum of its parts in terms of sound engineering and production.
Happy Holidays, and as always so much great information in your tutorials, and it's those little things. Never too late to develop better skills, Thanks Graham! Never stop making the world better through music! Stay Groovy1
Gain staging is so important. I normally mix at 48k, if my session arrives at 44k I don't touch it, there's no reason to. lots of great info thank you.
It's so sad to see people making such poor choices because of bad information being shared by well-meaning people who really don't know better. There is a very good reason why there are oversampling options in the really top notch plugins like FabFilter L2, which goes up to 32x oversampling. It's bad enough that you're starting off with low quality, but to compound it by processing and mixing it in 44.1kHz is truly a travesty.
Thank you so much for this incredibly helpful video! I've been playing music for many years and have recorded mic'd with all analog gear in the past but I'm really new to recording through a DAW and was struggling with some of the issues you've addressed here. So thank you very much!
Thanks for making knowledgeable the difference between (in gain) and (out fader) I got this knob on my DAW never figured out the purpose of this feature unless your video make it clear in one click. Thanks. Now new frontier is open..yeah just simple as that. I struggled with that problem and never had a clue to solve it.
The first two tips I absolutely agree with. The last one? Not so much. At minimum record at 48k. The file sizes aren't they much larger than 44.1 and you get some higher quality out of the recording. Most people recording at 96k (which many actually do it at 88,200 because the conversion is halved) are running HD Systems with DSP chipsets that are able to handle the processing power. All in all great video.
please put this in the description ,, and do the same for future videos. i really like ur vids but sometimes i dont have the time to listen to the whole video. 1- Gain staging 2- mix peak -5 DB 3- work on 44.1K
Totally agree. I also agree with it being pointless to mix at 96k, but 48 is a good compromise for me between higher quality and relatively low CPU drain
RE: Gain Staging: 1) Some plugins are more responsive to levels & level fluctuations than others. I.E. some plugins are coded to model analog gear in more depth or detail than others; as such, those plugins will be much more responsive to digital levels than others. 2) The way internal calculation resolution (word length) is handled in a given plugin is also a factor in how it deals with various levels. Plugins that are not coded very well will generally benefit from appropriate average levels more than well coded plugins (a poor implementation of an EQ plugins that has an internal calculation resolution of 32 bit floating point will do more damage to a signal than a well implemented EQ with a 64 bit float internal resolution). There are some very good designs that use 32 bit float & plugin coding continues to improve, so this is less of a worry than point 1. 3) Generally speaking, dynamics, saturation & other non-linear plugins will benefit the most from good gain staging; also many dynamics plugins require a certain range of average levels to be effective. The best way to avoid gain problems is to record with reasonable levels through quality converters, always at 24 bit (or more if digital FX are being printed on the input).
I was going to edit the above post, but since it has been up for 11 months, it seemed better to add an addendum: 1) Some plugins are more responsive to levels & level fluctuations than others. I.E. some plugins are coded to model analog gear in more depth or detail than others; as such, those plugins will be much more responsive to CHANGES IN LEVEL than others. E.G. a well modeled emulation of hardware that includes transformer saturation (and/or tube / transistor stages, photocells, etc.) will sound best with an input signal that matches its operating range. Many UAD plugins have well defined operating ranges; they typically sound best when the incoming signal has an RMS average at or near the stated operating range. P.S. Sampling rate is a complex subject; no single blanket statement can cover all the pitfalls and benefits of recording and/or mixing at a given sample rate. That said, here are a few typical considerations: 1) A given A/D converter may or may not sound better at a higher sample rate depending on several factors, the primary ones being the stability of the internal clock (crystal) at a given sample rate; the 'natural' sample rate of the internal crystal in non-VCXO designs (a converter designed to operate at a specific sample rate will sound best at that sample rate, or integer divisions of that sample rate, i.e. 96 kHz, 48 Khz. Some recent converters have two crystals - one for multiples of 44.1 kHZ and another for multiples of 48 kHz - the newer UA Apollo converters have two crystals AFAIK); or, if the converter is a VCXO (voltage controlled crystal oscillator), the implementation of the clocking system and/or the SRC (Sample Rate Converter); the quality of the nyquist filter can have a profound effect on the quality of the conversion (a high quality nyquist filter at 44.1 kHz is more expensive to implement than one at a higher sample rate - which is why most modern converters are 'oversampling' converters). 2) Some plugins handle lower sample rates better than others. Two techniques for implementing a high quality plugin algorithm at 44,1 kHz is: a) internal oversampling (the signal is upsampled at input, then downsampled at output; in the case of 'linear' processing like an EQ, this is one way to avoid unusual shapes near the nyquist frequency; in the case of 'non-linear' processing - a plugin that produces overtones - this allows for better high frequency overtone production and subsequent filtering of new frequencies above the nyquist). b) algorithms specifically designed to minimize problems associated with processing at 44.1 kHz (EQ algos with high frequency curve compensation - sometimes called 'decramping'; saturation plugins that rely on high quality nyquist filters instead of oversampling). There are numerous other factors that may contribute to different results when capturing and/or mixing at different sample rates. One persistent myth about recording at higher sample rates is that capturing more 'points' will produce better fidelity. This is not the case. as long as the clocking is stable and the nyquist filter is well designed, higher frequencies will be 'smoothed out' on playback (the one arguable exception is that mixing at higher sample rates will allow for higher frequency interactions; certain high frequency interactions have the potential to influence the magnitude & phase of frequencies in the audible band, though these artifacts are typically very subtle, and in most instances would require an extended frequency playback system to produce an audible effect). More simply put, conversion at a given sample rate is different than mixing at a given sample rate; each has its own considerations. Most of the confusion results from treating these two separate processes as the same thing (much like capturing at a given bit depth and mixing at a given word length).
Some VST instruments are desined to work with higth SR's to avoid aliasing - some plugins got a oversampling *2/*4/.. option to let you mix with with 44/48100 samples per sec, and still avoiding the aliasing
@@recordingrevolution Its true, and might matter esp if you're tracking VST instruments, where any latency btw performance and monitoring can fuck up a take... I don't think a live band would benefit at all b/c they'd be zero-latency monitoring the signal before it hits converters (or just hearing themselves in the room)... but for mixing, the latency subject doesn't matter at all, imo Who cares if there's a millisecond diff between hitting 'play' and playback, or the response of processing adjustments?
Latency is set by number of samples in the buffer, so a 32 sample buffer at 88,200 Hz is half the time long as a 32 sample buffer at 44,100 Hz. You get half the latency, but your CPU in turn gets more stressed.
Yep, I agree. Great advice. I have definitely been recording too hot. I felt in my gut that I could record lower but I had this stigma of recording with headroom but just barely. New sub sir.
Well i dont totally agree with the sample rate argument, i perfer higher fidelity when recording and capturing the most natural and tru to life signal. Not to mention if i pay for a interface that can give 96k then i want to use it to its maximum potential. I dont get the processor problems mixing in 96k cuz the way i record and mix i dont see a reason to dump a ton of plugings on all the tracks, maybe im a minimalist or i just get it right to begin with. I see it this way, if u need a bunch of plugins to get a great sound then you might want to learn how to properly capture a sound to avoid extra work in the mix and overloading your processor. Thats my experience and well i started back in the analog era when u had to get it right the first time, None of this ill just fix it in post crap everyone is used to now. Maybe thats why my work stands out to people cuz i take the time to get it right from the beginning and then improve on that. Remeber people in audio what u put in is what u get out, u can polish a turd but it will still be shit.
Agreed on getting the sounds right to start. If you track with a UAD interface and a good preamp and channel strip with just a bit of compression, you don't have to do much more on the track other than some EQ. The rest of the FX can go on a mix or master buss. I prefer to track at 192kHz when possible, especially when dealing with things with intense transients or really organic sounds like a cello or vocals. Even at 192kHz, I'll still use oversampling on FX that have it. Even though you're rendering to 44.1kHz, you still get a way better sound as long as you're using a really good down sampling algorithm. Which brings up a question. Rendering out of REAPER at native resolution and down sampling with r8tebrain is a lot faster, but does anyone know how it compares to REAPER's highest quality down sampling?
