This was so well produced and presented. Thank you for taking the time to do this. I have been using high sample rates to avoid this artifact with my analog model plugins for years. Harmonic saturation sounds great at high sample rates.
A - D - X are oversampled IMO. This is my first answer listening to from my tv. I turn on the monitors and I realised the opposite situation! Great video!
Big thanks for this video! It was pretty easy to hear the oversampled ones (B, C, Y), However, since I have been hoarding the "analog modeling" plugins recently, this is a very good reminder about the actual sound quality that can be achieved. Thank you 🙏
Nice challenge and very educational explanations. Thanks! Much to my surprise, I guessed all three examples correctly. I was focusing on clarity in the high frequencies which might fall under your "smooth & extended highs". I'm not sure what to take away from this, though. Obviously, oversampling does make a different if all other plug-in parameters are fixed. However, I wonder if the perceived extra clarity in the oversampled version could also be achieved by tweaking other parameters without oversampling. But I guess there's no objective way to argue either way and this is where the discussion starts getting religious.
BCY, totally audiolble on 10$ headphones. More oversampling - less high freq for me in this test. But high freq almost always anotal, aliasing atonal - no problem here at all
What can explain the drastic tonal shifts between the OS and non-OS version? I find it VERY hard to believe that oversampling would cause a 12 decibel difference in the fundamental of the low tom hit at 16:22. I totally understand you're talking about some difference in brightness and depth, but a 12 db difference in the floor tom's fundamental? That's the difference I measured between the tom hits! If oversampling was causing a 12 db discrepancy at certain frequencies, we would be hearing way more about it. Any ideas what's up here?
No idea. I'm not here to convince you or anyone they should or shouldn't OS, please conduct your own experiments. I did this across several older mixing sessions and personally, I'm convinced it's worth the effort. In all cases, I saw very similar improvements. I OS'd each chain in the recording and I noticed this tonal effect with the low-end was accentuated much more than shown here. In some chains, the OS added delay and flipped phase. I then printed each chain and slide them back into time as per the original audio. I also checked that the overall phase of each chain was the same. Again, the effects were still the same, though slightly less so.
@@SPLMixing I do agree with you. I try to use plugin oversampling where possible, but definitely on any kind of distortion/saturation. I think your video was super well done and it's an awesome and seamless comparison between the two, especially in regards to the difference in brightness/top end sizzle. I would just be concerned, personally, if a plugin's oversampling was causing a 12 db difference in low-end material. That seems more to me like a coding error and I would be really hesitant to praise it as an oversampling benefit, especially when there are plenty of instances where 12 db difference would destroy the tonal balance of the audio you're processing. (It also makes me wonder which plugin of yours was the biggest culprit in that 12db difference and if it's talked about in their documentation!) At the end of the day, absolutely conduct your own experiments. Everyone should become familiar with their plugin OS options and be aware of the benefits it can provide. Thanks for the video!
Thanks 👍 I wish I had an answer as to why. Armed in the knowledge of what plugins need it, I will mix that way now and react to the audio accordingly which means I will compensate for any tonal changes in real-time.
I got all of them right, but I also had a good understanding of what it is before watching this video. I listened to which mix sounded more buzzy. Because Aliasing can cause additional overtones because they are reflections of the overtones above the Nyquist frequency, it can sound a bit like distortion.
BB Tubes by Waves just came out (see my latest video). Blackbox by Brainworx (I think). Some other Waves plugins also have it built in but not accessible to turn on or off. Most of the Fab Filter stuff. Most likely Sonible stuff to some degree. Metaplugin can turn anything into an oversampled plugin though and it's cheap. Remember it's only relevant to plugins that are nonlinear and produce extra harmonics.
It would, if you were 2x upsampling a 48kHz session, but the strain on cpu resources increases significantly, plus your session requires double the hard drive space. Consider also that linear processes like digital eq do not produce any harmonics, so upsampling the whole session has no real benefit in those cases. Better to oversample on a per plugin basis instead IMHO. You probably won't see major differences until you're oversampling 4 or 8 times. At 48kHz, that's an equivalent session sample rate of 192kHz or 384kHz respectively. Try running a full mix session at that! 🙂
Yes, but I believe it also increases quantisation errors because of the math involved with crunching the larger file sizes. Larger files from increased sample sizes are impractical and tax your CPU and memory. Since not all plugins are non-linear and don't need oversampling, you're better off doing it only on what needs it on a case by case basis instead.
The oversampled ones sound so much more open. Like there’s room to put your feet up in the mix and stretch out. A, D and X sound more cluttered. Not bad but I definitely prefer B C and Y. Great tests!
I only got the last one correct 😅 I personally never cared much about aliasing as long as it's barely audible (~80dbFS). Softube and PA, for example, rarely have OS option. But if any plugins have OS, I tend to use 2x for track / channel and 4x on mixbus.
