96kHz Mixing is overkill. So why am I doing it?
Вставка
- Опубліковано 20 тра 2024
- ➡️➡️Join the HPS Discord:- / discord
➡️➡️ Support us on Patreon! - / hoppolestudios
➡️➡️ Check out our ULTIMATE GUIDE TO REAPER with Pro Mix Academy:- promixacademy.com/course/the-...
Mixing at 96kHz vs 48 kHz is a whole lot extra, and does it actually have any tangible benefit? Today we talk about the difference it makes, and whether it's relevant for you.
🔌🔌What gear to buy? We recommend interfaces, microphones, monitors and more at our Thomann (Europe) and Sweetwater (USA) links below
www.thomann.de/gb/thlpg_loh2r...
imp.i114863.net/DVNrad
🍻 BUY US A BEER WITH PATREON!🍻
/ hoppolestudios
Thanks to our Patrons who made this possible - your names are at the end of the video. You make a huge difference!
📱 FOLLOW US HERE 📱
Facebook: / adamsteelproducer
Instagram: / adamsteelproducer
All links are affiliate links
0:00 - Intro
5:26 - Nyquist Frequency and Aliasing
15:45 - Practical Demonstration
19:00 - Aliasing begone
So what sample rate are you using? Will you stick or move based on the whole aliasing thing?
Before I understood how whittaker-shannon interpolation worked, I did 96k every time. Now I record at 48k, and work with oversampled plugins, or move to 96k if I need/want to use a plugin that doesn't anti alias and then print back down to 48k.
That’s not a bad plan, I do find (depending on algorithms used) Oversampling can affect things in a way I’m not a fan of, but I’ve also done the Oversampling thing depending on needs. Reaper can now do that natively per plugin or per chain so it’s a real rabbit hole to dive down…
96kHz is the way forward metinks. Plugins work better, you need less processing in my experience, reverbs sound way better and loads more depth. (compared to 44.1 anyway)
I have used Virtual Instruments and Samples quite a bit and as soon as the major players started with 96khz so did I. At 48khz there were sometimes strange things happening, but only when I look at the signal like you did in this example. It wasn't as clean as it should be. But I couldn't say it affected the final version so you could tell yea this was mixed at 48khz and this one at 96khz. One other reason I switched to 96khz is resampling. Going from 48khz to 44.1khz is mostly black magic. With 96khz you have much more data so that all those maths can be more precise.
I stay at 48 because I am still running Behringer ADATs and they only do 48k. I dont always need them for mixes but like having them available. If I upgrade them ever I probably will go to 96.
What a bang up explanation. Can't believe I just discovered you.
As somebody who used to work in professional studios, getting back on the horse can be a daunting task. The knowledge I had to retain to put out a quality product for the artists I worked with came with years of learning and experience.
You, sir, have made getting back on the horse that much easier. Endlessly appreciate you, and will be enrolling in your Reaper class soon.
Cheers to you! Keep up the good work! Subscriber earned and I wish you many more!
thanks for this very informative video and not only did you explain everything very clearly or just say about of things you feel, you showed a very good example with the decapitator plugin. 👍🏼
In the analog world, when you put 2 frequencies together, you have 2 frequencies and nothing else when the analog circuitry is perfectly distorted. It's physics at this point.
Take those two same frequencies and add some non-linearity to the mix, such as in a tube amplifier with slight overdrive. This alters the signal and produces harmonics which is multiples of the frequency in odd order harmonics. With a single sine wave with distortion you have the sine wave and odd order harmonics. With 2 frequencies you have non linear mixing which produces some other frequencies. in addition to the harmonics, you produce the Sum frequency, and Difference frequency. This is in the analog world. This is perectly replicated in the digital world, except now a 3rd frequency is added to the input, the sample frequency and it's Aliasing frequencies. Note, there is no mention of if the mixed frequencies are above or below the sampling frequency.
Everyone does the due diligence with the nyquist frequency. The 3rd harmonic of 15 KHZ is 48KHZ.
When sampling at 96KHZ frequencies 32khz would produce the 3rd harmonic at 96KHZ, but 32 KHZ is much easier to filter out than frequencies arround 16KHZ. It's just the math. Yes most band mixes have very little energy in the 16KHZ frequencies and above.