Everything you do with hardware before the signal goes into your DAW is definately better...BUT it takes 2 things I don't have......equipment ($$$$$$) and experience
Untrue. When recording audio, the signal is naturally and electronically filtered by the mic/preamp at around 40-48 kHz. Recording at higher than 48k is pointless, there is no signal left to record in that upper domain of the spectrum.
Shouldn't you be mixing at the sample rate of the recorded audio? If you get a session full of 96khz/24 bit audio shouldn't that be what you hand the mix as to the mastering engineer and let them deal with conversion?
You're right about sampling rate. People don't realize that it only applies to frequencies over 20Khz, the idea that it's "clearer" is a myth.Yes there is a an argument for a difference in terms of anti aliasing filters effecting frequencies below 20k at 44.1 , but especially in pop music this is immaterial compared to losing half your mixing power. I've never met someone who can tell higher sample rates in a blind test. Look at most high end pro's mixing, they're almost never using high sample rates.
And I should say that putting critical plugins in oversampling when at 44.1 (like plugins on the mix buss or groups) can help with anti aliasing filter issues...
►► Create radio-worthy songs from your bedroom. Download my FREE Radio Ready Guide and learn my 6 step process → www.RadioReadyGuide.com
1. 1:04 - No gain staging
2. 7:15 - Sending too hot of a signal to the mix bus
3. 9:40 - Mixing at super high sample rates
Cheers, had this video on 2x haha
You are a beautiful person.
Brilliant. That'll do. No need to watch now.
Perfect. Should be in the description 😉
@@joelonsdale first time ever seeing someone put in timestamps, eh? 🤦🏽♂️
i think these excessive gain issues are from the conditioning that many of us received from recording on tape, where you actually benefitted from as hot a signal as you could get before clipping. The beauty of digital is; you can get away with much lower volumes and still retain the quality of a hot recording, thus leaving you plenty of headroom for signal processing....thanks again man, great stuff!
Another HUGE mistake many make. Don't keep adding treble to everything to clarify it. Remove the mud around 400-600 hz and you'll be surprised how that clears it up and does not wind up so brittle sounding in your mix. Often times what you remove is more important than what you add, not to mention, a lot of energy is in that region which leaves more overall headroom when you are gain staging. But most important of all, Listen! Removing TOO much is not good either because you will lose warmth. It is indeed a balancing act. Again, pay attention to what you are hearing.
@Aiden Macleod What a stupid thing to say, boosting has it's place you just have to know what you're doing and what you're going for.
@Aiden Macleod Big words coming from the guy that says to never boost with EQ. The way how you contradict yourself is incredible and unbearable.
EQ is everything. Get a GREAT sounding mix in mono using EQ before spacializing or adding anything else... your mixes will KILL.
@Aiden Macleod Right back at you dude, don't give up, keep pushing, you can get that worthless degree; in the mean time, keep believing internet myths like never boosting and such. The only douche bag here is you coming here with your nonsense and myths like if you are some kind of George Martin. Don't spread your bullshit pseudo-knowledge.
All you smart engineers technicians etc. Man said don't keep adding treble to EVERYTHING to clarify mix!!! Sometimes its beter to remove mud which is somewhere between 400 and 600 Hz. Even lower. It brings more dynamic to mix. And you have to decide what freqs to add to what and what to remove to what. If you have "boxy" vocal, you won't save it by adding trebble freqs. You have to find that "boxy " freqs to make vocal clearer. And so on for acoustic guitar and other instruments. Then you'll get mor headroom for mastering and finally get master much louder without loosing dynamics...
When you stink you Won't get rid of it with perfume. You have to wash yourself and then put the perfume on
THANK YOU for talking about gain staging. So much nonsense going around that it's not important. Thank you for another great video :)
Hey i love you videos too
yes very much
I don’t even have FL Studio but I love Michael’s content and explanation to recording which you don’t need FL to implement. Thank you for your contribution as well Michael! 👍
Who says it's not important? It's a must! You literally can't play a song through a DAW if things aren't gain staged. It will stop playback if there is to much sound pressure. Purestgain from Airwidows goes on every track, buss, folder, VCA and Master buss in every project. I rarely even use the track sliders in the DAW any more. I just put the Purestgain controls on the track and use those unless I'm lazy. The sound difference vs. a standard gain plugin is noticable after about 20 tracks.
Aiden Macleod disagree, you also need a bulk of rockwool in your room for good mix))
Happy Christmas!! Thanks Graham for posting this and encouraging so many people to record and make music!
You just gotta love these guys. :)
He's a great guy! Really helps so many people!
@@ProducelikeaproYou guys both are! I have incorporated so many of your techniques in my album recordings.
Hi Ten Degrees that’s wonderful to hear my friend!! You Rock! Happy New Year!!
Then you can take your music and let spotify steal it from you
The way he says 'hey friend' at the start makes me know I'm in the right place.
Adriano Murgia so true my friend
Facts
lol when he said “hope you’re having a musical week” I was like oh wow I’ve never heard anybody say that, thank you
Hahahaha
Bob ross kinda vibe lmao
It would be impossible to stress enough how important gain staging is. I used to make that mistake constantly.
I get people to ask me to help them fix a mix who have a bunch of fuzzy black rectangles for files a ton.
Their eyes invariably light up when they see how much that one tiny step did to improve the ability of a track to breathe properly.
For about 10 years I didn´t take care of gainstageing at all and always had critiques like " I hear a hiss on the vocals", " Something is causing distortion" etc.. These days, and in particular after joining Graham´s Mixing University, my mixes get a proper gainstageing and...BOOM... they sound better!!!
Pumped to have you as a student!
Maybe you just got better at recording after ten years some people do
@@thedevilsadvocate5210 Yes, hopefully for sure! But some things I always just did wrong I now do right ;-)
How could you not gain stage? Most of the stems I get will keep a DAW from playing them back for more than a few seconds before they stop playback to avoid overloading the system.
Hi Graham. Quick pointer on the the Waves SSL Channelstrip. The fader is output not the input. The switch is only for the metering. Also the input knob adds saturation to the sound (not a drastic amount but still) So I tend to use clip again for the input, the input knob if the compressor doesn't have much to grab on to, and the output fader to balance out the processing.
Essentially what you said, but the fader thing I had to point out just in case.
Thanks for pointing that out!
@@recordingrevolution You're welcome. Thank you for all the great content you put out over the years !
The saturation is WHY people use an emulation of a channel strip like the SSL, though UAD's is the only emulation that sounds like the real thing. The Waves emulation just makes noise, and the Slate Digital is only slightly better. Still, they do as a touch of analog feel, but not the warm richness and truly harmonic analog distortion that the real hardware or the UAD adds. If you're not wanting the saturation, you're much better using FabFilter EQ and compression. Even if you do want the analog feel, many would argue you're better off using the FF plugins with the opto compression in Pro-Q2 and Saturn for saturation. In any case, there is no point in using an SSL emulation if you're avoiding the manufacturer"s attempt to sound like the real strip AFAICT. Especially the Waves version. We ought to all be boycotting Waves anyway, as they are an Israeli company who steals technology, often markets snake oil plugins with trendy names endorsing them just to grab a few bucks from naive young audio engineers who want a magic button, and support the racist, apartheid Israeli regime that commits more war crimes against humanity and war crimes in a week than any other country"s government's commit during the entirety of their existence. Also, their plugins rightly belong to the American tax payer, side they pay for whatever R&D they don't steal with billions in foreign aid that never gets repaid. We ought to not just be boycotting them, but freely distributing their entire library to the American tax payers to whom they belong.
@@xaosnox
Just say you don't like a plugin. Nevermind the anti semitic rant
Higher sample rates don't mean "smoother" audio representation! This is a common misunderstanding of digital audio which isn't always that intuitive. Sample rate dictates the highest frequency you can record. Humans can hear up to 20 kHz at best so we don't need to record frequencies higher than that. Sample rate must be at least twice as large and that's why 44.1 kHz is enough for human ears. Higher sample rates do not increase fidelity. The sampling theorem says you can reproduce the signal 100 % as long as the sampling rate is at least twice the highest frequency in your signal. DACs create "smooth" analog signal from the sample points regardless of the sample rate. Higher sample rate means redundance in this sense. 88.2 kHz sample rate is not 2 times "smoother" than 44.1 kHz, it is the same fidelity defined twice. Somebody telling you twice you won the lottery for $5 million doesn't mean you won $10 million! 44.1 kHz sampling rate already defines the sounds we humans can hear with 100 % accuracy. That's the power of digital audio and something a lot of people do not understand well because it goes against our intuition.