@@OfficialGOTCHA33 Okay, after some investigation, I'm going to say "No", you cannot properly null test this, as every oversampling processor I've used appears to phase-shift the extreme lows and highs, but also (critically) slightly across all frequencies. This means a null test is out of the question if you want to completely extract only the aliasing portion of frequencies from the original. You can, however, run standard sine sweep from 20Hz to 20kHz through a plugin (like I did in the video) and hear the effect.
If you enable oversampling, the upper range content can be filtered out before it producing aliasing. Oversampling is not looking to add more highs. it's looking to reduce aliasing significantly.
Thank you very much for the explanation, the best I have seen on the subject to date. There you talk about mixing at a higher sample rate. Do you recommend recording at 44,100 and oversampling the plugins/session, or recording at a higher sampling rate? Or would this bring no benefit and just more processing and storage usage? Thanks!
It really depends on the strength of your cpu and storage capacity. I'm partial to recording at 88.2 kHz and working at that sample rate. My PC can handle it, so long as I don't get too crazy with plugins. As far as investigating the potential invreased benefits of working with higher sample rates versus the equivalent oversampled audio, this is something that I plan to test out in a future video. No doubt it negates some of the aliasing artefacts, but if you lack cpu power and space, then oversampling the plugins that need it most is your best strategy.
my initial prediction was A- oversampled D- oversampled X- oversampled curious to see if my guesses are right or not but in all of my answers for oversampled the transients felt slightly softer I would assume from the FIR linear phase oversampling that DDMF uses, and in the mixes i guessed were not oversampled I thought I could hear some aliasing in the guitar in particular.... time to see if i'm full of shit or not!
UAD sound great, but having only used them in other studios, I'd suspect that they might include some form of oversampling. Then again, I've not formally tested them to know for sure. Many plugins still sound great and still exhibit aliasing when put under the microscope. It's the really badly designed older plugins that you want to watch out for. Many of the Soundtoys plugins need a good update, for example.
Difficult to say, as I don't know what your laptop speakers sound like. You most likely will not hear a difference without first hearing it on a decent set of headphones or monitor speakers.
It's very situational dependant tbh. Some times aliasing can be so insignificant nobody would be able to hear it. Where as in another instance it might be so obvious literally everyone could hear it.
@@SPLMixing Yeah for sure. I was mixing the other day and noticed aliasing from the saturation in the channel strip plug in I had recently started using. I hadn't noticed it up until then as I hadn't pushed the saturation so hard. Going back with it in mind to listen for aliasing specifically, I picked it up subtly on a few other tracks too, ones with less saturation. However there were still a few you just couldn't really hear anything on as the saturation applied was really low. With the one that I initially noticed it on I was also using subtle frequency modulation on (as it was a really long dead bass note that needed some living up). This set off all the funkiness with intermodulation effects that aliasing can have when things are moving around. This made it so much more apparent it was there, although I would of heard it anyway as it was bad regardless, just it made it stick out even more. It's a shame the plugin doesn't give options for OS. It was a really really handy plug in otherwise, and actually has really nice saturation. ;( I haven't tested it in 96Khz projects yet, this one happened to be sent to me in 44.1K, so I was working in that, I normally work in 96K, so maybe it's a plug in I can use with a higher sample rate, I need to test it.
@@SPLMixing Yeah I've heard about this before a few times. I've actually been meaning to check it out, but have not gotten round to it. It's either that or move to reaper, but I'm not quite ready for that kind of commitment just yet (throwing away 20 years of experiance using cubase) so using DDMF meta seems like a good solution. 👍
A,d and y sounded oversampled to me, i believed i could hear non musical harmonics in snares fx etc had more of a fuzzy noise layer over it where as the saturation in the d and y versions especially, sounded like harmonics were integrated and musical and therefore clearer mixes... Im being honest here cos i think it would be hilarious if im completely wrong lol ....... didnt do too well in the end lol 😂 its my ears or maybe i can get away from this bruising by blaming my headphones lol 😆
@@SPLMixing thats what i thought too thanks, your video sent me on a learning binge to try and minimise the build up of aliasing in my mixes, a lot of the plugins i use often seem to have oversampling built in so i use that, hopefully until my ears can hear it better at least i can keep aliasing as low as possible. Thanks for the reply 👍
Wow, I got them all right! :D I guess all those months comparing plugins that have oversampling with different ammounts of it really paid off (being a mastering engineer helped too, my ears are sensitive to the highs). Btw, just like how you said, limiting and saturation are greatly affected by Aliasing. But there's one kind of plugin you didn't mentione that distorts like crazy: The clippers. If you've tried different clippers, you know that some of them have options for crazy ammounts of oversampling (like 256x).
Yes, I agree, also seems to have a space around each instrument too. Metaplugin just got updated to include 16x oversampling and a whopping 64x oversampling in offline bounce mode!
@@SPLMixing Is that a fact? I've wanted to buy that for a while but didn't want to spend the $50. Now I'm thinking about getting it soon. I heard a guy say that he uses AirWindows tape plugins and runs it through Meta with a lot of oversampling and he says it will blow your mind if you try it. I demoed meta and did this but didn't hear much difference other than a spike in CPU usage so I was not impressed with that. I think it's really only feasible to use it on tracks you can immediately bounce and commit to. I have some plugins that just running them on 2x oversampling uses up a ton of CPU. I don't have a lot of CPU but I have a decent amount.