Recordings with high energy in high frequencies and high harmonic content are poor on CD. I first noticed this issue with Pink Floyd's Dark Side of the Moon album on the Money Track. Having the analog pressed recording from studio tape, and then getting the CD, the coins sound completely different. You can't miss the amount of unwante frequencies muddy what used to be very crisp sounds of coins ringing. If you can find an analog recording, note the very stark differences. One is clear crisp sounds of the coins. The other is a jumbled mess of noise.
There is more to the Aliasing than just the Nyquist frequency. It is the harmonics and the sample frequency in the digital world. Yes you can filter the harmonics going in, but digitally add some non-linearity, and the problem is returned after the low pass filter.
Thanks for all this info, really clear a lot of things. 👍🏻
Your advice, gleaned from your vast experience, was very helpful. Thank you very much🙏
Still using 48 Khz - 24 bit. For >3 dB EQ moves I prefer hardware. For smaller ones software is just fine. Most of the aliasing and foldback distortion is at a very low level. What 500 module is working on your voice in the background?
I’ve listened to the first 4 minutes of this and I haven’t a clue what you’re on about. I play hi res music at 96khz and it sounds a lot better than vinyl and cd’s
The thing is, that theoretical physics aside, anything recorded, processed, mixed, mastered, and played back at 96/24 will sound better to a practised listener . And this isn’t just some sort of imagined placebo effect either; it’s been shown to be the case in meta-studies of audiophiles and sound engineers.
So why is that? Well… while it’s certainly true that 48/24 provides all the resolution any human could ever need at any volume they are able to withstand, there’s a little known engineering issue that comes into play: there is no commercially viable let alone available DAC that can easily resolve 48KHz (or 44.1KHz for that matter). This is actually due to a quirk of physics where ripples in the bandpass filter introduce distortion at lower sample rates. Hence we’re better off converting back to analogue at 96KHz.
Thanks Adam. Some of the older free plugins I've been playing with clap out at 96k. Never mind, If I wan't to try this I can download the great plugins and misbehave like lots of others. 😄
Thanks Adam
Thankyou for this vital info! So if I have a 48 rate sample based vst (many of kontakts) but in ableton i set it to 96 sample rate. Would this help improve fuzziness (alaiasing) of some of the sounds i am experiencing? I wish I came across this link before buying all these crusty low quality plugins 😢 lo fi artist would probably love it LOL
I'm here because recently Colt Capperrune did a video on 48k vs 96k. Ultimately, the core of his argument was that 96k sounds more "accurate", "realistic" and "clinical" (not in a good way), is more CPU intensive and requires more storage space. At one point he made somewhat of a comparison of 48k having more "vibe", "aggression", "crunchiness", and then comparing that to how recording to tape colors things. Anyway. The CPU and storage requirements are legit, factual reasons for using a lower sampling rate, I get it. But, the tone/sound argument is completely subjective. I've been recording and mixing @96k for awhile and really only did it as a way to future proof, but, maybe that's wrong. Anyway -- I guess I'd have to record and mix something myself at both sampling rates and compare to see if the tone argument holds water - he might be right. And even then, it would be my opinion vs his or someone elses that feels the same way. Good info here -- thanks.
Its strange but this is exactly why i have just moved to 96k. For all of the same reasons youve illustrated, and why its a kind of future proofing against the aliasing issue . Bravo for speaking your mind Adam. Its the only way to be!
thanks a lot
I record my first E.P. with different sample for some of the take from 44.1 or 48-then edited and mixed in 48-export/transfer into the master to 44.1 because of the majority of take recorded in 44.1-release it in digital platform and does have that artificial digital artifact sound of resampling and desampling
As soon as I found this video I follow your advice and transfer the original mix into 96khz 24bit (mind you, before it's also 16bit)
Thanks
interesting, since the covid lockdown went away early 2022. I mixed a lot
of theater and festival shows on "older" systems (48K) and the new consoles (96K).
When putting heavy processing on channels, like gate/dynamic eq/limiter/6-band parametric
(over the top of course) and some long reverbs, I always found the 96K mixers better sounding
and more "open". I always thought it was in my own mind and my old ears. You just confirmed
that my gut feeling was sort of right. What about latency difference between 48K and 96K?
Doing lots of inear mixes on a 48K console always makes me feel uncomfortable with the result.