Similarly bit depth doesn't define fidelity in the way many think. When dithering is used correctly, bit depth defines only the level of the noise floor. For every additional bit, the noise floor drops 6.0206 dB so that theoretically 24 bit allows about 48 dB lower noise floor compared to 16 bit. However, 16 bit already allows over 90 dB of dynamic range and when shaped dither is used, well over 100 dB of perceptual dynamic range. In consumer music no more than about 80 dB of dynamic range is need. This means 16 bit digital audio already gives more dynamic range than we need. In music listening 24 bit is total overkill, but very useful in music production because audio levels aren't optimized. So, 24 bits when producing music, and 16 bits when listening to it.
So why does your DAC sound different with different sampling rates? The most important reason is expectation bias. You expect higher sample rate to sound better so that's how you hear it. Simple placebo effect. It's amazing how much our expectations affect how we hear things. Careful blind test removes this aspect. Differences in audio tend to disappear in blind listening tests. Another reason is that your DAC really creates different analog sound from different sample rates. The difference is real, but it doesn't mean the higher sample rate version has more fidelity. It means there is a difference, two interpretations of the signal. There are perhaps slight differencies in high frequency phase response, but that doesn't mean difference in fidelity. Your DAC giving the best sound at 192 kHz is not because 192 kHz has more fidelity than lower sample rates. It's because your DAC has the most pleasing imperfections to your ears at that sample rate, who knows maybe because of expectation bias?
Summary:
- digital audio allows 100 % fidelity within the frequency and dynamic range limits.
- sample rate defines the highest frequency we can have in the recording.
- bit depth defines the level of the noise floor (dynamic range).
- 44.1 kHz sampling rate is enough for humans. Do you hear above 16 kHz?
- 16 bits is more than enough in music listening.
- 24 bits is beneficial in music production because levels aren't optimized.
- be aware of expectation bias.
In regards to sample rate. I was always taught that you should be recording at 96kHz (or better if you choose to do so). The reasoning behind this is while working with outboard equipment (consoles, comps, eqs, etc.) you are not recieving a max frequency of 20khz. You are recieving whatever frequency that specific gear is outputting which is not a perfect sine wave. Just because you can't hear it does not mean it is not there. Recording at a higher sample rate creates redundancy (as I believe you stated, on mobile. Difficult to check lol) to help ensure you don't have any aliasing. For anyone who reads this and doesn't know, aliasing is when a recorded sound has a higher frequency than the sample rate can properly capture (divide the chosen sample rate by 2. This gives the highest possible frequency you can record with little to no problems). The effects are frequencies that playback incorrectly like a constant hum that is not there and not caused by your equipment rather caused because there happened to be a frequency higher then the sample rate can handle. IMO if you exclusively operate within your DAW I do find it a bit useless to record above 44.1k or 48 since you won't be using any outboard gear, or very minimal if you do. Aliasing is also another reason why many monitors say they can playback sound at much higher frequencies then you can hear.
Some plugins don't oversample. And that can be obvious. Not just very subtle, but very audible. So you need to test them on your low sampling rate and your high sample rate to make sure there isn't any difference.
"The sampling theorem says you can reproduce the signal 100 %...." -- Nyquist was speaking in the mathematical realm. At some point we must deal with a host of problems in different disciplines when entering the realm of physical implementation.
@Joseph Winett That dude uses strawman arguments and sine waves in his examples. Sine waves don't have harmonics, audio is more complicated than his tests.
Radio Sound is the best source for human hearing tho , try to listen ( flac ) has higher mb compared to mp3 ( smaller file ) has less quality sound compared to wav ( so obvious)
I recently started applying steps 1 and 2 when I found out what they were. Never had a problem with 3. Thanks Graham.
When it matters, many good plug-ins offer an option called 'oversampling' which effectively lets you work in a 44k session, but flip a switch when it's time to export/bounce/render the audio, which tells it to basically calculate at 96k under the hood when oversampling is on.
This rarely matters unless you have a specific problem that you are trying to solve.
For what it is worth, to my ears the best sounding guitar effects processor I have ever owned (Digidesign Eleven Rack) sounds like a buzzsaw at any setting other than 96k!
"The higher the sample rate the smoother the signal will be"
This is a bit of a misconception! Based on the Nyquist-Shannon Theorem, when your Digital-Analog Converter (DAC) 'interpolates' or connects-the-dots between the samples being fed to it, it does so in a way that would be equivalent to the continuous, real-world electrical signal that existed below half the sample rate (bar aliasing, which is combated with the reconstruction filter being set a bit below half SR). The content below 22050 hz in a 44.1khz sample will be *identical* to the content below 22050 hz in a 192khz sample.
One might go as high as 48khz to give the anti-aliasing/interpolation filter room to breathe and appease those who claim to have super-human hearing extending above 20khz, but going beyond that wastes processing power & storage space for no additional benefit to you or the listener. It can even be detrimental when accounting for factors like jitter.
As you said, super high sample rates are unnecessary!
Tristana Wild Correct, unless your plugins don't oversample.
aceyage oversampling is useful for some specific DSP mathematics, especially frequency domain stuff, but it’s not going to add to the fidelity of your signal below the nyquist frequency of your overall project. It’s not particularly relevant to the issue at hand.
Tristana Wild There is some really audible aliasing wiith some, usually older, plugins or those that don't oversample. That’s where higher sample rate can help as those mirroring frequencies are now out of the hearing range. I won’t use plugins that don’t sound exactly the same at 48k or 96k. Working at 96k is just unnecessary CPU wastiing otherwise, agreed.
went back to my first real practice mix (I'm a noob, learning by fucking up to be honest!) and went through EVERYTHING to gain stage properly. one track, one plugin at a time... what a world of difference, really brought out a lot more clarity from everything. I was able to knock a lot of the different points of compression down a bit because of it, and really opened up the sound of the song. Thanks for this!
about sample rate . For me it's true for mix only , because in sound design we use a lot of time stretching and so high sample rates are essential
Ohhhh.. so when you time stretch should you resample at a higher rate 1st?..I understand that it's ideal to run the project at 96k but if I'm just making a track or scoring for a video. How would I handle that situation.
@@rendilly No. You acquire at a higher sample rate. When you time stretch or repitch, you're starting with audio with a maximum frequency around 24kHz. Slow down by half and you're now at 12kHz. If you acquired at 96kHz, then slowing down by half still leaves you with audio at 24kHz and plenty of high end. If you're acquiring at higher sample rates, you need a quality clock source as the timing of your converters becomes more crucial.
You've made it clear talking about mixing in 44.1 kHz. Such a great point!
Just a correction regarding the sample rate. It is not like higher resolution in a photograph. That's a big misconception. Whatever sample rate you are using to record, it is always 2 samples per frequency. So, 96k doesn't provide more accurate sampling, but rather expands the frequency sampling range to 48KHz, which is way beyond human hearing. There are other things that go into play when recording at higher sample rates, but that's another discussion. I agree with you regarding the benefit/ drawback of higher sample rates.
And you should always mix in 32bit floating point environment at least, never 16bit.
Aside from the arguable benefits that you mentioned of higher sample rates (not opening that can of worms here, but for what it's worth I track @96/32float and have 60/70 tracks on a 4 core i7 imac and I haven't maxed it out yet.)
I do believe many synth plugins actually sound noticably better at higher sample rates, and I've also heard some performance related issues reported because sometimes the synth plugs will be processed internally @96k, with the DAW @41/48k. While I imagine quality and updated plugins will make this a less frequent issue, it is something to consider.
as in, the daw sampling it down live?
@@vooshmoozik6185 I'll see if I can find a link to the performance issue in that regard but here's one angle of the conversation.
www.keyboardmag.com/lessons/should-you-record-at-96khz
I agree with everything regarding the Sample Rate with keeping in mind that most people are doing their stuff in a homestudio environment or just working at a small professional level. If I'd go in a bigger studio with a major project I'd want at least the 96khz Sample Rate. There is a (small) audiophile market that wants to be served, too. Personally I work with 48 khz and I'm totally fine with that. I'm even considering getting back to 44.1 khz just for the performance gain.
I agree - just like a good song, I haven't heard a mix musical enough in years to warrant a high sample rate. Thanks for the tips!
i haven't made music in years...finally have enough to buy a new imac...I've spent the last 6-12 months watching videos like this to figure out what I was doing wrong years before, so when I finally can start, I'll have more knowledge right away..
Thanks for making this video....
One thing I think was missing from point #3 (recording/mixing at higher sample rates) is when recording at 44.1kHz the anti-aliasing filter (low-pass filter) that prevents frequencies above 22.05kHz from being recorded can also roll-off some of the high frequencies in the range that you want to record (below 20kHz). This is essentially rolling off high frequencies with a high-shelf EQ on every track you record. By recording at higher sample rates, the anti-aliasing filter can be designed better because it needs to reach 0dB before 48kHz (if recording at 96kHz sample rate) instead of 22.05kHz, so it has plenty of time to reach 0dB without affecting frequencies that you care about under 20kHz.
It's explained better here:
www.indiana.edu/~emusic/etext/digital_audio/chapter5_rate.shtml
That is the strongest argument I've heard for recording at higher sample rates. Maybe recording at higher sample rates to eliminate anti-aliasing filter problems and then batch converting to 44.1kHz for mixing to save CPU power would be the best of both worlds?
I'm a beginner, but having this knowledge up front will help me to avoid mistakes going forward. Thanks for putting out these videos, subbed. I call myself a beginner because even though I started many years ago with Sonar 8.5 studio, then upgraded to Sonar X3 studio, I never put in the time and effort to really learn the software. I'd bring it up and basically fly in the dark, just figuring out the very basics of laying down a few midi tracks and vocals with some reverb. I recently downloaded Cakewalk by Bandlab and now I'm making a concerted effort to learn it. I had given it up for years because obviously if I didn't know what I was doing, the creative process would drag to a crawl. Instead of making music, I was struggling with the software to the point where it wasn't fun anymore. It's a ton to learn but I'm being patient and not trying to learn it all overnight. But every day I try to learn something new and build my knowledge base.
This video is pretty spot on based on many home recording projects I've heard as well. Understanding dynamic range and gain structure are basics anyone getting into recording should understand if they want their recordings to sound good. I went from analog to digital when DAW's were first being offered around 1990. I had Turtle Beach 56K and Cakewalk by Twelve Tone. There was no forgiving if you tried to record too hot of a digital signal; it just overloaded and broke apart.
Most don't seem to have a grasp of EQ or the fact they compress the hell out of the signal as well these days. Most music being recorded not just in home studios but commercial ones lack range and nuance and most performances are mediocre relying on studio magic. Don't get me going!! Good video for beginners though.
Doesn't it start with the shitty song?
I argue about all 3 of these all the time. Some people just think louder is better and higher numbers are cooler. It is what it is. Great video as always!!
Not sure if anyone's mentioned this yet, but you can also get your level by CMD+clicking in the volume section of the track to toggle through to see peak level (you'll toggle past delay compensation value to see this), which some might find useful. Cheers.
When recording high frequency instruments like violins and cymbals, you might still want to consider a higher sample rate, so there is less risk of aliasing, i.e., the ultrasonic frequencies traveling back down into the audible range as inharmonic distortion.
Especially true (and clearly audible) when doing any pitching up of hi frequency audio. You want to do that in a separate session running at a high sample rate, then low pass your sound somewhere around 20k, and then import it back into your session at 44k. The difference can be massive.
Thanks 🙏🏽 I never realized I never gain staged properly! All these years I thought you wanted sounds right before the clip! 😂
good tips. I do my own gain staging from the sound design, I always stay in the green on the mixdown and give 8 db headroom for the master, and I do use 48k/24 bit when exporting.
Lots of good information here. However, I like to mix at 48k because my focus is on sync licensing, which requires 48k 24 bit mixes. 44.1k is good for commercial releases. CDs are only compatible with 44.1k 16 bit, but every music library and sync licensing opportunity that I have seen requires 48k 24 bit. That said, I use 32 bit floating point for recording, mixing and mastering and dither my masters down to 24 bit for sync licensing and resample and dither to 44.1k 16 bit for mainstream releases. Other than CPU conservation, is there an advantage to recording at 24 bit instead of 32 bit FP?
This is the first time I've posted here, but I've been watching your videos for several years now. This is one of the top sources for very concise instruction for mixers, whether they are experienced or just learning the skill. I consider myself closer to the latter.
I would like to address the last issue, Sample rate. In my limited experience (I mean that sincerely), I have the understanding that sample rate is like pixels in digital photograph. So the more 'pictures' that you take of one second of music, the more accurate the digital representation of the analogue source. For mixing, 44.1k is plenty and you will probably never hear (some say that they can) a difference. But...
One of the issues that comes up in my little recording world is dealing with some singers who don't sing accurately; so pitch correction is needed. This, I believe, is where 'more samples' benefit. If you have a higher sample rate, you can more accurately correct pitches. This could also apply to alignment. And a big one is creating harmonies from a melody vocal. I do this a lot. I use Cubase (newest full version), which has an amazing pitch correction feature. But for building harmonies, I always use Melodyne Studio. I find when using higher sample rates, there are less artifacts; there's almost never none; but for harmony, they aren't noticeable. So, I'm of the mind that one of the purposes for higher sample rates are for things like this. The actual mixing doesn't require the higher sample rate, but the editing beforehand does. So recording in a higher sample rate is, for me, to allow for more accurate editing. We don't all have singers that can do the job, especially if they are going to home studios owned by newbies like me. So everything is recorded, edited, mixed and mastered in the high sample rate and only after mastering is it re-sampled that dithered.
Right now, I have been recording at 88.2k and have been for over ten years, but my reason had more to do with the computer's math co-processors only having to divide by two to lessen the load on the cpu. I don't know anything about this stuff, but it sounded logical. But those days are long gone; computer power and OS,s are a none issue now, at least for audio recording. We now have super fast computers and inexpensive large storage.
I'm going to move up to 96k in the new year if only to be able to cut it in half for video tracks, which are, as mentioned in the comments, the sample rate for video.
All of this is just my accumulation of knowledge put into my recording practices. If I have misunderstood any of this, please let me know and I look forward to hearing about it.
So Merry Christmas to you and yours, since it's been a year since you posted this and it's that time again. Holiday wishes to all.
@recordingrevolution, you have a very good point at 10:00 and I go beyond that: the same applies for bitrate!
If you let your whole project in 16 bits, the plugins still providing 24, 32 and even 64 bits in their outputs.
I agree that in mixing phase you will be listening at 16bits, but when you generate a mixdown file, you probably will have the option to set to 96KHz and 24bits for mastering! (at least, Sonar has)
That way you still having as 24bits "like" files.
That's why I still recording in 44.1KHz/16bits (as you said, the "wins" are too low), but I can count on "upsample" and "upbit" from plugins.
Not using Reaper was a mistake I made. So glad I found it.
I know nothing about Studio One. I was using a crack of Cubase 5 when I started and didn't understand anything. The first thing I did in Reaper was to send my vocals to another track for an EQ'ed Hall....WITHOUT READING ANY INSTRUCTIONS........I love Reaper
@@RichieHmusic It is intuitive. The other thing I love about Reaper is value for money. The Reaper folks haven't bought into that subscription model, and have decided to ask a fair price. It will pay off for them in the long run, big style.
@@RobertLinthicum totally agree. I refuse to use anything with a subscription model or iLok on principle. People who say REAPER isn't intuitive escape me. It's easier than anything to start with, and every single thing (except getting rid of the big clock) is customizable. Plus, with the ability to create custom actions, or macros, you can cut your editing and mixing time by 75%. People who say REAPER is to hard are the kind of people who say reading is to hard because they don't want to learn the alphabet. Very little that commenter says makes sense. Studio One is a nightmare compared to REAPER. Even Logic is agonizing after you get used to REAPER. There's no other DAW that will let you set it up so that everything you do is a key command or mouse click away, and everything you don't need is out of your way.
@@xaosnox Amen! Long live REAPER.
Reaper is just awesome. I absolutely love it.
You should be checking your track levels before you start mixing, when adjusting your gain stage make sure the track fader is set to >0< then adjust for gains to -18 or -12 whatever,,drums, bass , lead vocals on the louder side.