Its not to the extend of being unable to perform, but like 46 channels into 10 stereo mixes seems
always "cluttered" to me. Any thoughts on that? When tracking, latency in the headphone mixes
seems pretty important to me. Keep up the good work, Adam.
Thanks!
As for latency, that and sample rate are not linked. Latency is DSP/converter dependent, and I think we will see that go down as processing power goes up.
I think a lot of people confuse the two because they change sample rates and the total system latency goes down by default. Why? Because by default we ask for the same number of samples as a buffer, and they go by twice as fast so reported latency is lower. However system performance is degraded because the DSP has half the time to process the same samples! How to fix it? Double the sample buffer! You end up in exactly the same latency as before (or close enough depending on internal converter latency) with roughly equal system stablilty.
@@adamsteelproducer I run at 96khz for lower input latency, and it works much better than using a smaller sample block at 48khz. But aside from that I see no real reason to use higher than 48khz + oversampling for any non-linear processing. Every volume slider would suck up twice or more cpu cycles.
i agree, also omnisphere plug in sounds way different at 48khz compared to 96khz.. so the sounds of the instruments change at different sample rates depending on the plug in. plus at 96khz everything sounds more detailed if you are producing
Thanks Adam. This is helpful stuff, recently joined team 96K. It is worth it.... Question now is... How do I know if I have a good resampling algorithm in my DAW? Also... Do you find you're losing anything by downsampling at the end? For delivery of a 48K master?
did you have any luck in your research mate?
@@user-ls7xf4lk6t yes.... I use Cubase and he SRC algorithm proved to be high quality. I used a sine wave sweep to test. with regards to downsampling for final master, I find you do not lose any quality as long as you have a good anti aliasing filter in place. protect the audible range at all costs.
@@user-ls7xf4lk6tI'd like to know this as well.
@@user-ls7xf4lk6tI'd like to know this as well.
@@user-ls7xf4lk6tI'd like to know this as well.
Hi Adam, have you tried using AMD cpus for your mixing/recording computer? Do they work okay with plugins? I saw some plugin manufacturers that they only mention Intel CPUs. Currently looking at them now cause they have good benchmark and is more affordable for me than Intel. Thanks and will appreciate any tips. Cheers!
Hi Edward, both of my mix PCs use AMD CPUs. They’re based on licensed tech from Intel, so don’t worry about that. They’re all great.
@@adamsteelproducer thank you so much Adam! Do you think a Ryzen 9 5900X is a good cpu for 96k mixing + acustica plugins? Will appreciate any advice.
Yeah I use a 5950X and it’s overkill, a 5900K should be just fine unless you’re using hundreds of acustica plugins, in which case you’re on your own financing a supercomputer!
@@adamsteelproducer thank you so much Adam! Really appreciate the help. Im going with this cpu then 🙏
48k has to do with the fact that there is then an even number of samples per frame of video.
I wonder if there is a filter to build to add to the chain before the computer interface channels so the negative effect doesn’t or occur at all, then 48 would be fine
Modern interfaces have anti-aliasing filters in them for this exact reason. there’s no way to filter out what will be processed after recording though
It is more efficient to oversample just what you need to oversample than to run everything at a higher sample rate. Upsampling and downsampling just for non-linear processes is always going to be more efficient than running at 96khz throughout the entire project. Even gain plugins will be sucking up unnecessary resources.
For a lot of people, yeah.... That's true. There's quite a few factors that AI add up that can make 96k make more sense for some people. Too much to gab about to explain it easily. For a lot of people 48k with Oversampling is going to make the most sense.
@@ramspencer5492i still think that it only makes sense to have higher sample rates for general usage if you are going to do a lot of massive time stretching. Although i would argue that 96khz is not even enough for that. Or if you are using plugins that cramp and stuff. But that is working around bad tools.
so if the original mules were recorded at 24b 28Khz and you drag them into protools they then have to be upscaled to work in a 24b 96k session yes. So its not as straight forward as your pointing out. Track at 96Khz and then mix at 96Khz would obviously be the best way but you didn't mention the original tracks bit rate or sampling Freq
Nope. 44100 Hz is the sample rate the same between PAL and NTSC (Black and white - 60 fields, not 59.94) encoders. 48 KHz is a multiple of 24 (fps in film).
So if we run mixes at 96KHz, are you saying that we wouldn’t need to do any oversampling in any plugins?