About #3 of no high sample rate because it kills your CPU? Here are a little two somethings that everyone seems to either ignore or forget:
1) It's only a MISTAKE when you are recording at high sample rates in your DAW when your actual input does not SUPPORT anything higher. If your audio interface doesn't go above 48, then there is no point in setting your DAW higher either.
2) You can very well use high sample rates without killing your CPU. You record the raw signal in high sample rate, then COPY the original file, CONVERT to a lower sample rate, open different project with the lower sample rate, do all your edits. When you are done with everything, you duplicate the project then set to high sample rates, replace the source files with your high sample rate files, because all edits done have been done from that one specifific original file. Bounce out, phase-shift, compare to the lower sample rate export. Do minor adjust. Done! You just have futureproofed tracks without straining your processor. You're welcome :)
There are some benefits to high sample rate recordings other than just higher sound fidelity though but this has to be judged by everyone individually. If you have to do a lot of processing and time stretching, then you go as high as you possibly can. Fewest tracks require that though. But for people doing sound design, that's a very different ballgame.
Edit: btw those large, clunky files people complain about? Fewest people CLEAN UP their project to only have the files remaining that are actually necessary, alongside a copy of the main source files. They keep every scrapped piece in there. Please clean your bin, fellow recording engineers. Also: Those files can be COMPRESSED for archiving :) Just saying. Multiple backups notwithstanding.
Video uses a 48K sample rate by default so I no longer use 44.1K because for me, video is a more likely distribution format than physical CDs.
Makes sense to leave headroom for mastering. I'm going to try that. I always went for maximum volume before mastering. Although mastering didn't make it clip, it did limit how much mastering I could add to it.
From a technical standpoint, I don't have anything to offer to this discussion but, from a usability perspective, I think that a lot of "mistakes" can be circumvented. For example, beginning users in ProTools or Cubase will start their first projects based off the templates. The initial focus when starting out is getting a clean recording from their sound interfaces and making sure it's not clipping on the meter on the hardware device. Eventually, with more practice, it becomes fluid practice to observe hardware meter and track meter and master fader meter. It's a balancing act for the eyes initially. Most tutorials state to adjust volume at the track level and leave the master fader alone during recording and mixing. Things can get confusing for those composing their songs in sequencers. I remember seeing one tutorial in a hard copy guide to record your instruments as high as you can without clipping (sequencer recording on some midi capable keyboards). There's definitely a lot to learn that it can become overwhelming for first time home users. Which is why, after watching your video, I had a brainstorm that, in addition to project templates that DAWs provide, they should include a finished project that shows it's evolution from conception (all tracks recorded), mixing stage, mastering stage, and what it might look like when targeted for various venues, i.e. streaming, CD burning, radio broadcast. This way, beginning home recording enthusiasts have a standard of measurement to use when developing their projects. The analogy I have for this is sculpting museum figures. These people had the benefit of seeing what a finished sculpted figure looks like as its displayed in a museum. So they have a working visual standard of a product worthy of inclusion in a museum. Home recording people would benefit from the same. There's nothing in the purchased product that shows and/or adequately describes what a project should look like for bouncing to disk for the various venues.
Thanks for yhe clarity on Gain staging I do it on the Clips but didnt know that the faders only effect after you clip gain
Hey great tips, especially gain staging. But I must say that while it's hard to hear the difference between 44.1kHz and 96kHz sample rates, the processing that's done in your DAW definitely benefits from the highest sample rate you can manage. Depending on what your 're doing with your plugins, high sample rate can sound noticeably better. For instance if you are doing any kind of time stretching the higher the sample rate the better. In general, higher sample rates result in cleaner processing, because there is more data per second. It has to do with the math involved, but suffice it to say that the higher the resolution of your audio, the better it will sound after processing. Also when you down convert to 44.1kHz it sounds better the higher it is before conversion. Again it's the math your computer is crunching that makes the difference. Then again, you still may not actually hear the difference, but someone will be able to. Great tips for beginners though Graham.
Hi Monte, please read my explanation above why the FX processing sounds better with higher sampling rates. Your perception is correct.
@Beeblebrox One Sometimes, "digital" can be a source of misunderstanding, even for pros. you might find that video interesting: th-cam.com/video/cIQ9IXSUzuM/w-d-xo.html
Please watch it to the end. This is the best explanation worldwide imho, and I say that as a person who has a degree in audio engineering. I love that guy. Highly recommended!
@Beeblebrox One By the way: This is for the readers- dithering is not used for down-sampling, it is used for the process of bit reduction. (24 down to 16)
@Beeblebrox One Oh really? Almost every "top" engineer I've read says use higher sample rates, for the obvious reasons. Dithering does not cause problems as far as I know. In fact, when dithering adds that small amount of noise to the signal, it actually improves audio resolution, especially in quiet passages. And, as Johannes Mazur pointed out, dithering is used when reducing the bit rate, not the sampling rate. A little bit of knowledge can be dangerous. You can mix and record at low sample rates if you want to, but I'll continue to record and mix at high sample rates, thank you very much.
exactly... and if you make a final decision and bounce the processed track the cpu usage is negligible...
The issue of high sample rates is actually quite nuanced. With audio signals entering through an interface, you likely won't hear any difference by recording at 44.1 kHz or 96 kHz. However, with sounds generated inside the box, it can make a big difference because, for example, an instrument recorded at 96 kHz may not have aliasing that would occur if it was recorded at 44.1 kHz. This is not a "wine tasting" difference, but something that sounds obvious even to someone who's mixed FOH all their life :). So...instruments that oversample internally should solve this, right? Well...mostly, but the problem there is an algorithm optimized for real-time oversampling may cut corners compared to offline, non-real-time upsampling. In most cases I record at 44.1 kHz, but for virtual instruments, save the MIDI file, open it up in a 96 or 192 kHz project, render at the higher sample rate (which means there will be no aliasing in the audio range), then downsample back to 44.1 kHz. Because I'm downsampling audio that has no aliasing, there won't be any at 44.1 kHz whereas the virtual instrument output would contain aliasing. This is a very basic example, matters are much more complex with plug-ins like amp sims, some dynamics processors, etc. The bottom line, unfortunately, is that there is no "one size fits all" solution. The people who say "recording at 96 kHz is ridiculous" and those who say "recording at 96 kHz sounds so much better" can both be right, depending on the project, the software being used, the plug-ins, when the plug-ins were coded, who coded them, etc. etc. etc.
Complete and utter BS. You must have sh!tty gear.
I’ve been following you for little while, but I just want to say I really love your videos. You go into so much depth without getting off topic and rambling. Thanks for all the info you share
"How you were raised" LMFAO! Legend.
The sample rate tip was huge to me, I never knew that before. Keep up the great tips I always refer back to you every mix!👍
Great gain-staging points btw. Anyone using Ableton Live, if you want to Gain-Stage, you can use the "Utility" plugin. Don't use the "Limiter", it colors the sound a tiny bit. But the Utility is equivalent to lowering clip level, 100% transparent.
What a difference!!! I'm used to live mixing on a big (Allen & Heath GL2800 56 channel) analog board. Their manual has a section on gain staging. They want every channel, every subgroup, all the mains... up in the yellow with peaks at +3 db. That, they say, gives you the best signal-to-noise ratio possible. For live applications, that holds water. Apparently that is NOT the case for digital recording. Guess I better get to learning if I want to get into using ANY computer software, huh??? Already clicked on the link for your FREE Radio Ready Guide.
This is GREAT!! Recently got back into music/production, these are things I didn't know that save time instead of a lot of trial and error. I was working to hard on the end process trying to get the "right sound", when I could have corrected it at the beginning. Good to know. I knew nothing about gain staging before today!!
Thank you Graham! Just applied those basic rules to my latest project and it sounded infinitely better! Never really took the time to understand it all before. Great video, clear and straight to the point.
Speaking as a softsynth/DSP developer. Common misconception: sample rate is NOT a stepped resolution. Most audio engineers think that a digital signal is defined as tiny blocks with each having an associated amplitude. Some imagine it as a staircase. That's simply not true.
What sample rate means is that at a specific interval (44100 times per second for standard cd's) an amplitude is defined. They are points in time, not blocks. When this data is converted by the DAC to an analog signal, a smooth line is 'drawn' between these points. If your digital signal represents a 20 kHz sine, that sine is perfectly reconstructed by the DAC. Or any sinusoid below the Nyquist-frequency. If you run this DAC-converted, digital signal through an oscilloscope, you won't see any 'staircasing'. Your sinusoid will be perfectly replicated. That's why having audio at higher than 44.1 kHz sample rates is ridiculous. We can't hear beyond 20 kHz. And if you claim you can, congratulations; you are a dog.
Think of it like this; is having a screen that emits a spectrum of light that extends beyond our visual spectrum a good idea? No. We can't see ultraviolet or infrared light, so there is no point in adding this extra data in a visual signal. It would even be harmful in the case of uv-light. Then why would we need an audio signal that contains data that's imperceptible to our ears? We don't. It's pointless.
On an actual console you start with the channel faders at 0db then you adjust the pregain so it isn't clipping while EQing the sound and then once you get it set, pull the channel fader back and do next channel, once all set then start mixing... That way you know once you hit unity gain 0db on the channel fader anything above that will start clipping.
Thanks for this one. Gain staging is not a step I was aware of until now. I did a quick hunt on how to do it in my DAW and think this will make a massive difference. Much appreciated
1. You should do the gain stage with a vumeter aligned to -18 rms, thats the optimal signal/noise level
2. Each track is different, the drums transients are different in each song, so u can't focus on peak level, there's an old system created by bob katz, the K level, for comercial music u should set the master to 0 vu calibrated to k-12, that's like -12 rms aprox. Not only for avoid clipping, if you always mix at the same perceived volumen and yours monitors/headphones knob volumen is always in the same position your brain will remember how sound everything and you will take better decisions. Also with this system you always mix on the same point of isophonic curves (Fletcher-Munson)
3. I mix at 44.1khz because my cpu can't run all the plug-ins at 88.2khz, but if I u can, mixing at 88.2khz its the optimal between quality/AD-DA errors and its the exact double of 44.1khz so the downsample is a very simple mathematic operation for the SRC (just divide by 2), also at 88.2 kHz all the plug-ins goes automatically oversampled (even plugins without oversampling option) so avoid aliasing and all the plugins works x2 quality.
THanks a lot! Immediately remixed my most recent project with these tips and it fixed all the problems I was banging my head off the wall for!
Cool, thanx graham. I know this but i'm gonna go back today as i start recording my latest and be diligent. Good reminder
I agree with the first two mistakes. The last one, well...
I think the best way to explain it is this:
If you only plan on using a picture for Instagram or a Facebook profile picture, or even an album cover...
Will you take it in a 1mp camera, or will you try and get the best possible image quality?
Exactly the same goes for audio. Just because it will end up in a 44/16 mp3 doesn't mean you don't have to record the best possible way.
I have to agree with you on the high sample rates limiting how many tracks you can work with without pushing the CPU. I capture my individual audio tracks at 48 khz /24 bit and couldn't be happier with the quality I get. Rather than doing a two track mixdown on the multi-track station I pass this signal via analog to a stereo tube preamp where I can slightly saturate the tube preamp for a warmer feeling tone that is then recaptured at the same 48 khz/24 bit sample rate on the mastering station. Once the 2-Track stereo audio file has been mastered, I then reconvert down to the standard 44 khz/16 bit sample rate during the mastering process.
Hey Graham, I guess you won't read that, but there is something that makes mixing at 96k not that pointless at all.
Due to Nyquist, the ordinary stock EQ gets an asynchronous bell shape at high frequencies. This is what they (umm..we) call non-analog behaviour, as it is a digital EQ side effect. The most popular developer-work-around is a switchable built-in oversampling but that introduces latency and eats some CPU-Power. When you mix with higher sampling rates you don't need oversampling. The curve of the high EQ-band behaves like its analog counterpart and sounds more natural as the Nyquist frequency slides up. That is the true benefit of mixing with higher sampling rates. You don't need to switch on internal oversampling (where it is available within a plugin) and therefore actually safe DSP-power.
Great advice/pointers fellow human ! I appreciated especially your "controversial" thought about sample rates ... some of the best records in history were made on analogue machines that dont come close to as intricate and multifaceted as we have with our DAWS and amazing array of 21st century recording tools/options ... like good coaches in Basketball keep saying 'keep it simple' ... simplicity executed with great merit always goes further than gimmicks and shortcuts/complications.
I like mixing at 96 kHz because I feel like it definitely makes it easier to hear your eq moves, especially when it's subtle, or a narrow cut. I can really hear the frequencies better. Really depends on your hardware though. I can only manage a full mix at 96 kHz on my Mac Pro, whereas my MacBook Pro can only manage a full mix at 48 kHz. Tracking at 96 kHz has lower latency too which helps with amp sims, but it's very taxing on the cpu. Best advice, try a few different sample rates out and see what your computer can reasonably handle and just go with that.
There is a noticable difference with the initial jump from 44.1 to 48 which is easy to hear. Beyond that there are minor differences. Many people mix with 44.1 and it sounds absolutely fine. It's worth noting, a mastering engineer once said to me, if you want a high resolution, but still need to convert down to 44.1, then 88khz is the way to go because it mathematically divides neatly. It makes sense. Oddly enough he said 96khz is not always the best for bouncing down to 44.1. in terms of rounding up or down with the truncated slices and peaks.
i was told the same thing by one of my mentors. Record at 88.2 because it is twice as high as 44.1 and will be better once it is rendered down to 44.1
I though this myth was laid to rest years ago but it keeps popping up! It's only easier to downsample from double the frequency if your downsampling method is just averaging adjacent samples. Maybe back in the 90s some sketchy companies did that in their hardware but nobody does it like that now because the maths to convert between common sample rates is both simple and lossless. It makes no difference at all if you throw away 60% of your original signal rather than 50%, none at all.
@Beeblebrox One Dithering affects the bit conversion (e.g. from 24 to 16) and it has to be done just ONCE (preferably you leave it for the master engineer) and not sample rates.
Hey Graham! I just want to point out 2 things for adding to the debate.
One is that Clipping isn't a bad thing, we use it so much and because of the keep-repeating that it's bad, people take years to understand that it is actually a very important tool, specially in aggressive music. For the beginners: please take this one wisely, it's not clipping in whatever way, there's almost infinite possibilities here and you must know how and why you're doing it.
And the other one is that there's some other aspects with that last topic (as you said, controversial), I read Al Schmitt, Geoff Emerick and some tech guys talking about Nyquist filters and suppression of subharmonics of those upper frequencies. Everyone with horse power is working at 98khz and up, but as you mentioned CPU must run alongside the intent.
Thank you and keep the good work!
You are referring to analog clipping, right? Digital is the opposite. Plus you can always add more volume and punch using limiters, EQ, and compression on the master, and in the digital world, it's best with more headroom to add frequencies/volume to your final product.
actually no. any clipping is tool. all tool can do more than it's supposed to do and very often they do. however, some basic understanding is lacking here.
clipping is a different thing that also can happen inside compressors and eqs (there's a famous trick of clipping API eqs). clipping is about reaping peak information hoping to have some pleasing distortion back and this effect have very little to do with the exact common understanding of compression and eq. let's get more visual: guitar amps, specially obvious in high gain preamps.
so you can clip whatever you want (tubes, transformers, ICs, plugins, etc.) as long as the sound you're hearing is what you're aiming. just a warning for kids: clipping can or cannot hurt your gear, sometimes it is worth.
great advise! I've been doing this for a very long time but your simple advice still ring as a gold standard. It gave me a confirmation of what I suspected all these time. Keep doing the good things!
As I understand it mixing at 48k mean you don't have to up sample for video and for sync licensing. It's just a tiny bit more CPU power and fits more formats for licensing. IIRC.
If you print after tracking at 192 it will sound better and you can save a lot of processing power too...
This
It's useful to hear what you say about the sample rate when mixing, for just getting on with arranging and mixing, as well as clarifying points on track levels. Nice one.
I can't thank you enough for your excellent information I have in fact made all of these mistakes and wonder why all my mixes sound terrible thank you so much keep up the good work you're awesome
eq your return tracks. thats a good one they dont tell you. you dont need a reverb and delay sending back the full whack.