That’s the idea yes. Why oversample up to 96k when you’re already there?
@@adamsteelproducer I suppose so. Do you also think that 2x Oversampling running at 48k is more cpu intensive then running those same tracks and same plugins at 96k without oversampling?
@@kablah19 most plugins that oversample these days provide anti-aliasing filters which are often more effective than running at a higher sample rate alone. you can still alias at 96khz. not to mention that running every plugin at 96khz is going to be way more cpu intensive than just oversampling what is needed to be oversampled. even a gain plugin will be working harder than it needs to.
As soon as you have any processing that stretches and contracts audio (chorus and especially pitch/timing correction), things sound terrible really quick with 44.1kHz because the time data is just not there. Even 48kHz is a big improvement and pitch shifting full notes, heavy chorus or leslie effects suddenly sound great.
Note that many audio interfaces sound way depending on whether the sampling rate is a multiple of 44.1 or 48 kHz. Depends on the native clock rate, but usually the native clock is based on 48 and it just rounds samples to fit in the 44.1 grid. So 192kHz tends to sound fantastic compared to 44.1kHz but 48kHz sounds just as good as 192.
I wish you could mix and match sample rates on the fly so some tracks record on 48 and others at 192 if there's gonna be a lot of processing.
Do you know what omnispheres sample rate is?
There are some
Videos on UA-cam showing that up and down sampling have absolutely no effect on sound. 500 passes still null test to zero. This is why your plugins can up and down sample all day long
I understand the thinking behind this decision but honestly it’s pretty much a non issue, as long as your aware of it and oversampling on your plugins where necessary, for instance if your saturating a kick drum there’s no point in oversampling as unless your creating an actual square wave it’s now going to reach nyquist but when saturating something like cymbals then it’s a good idea to oversample.
Any decent modern oversampling algorithm is for all intents and purposes transparent, if you have the computer resources and power to run everything at 96 then go ahead but I think it’s really paranoia, I’d recommend the Dan Worrall “sampling rate: the higher the better?” As it goes over all of this.
I'm also switching to 96k. Now that I've finally gotten a computer that can handle it.
Simply said, higher sample rates gives you a higher spectral head room to avoid aliasing.
I suppose that’s a succinct way of putting it! I think it still warrants the length of video to try and explain what that actually means to most people
Using high samplerates is literally the most ineffective way of almost solving an issue that doesn't matter.
every plugin worth using implements AA
Thats obviously! Take amp simulation and check so8nd of guitar with and without oversampling. Huuuge different
So I understand plugs will up/down sample to 96k which is a lot of conversion/artifacts. I also understand how 48/44.1 can represent the Nyquist freq.
My question is: why do folks in the 44.1/48 simply audio repro. As a single freq. up to 20k/22k etc…when there can be multiple tracks representing these freqs. At different phases and further more the nuance and air of a track falling between samples-it seems that these will be less likely to be captured. How does they account for this? To me, altho it’s a burden to my cpu 96 sounds silkier, but maybe I’m trippin’. I think it’s especially noticeable in quiet live recorded material.
Would using a analog eq prevent this problem..?
Technically yes, there’s no aliasing in analog. Phase shift still happens though, but that’s a different issue
why would anyone shelf at 20kHz ? a system should not worry about invalid inputs
I’m not sure you understand. Not only was I using that number simply to highlight the issue, but any filter has a nominal frequency but affects far more either side of that, so a high shelf frequency can be used to give a gentle curve below without adding resonances
@@adamsteelproducer you are correct but the amount of gain that you have to apply to a 20k shelf to hear the artefacts is an unreasonable amount
or at any frequency for that matter
Still just an example. It’s more audible in saturation/distortion but harder to explain that way
@@adamsteelproducer so tell me if i am correct, your final argument is "i use 96 because some plugins can't handle aliasing"
My attention span is against me on this one...
Because your mad. In a good way.
@3:13 fart
😂💨
Bump your plugins in reaper to 192. Thank me later.
My electric company would thank you, but I don’t see a big difference on most plugins. Don’t do anything blindly with no context
@@adamsteelproducer well. All the aliasing is gone. Totally gone. I’ve been doing this for months and as owner of so much hardware. It’s making more sense why the late Al worked at 192.
I didn’t say the sessions need to be at 192 just bump up ( oversample ) to 192. I still track at 48k.