Sends
I'm not sure why, but when I was mixing in 44.1 I was getting a weird artifact on one my kick samples. When I went to 48k the noise went away. So the only reason I went to 48k was because I thought I was avoiding artifacts. But now when I think about it, if the standard and industry mostly use 44.1, I might as well go back down and try to use another sample or try and treat the artifact. Great advice!
Mixing Mistakes
I apologize in advance if this has already been addressed...
I am not sure if this is common knowledge but your DAW's, Ableton does, audio strips can be used to enhance, colour, your recordings. Similar to an analog desk!
I discovered it over time. Then another musician heard it too! Very cool, if it is/was intended.
I have heard of some mixers that use the analog desk as a compressor... Max out the gain for each audio track then add the audio. Then you can add distortion, too... Let me know it you try it out? And then other mixers rarely do anything as the Mastering process will dictate the final sound, anyways!!
This leads me to contest the dB level reliance on -20 dBs. A little saturation is very nice...
In addition, I am willing to bet that all the stages are like the above - groups, master, returns... Best of Luck all!
@mrvllous4
Marquis de J.
Thanks for this, particularly the first two. I deal with audio from multiple sources often and struggle with levels all the time. So, hearing you state the obvious gave me some guidelines. Cheers.
Great video, thanks for this information! It's funny how the same information, provided at just the right time, makes all the difference to how you understand a task.
Two point here:
1. For my workflow 48 KHz is a must. I do not know if the audio will be used in a CD, DVD, BR or on the web. Broadcast standard is 48 KHz.
2. I do not necessarily agree with the original statement that high bit rate is bad/taxing for the plug-ins. A lot of plug-ins have x2, x4 or even x8 oversampling (higher bitrates allows DSP to be more precise). Meaning that the plug-in before it processes the incoming sound, up-samples the data X times the incoming data rate. That is taxing the CPU. Imagine 100 instances of that plug-in. On the other hand if you already have high enough data rate, that process of up-sampling and down-sampling on each plug-in’s in and out does not have to happen (saving a lot of CPU processing power). If I understand it correctly, modern computers use DMA to move the data and CPU to process it. The CPU issues a command to a DMA chip to move a chunk of data from point A to point B, while the CPU is free to do other things. So the amount of data being moved is not that significant to CPU performance.
I want to point out that the anti aliasing filter probably causes the most trouble on low end gear. With a 44.1 sample rate the filter is set around 22.05 kHz and eats into the the audible range. Using 48 kHz and higher sample rate could help that shortcoming by having the filter set higher while recording. Your mileage may very depending on your gear. You could use REW to do a frequency sweep on your ADC.
Thank you so much for a great video! I will take this tips to my Sunday session in my new studio :)
I make the first mistake sometimes. I think that Trimming some of my hotter recorded signals would be beneficial. I should use the output knob on my avalon more because there is a very clear sweet spot on the gain. Thanks for this clip!
Hi dude and thanks for taking to time to educate the wonderful people of TH-cam.
I have some thoughts on the content and the message though.
1) The plugin sweet spot thing is a myth. What your talking about has no basis in DSP. The math for samples -0.1 dBFS is the same for -18dBFS. The software does not do louder, it increases or decreases a range of bits.Hence also the 'drive' knob on many modelled plugins. What you can do is overload the plugin and run out of bits or reduce the bit-rate artificially and create digital distortion. Which brings me to the next one.
2) You cannot run a digital signal 'hot'. Any signal over 0 dBFS will clip and distort, any thing under will not. Say you run a 24-bit session and have a cheaper version of pro tools, your mixer is still native fixed 48-bit. This is why you can send an overloaded signal (hence why it's yellow and not red at the top of the meter) out of channel to a buss and perform compression and not have it clip or distort coming out your master channel. It will come out as mangled as you want by your cool and calculating compressor, but your mixer won't care since you basically have 288 dBFS dynamic range. Getting crazy hear and translating this to human perceivable dBSPL, hearing a signal that loud would probably burst some parts of you and would be akin to hearing the voice of God (strictly a Dogma reference). but back to math. akmedia.digidesign.com/support/docs/48_Bit_Mixer_26688.pdf
3) Recording a higher bit rate is very beneficial for any digital signal. Look into Oversampling and why you need it and Digital Aliasing. Shit, here, I'll make it easy for you: theproaudiofiles.com/digital-audio-aliasing/. Look at is this way. You're making more snapshots of audio per second. Your right with your statement about it sounding good, and that should be your main concern- but it is not a mistake to record at higher sample rates. Say you want to pitch your voice down 12 semitones or more extreme to double or triple that. Your voice will lose all it's upper harmonics. Because of the Nyquist frequency of a higher SR you can now have more! Same thing if you wanna stretch audio. More samples to stretch and therefore a higher fidelity for a longer time with less inter sample distortion.
If someone wants to take a crack at what Ive said here, please do. Beware though that a copy-pasta from manufacturers stating they modeled "precisely after the original hardware, so it is preferable to provide input levels similar to those expected by an analogue master bus compressor", will find the word PREFERABLE in there useful food for thought and if not can expect a reply containing the words "Marketing", "Signal to Noise Ratio" and "Impulse Response".
I get your point. I worked for 25 years as an audio engineer and producer from the late 70's and saw the end of analog and the beginning of digital recording. Recording analog 'hot' signals use to mean banging the gain to saturate the master Ampex 456 multi-track tape to reduce the SNR and splat the signal onto tape as a pseudo compression effect. Worked a treat. I soon learned this was not possible working in the digital domain. Regarding recording at a high bit rate, I think the general main point being made is valid. Notwithstanding your point regarding high oversampling rates, digital aliasing and examples given the benefits are marginal compared with hammering the CPU. But as always - people have different experiences and reasons for engineering signals to achieve their aims. The key thing to remember, the whole thing is a about the music which is greater that the sum of its parts in terms of sound engineering and production.
I completely agree. Having a guy argue stuff as rational fact instead of how it makes you feel just tickles my Engineer.
Happy Holidays, and as always so much great information in your tutorials, and it's those little things. Never too late to develop better skills, Thanks Graham! Never stop making the world better through music! Stay Groovy1
Gain staging is so important. I normally mix at 48k, if my session arrives at 44k I don't touch it, there's no reason to. lots of great info thank you.
It's so sad to see people making such poor choices because of bad information being shared by well-meaning people who really don't know better. There is a very good reason why there are oversampling options in the really top notch plugins like FabFilter L2, which goes up to 32x oversampling. It's bad enough that you're starting off with low quality, but to compound it by processing and mixing it in 44.1kHz is truly a travesty.
Thank you so much for this incredibly helpful video! I've been playing music for many years and have recorded mic'd with all analog gear in the past but I'm really new to recording through a DAW and was struggling with some of the issues you've addressed here. So thank you very much!
Thanks for making knowledgeable the difference between (in gain) and (out fader) I got this knob on my DAW never figured out the purpose of this feature unless your video make it clear in one click. Thanks. Now new frontier is open..yeah just simple as that. I struggled with that problem and never had a clue to solve it.
The first two tips I absolutely agree with. The last one? Not so much. At minimum record at 48k. The file sizes aren't they much larger than 44.1 and you get some higher quality out of the recording. Most people recording at 96k (which many actually do it at 88,200 because the conversion is halved) are running HD Systems with DSP chipsets that are able to handle the processing power. All in all great video.
This was SO helpful.. Excellent advice.. 👍
please put this in the description ,, and do the same for future videos. i really like ur vids but sometimes i dont have the time to listen to the whole video.
1- Gain staging
2- mix peak -5 DB
3- work on 44.1K
Geez , So ungrateful 😂
@NoiseFeedMusic that may cause me to forget to watch it at all.. plus i always give graham a thumbs up
@@dbrain7451 bruh.. read my other reply
Elos byuri five minutes in my baby started crying.
Speed up the video then, Elos. I do that a lot on longer vids. 1.5X is good
If you're planning a video, it's worth having a sound at 48 kHz.
Absolutely..
Totally agree. I also agree with it being pointless to mix at 96k, but 48 is a good compromise for me between higher quality and relatively low CPU drain
Already had my hard time for life because of that (I work with video). Thanks!
Agreed. Mixing for video is different!
And video games too
Agreed.. only time to perhaps go beyond is when adding some saturation from those analog modeled plugins...
Studio one has become so good and cheaper plus unlimited tracks with more features, thats why i switched from pro tools to studio one
Thanks Graham, I'm glad you covered this topic. You've really helped me put some clarity and a better understanding in my mixes.
RE: Gain Staging:
1) Some plugins are more responsive to levels & level fluctuations than others. I.E. some plugins are coded to model analog gear in more depth or detail than others; as such, those plugins will be much more responsive to digital levels than others.
2) The way internal calculation resolution (word length) is handled in a given plugin is also a factor in how it deals with various levels. Plugins that are not coded very well will generally benefit from appropriate average levels more than well coded plugins (a poor implementation of an EQ plugins that has an internal calculation resolution of 32 bit floating point will do more damage to a signal than a well implemented EQ with a 64 bit float internal resolution). There are some very good designs that use 32 bit float & plugin coding continues to improve, so this is less of a worry than point 1.
3) Generally speaking, dynamics, saturation & other non-linear plugins will benefit the most from good gain staging; also many dynamics plugins require a certain range of average levels to be effective.
The best way to avoid gain problems is to record with reasonable levels through quality converters, always at 24 bit (or more if digital FX are being printed on the input).
I was going to edit the above post, but since it has been up for 11 months, it seemed better to add an addendum:
1) Some plugins are more responsive to levels & level fluctuations than others. I.E. some plugins are coded to model analog gear in more depth or detail than others; as such, those plugins will be much more responsive to CHANGES IN LEVEL than others. E.G. a well modeled emulation of hardware that includes transformer saturation (and/or tube / transistor stages, photocells, etc.) will sound best with an input signal that matches its operating range. Many UAD plugins have well defined operating ranges; they typically sound best when the incoming signal has an RMS average at or near the stated operating range.
P.S. Sampling rate is a complex subject; no single blanket statement can cover all the pitfalls and benefits of recording and/or mixing at a given sample rate. That said, here are a few typical considerations:
1) A given A/D converter may or may not sound better at a higher sample rate depending on several factors, the primary ones being the stability of the internal clock (crystal) at a given sample rate; the 'natural' sample rate of the internal crystal in non-VCXO designs (a converter designed to operate at a specific sample rate will sound best at that sample rate, or integer divisions of that sample rate, i.e. 96 kHz, 48 Khz. Some recent converters have two crystals - one for multiples of 44.1 kHZ and another for multiples of 48 kHz - the newer UA Apollo converters have two crystals AFAIK); or, if the converter is a VCXO (voltage controlled crystal oscillator), the implementation of the clocking system and/or the SRC (Sample Rate Converter); the quality of the nyquist filter can have a profound effect on the quality of the conversion (a high quality nyquist filter at 44.1 kHz is more expensive to implement than one at a higher sample rate - which is why most modern converters are 'oversampling' converters).
2) Some plugins handle lower sample rates better than others.
Two techniques for implementing a high quality plugin algorithm at 44,1 kHz is:
a) internal oversampling (the signal is upsampled at input, then downsampled at output; in the case of 'linear' processing like an EQ, this is one way to avoid unusual shapes near the nyquist frequency; in the case of 'non-linear' processing - a plugin that produces overtones - this allows for better high frequency overtone production and subsequent filtering of new frequencies above the nyquist).
b) algorithms specifically designed to minimize problems associated with processing at 44.1 kHz (EQ algos with high frequency curve compensation - sometimes called 'decramping'; saturation plugins that rely on high quality nyquist filters instead of oversampling).
There are numerous other factors that may contribute to different results when capturing and/or mixing at different sample rates.
One persistent myth about recording at higher sample rates is that capturing more 'points' will produce better fidelity. This is not the case. as long as the clocking is stable and the nyquist filter is well designed, higher frequencies will be 'smoothed out' on playback (the one arguable exception is that mixing at higher sample rates will allow for higher frequency interactions; certain high frequency interactions have the potential to influence the magnitude & phase of frequencies in the audible band, though these artifacts are typically very subtle, and in most instances would require an extended frequency playback system to produce an audible effect).
More simply put, conversion at a given sample rate is different than mixing at a given sample rate; each has its own considerations. Most of the confusion results from treating these two separate processes as the same thing (much like capturing at a given bit depth and mixing at a given word length).
What are you trying to say? Higher sampling is better or it doesn't matter?
I heard you can mix at 441 and then raise the sample rate to 96 when you print the track and boom back to 44
If you mix at a rly high sample rate you will get a better sound on frequencies below 20hz and above 20khz.
Some VST instruments are desined to work with higth SR's to avoid aliasing - some plugins got a oversampling *2/*4/.. option to let you mix with with 44/48100 samples per sec, and still avoiding the aliasing
I track at higher sample rate because of less latency. Merry Christmas
@NoiseFeedMusic He's actually right. Tracking at higher sample rates can allow for lower latency. Seems backwards but it's true.
I guess it depends on your interface. I run an apollo so I don't really worry about latency nor buffer when tracking.
@@recordingrevolution Its true, and might matter esp if you're tracking VST instruments, where any latency btw performance and monitoring can fuck up a take... I don't think a live band would benefit at all b/c they'd be zero-latency monitoring the signal before it hits converters (or just hearing themselves in the room)...
but for mixing, the latency subject doesn't matter at all, imo Who cares if there's a millisecond diff between hitting 'play' and playback, or the response of processing adjustments?
Latency is set by number of samples in the buffer, so a 32 sample buffer at 88,200 Hz is half the time long as a 32 sample buffer at 44,100 Hz. You get half the latency, but your CPU in turn gets more stressed.
You could just reduce the buffer size...
Yep, I agree. Great advice. I have definitely been recording too hot. I felt in my gut that I could record lower but I had this stigma of recording with headroom but just barely.
New sub sir.
Well i dont totally agree with the sample rate argument, i perfer higher fidelity when recording and capturing the most natural and tru to life signal. Not to mention if i pay for a interface that can give 96k then i want to use it to its maximum potential. I dont get the processor problems mixing in 96k cuz the way i record and mix i dont see a reason to dump a ton of plugings on all the tracks, maybe im a minimalist or i just get it right to begin with. I see it this way, if u need a bunch of plugins to get a great sound then you might want to learn how to properly capture a sound to avoid extra work in the mix and overloading your processor. Thats my experience and well i started back in the analog era when u had to get it right the first time, None of this ill just fix it in post crap everyone is used to now. Maybe thats why my work stands out to people cuz i take the time to get it right from the beginning and then improve on that. Remeber people in audio what u put in is what u get out, u can polish a turd but it will still be shit.
Agreed on getting the sounds right to start. If you track with a UAD interface and a good preamp and channel strip with just a bit of compression, you don't have to do much more on the track other than some EQ. The rest of the FX can go on a mix or master buss. I prefer to track at 192kHz when possible, especially when dealing with things with intense transients or really organic sounds like a cello or vocals. Even at 192kHz, I'll still use oversampling on FX that have it. Even though you're rendering to 44.1kHz, you still get a way better sound as long as you're using a really good down sampling algorithm. Which brings up a question. Rendering out of REAPER at native resolution and down sampling with r8tebrain is a lot faster, but does anyone know how it compares to REAPER's highest quality down sampling?
Everything you do with hardware before the signal goes into your DAW is definately better...BUT it takes 2 things I don't have......equipment ($$$$$$) and experience
Untrue. When recording audio, the signal is naturally and electronically filtered by the mic/preamp at around 40-48 kHz. Recording at higher than 48k is pointless, there is no signal left to record in that upper domain of the spectrum.
Great points and thanks for making them.
nice presentation. love the optics.. super job!
Shouldn't you be mixing at the sample rate of the recorded audio? If you get a session full of 96khz/24 bit audio shouldn't that be what you hand the mix as to the mastering engineer and let them deal with conversion?
You're right about sampling rate. People don't realize that it only applies to frequencies over 20Khz, the idea that it's "clearer" is a myth.Yes there is a an argument for a difference in terms of anti aliasing filters effecting frequencies below 20k at 44.1 , but especially in pop music this is immaterial compared to losing half your mixing power. I've never met someone who can tell higher sample rates in a blind test. Look at most high end pro's mixing, they're almost never using high sample rates.
And I should say that putting critical plugins in oversampling when at 44.1 (like plugins on the mix buss or groups) can help with anti aliasing filter issues...
If your planning on using that mix for a video, pretty sure you need 48kHz.....Great tutorial too. Merry Christmas Graham.