The issue with this is not all plugins have a non-linear behaviour and therefore require oversampling, so you're upsampling the whole mix for a small percentage of plugins that do. Second, you're only giving your mix the equivalent of 2 x oversampling (@48 kHz sample rate), whereas upsampling goes a lot higher than this. Third, the increase in file sizes for all files could tax your cpu, and potentially lead to computational errors, halt your system, and take up loads more space than necessary.
This is the best explanation I have found on what is happening with both aliasing and oversampling. The comparison is really helpful. Thank you!
This was so well produced and presented. Thank you for taking the time to do this. I have been using high sample rates to avoid this artifact with my analog model plugins for years. Harmonic saturation sounds great at high sample rates.
Excellent explanation, thank you!
A - D - X are oversampled IMO.
This is my first answer listening to from my tv. I turn on the monitors and I realised the opposite situation! Great video!
the short example was very clear!!!...thanks for the video dude!
Big thanks for this video! It was pretty easy to hear the oversampled ones (B, C, Y), However, since I have been hoarding the "analog modeling" plugins recently, this is a very good reminder about the actual sound quality that can be achieved. Thank you 🙏
You're welcome
BCY I’ve got them all right. 🎉
I focused primarily on that sharp high end. And harshness in vocal.
Awesome explanation Daniel! keep it up! I will listen carefuly with headphones and my new monitor if I can hear it. cheers friend!
perfect! now i know everything i needed.
Nice challenge and very educational explanations. Thanks! Much to my surprise, I guessed all three examples correctly. I was focusing on clarity in the high frequencies which might fall under your "smooth & extended highs".
I'm not sure what to take away from this, though. Obviously, oversampling does make a different if all other plug-in parameters are fixed. However, I wonder if the perceived extra clarity in the oversampled version could also be achieved by tweaking other parameters without oversampling. But I guess there's no objective way to argue either way and this is where the discussion starts getting religious.
Clarity and space around instruments. Less crud clogging up parts of the mix.
Happy that I managed to guess all three correctly and it was not as subtle as I expected even! Now I am a fan of oversampling all the way!
Great work! And here I was thinking everyone would complain they couldn't hear a difference 😅 I think it's hard to look past the benefits.
B,C, and Y do not use oversampling. The midrange is clearer to me. Oversampling sounds a bit compressed.
BCY, totally audiolble on 10$ headphones.
More oversampling - less high freq for me in this test. But high freq almost always anotal, aliasing atonal - no problem here at all
is there any plug in that cleans and balance the signal as such?
DDMF Metaplugin running with oversampling turned on.
@@SPLMixing thanks!!:)
@@SPLMixingits best to put it after all vst chain on mixer channel i suppose?
That allows you to chain multiple VST plugins and oversample them.
What can explain the drastic tonal shifts between the OS and non-OS version? I find it VERY hard to believe that oversampling would cause a 12 decibel difference in the fundamental of the low tom hit at 16:22. I totally understand you're talking about some difference in brightness and depth, but a 12 db difference in the floor tom's fundamental? That's the difference I measured between the tom hits! If oversampling was causing a 12 db discrepancy at certain frequencies, we would be hearing way more about it. Any ideas what's up here?
No idea. I'm not here to convince you or anyone they should or shouldn't OS, please conduct your own experiments. I did this across several older mixing sessions and personally, I'm convinced it's worth the effort. In all cases, I saw very similar improvements. I OS'd each chain in the recording and I noticed this tonal effect with the low-end was accentuated much more than shown here. In some chains, the OS added delay and flipped phase. I then printed each chain and slide them back into time as per the original audio. I also checked that the overall phase of each chain was the same. Again, the effects were still the same, though slightly less so.
@@SPLMixing I do agree with you. I try to use plugin oversampling where possible, but definitely on any kind of distortion/saturation. I think your video was super well done and it's an awesome and seamless comparison between the two, especially in regards to the difference in brightness/top end sizzle. I would just be concerned, personally, if a plugin's oversampling was causing a 12 db difference in low-end material. That seems more to me like a coding error and I would be really hesitant to praise it as an oversampling benefit, especially when there are plenty of instances where 12 db difference would destroy the tonal balance of the audio you're processing. (It also makes me wonder which plugin of yours was the biggest culprit in that 12db difference and if it's talked about in their documentation!)
At the end of the day, absolutely conduct your own experiments. Everyone should become familiar with their plugin OS options and be aware of the benefits it can provide. Thanks for the video!
Thanks 👍 I wish I had an answer as to why. Armed in the knowledge of what plugins need it, I will mix that way now and react to the audio accordingly which means I will compensate for any tonal changes in real-time.
I got all of them right, but I also had a good understanding of what it is before watching this video. I listened to which mix sounded more buzzy. Because Aliasing can cause additional overtones because they are reflections of the overtones above the Nyquist frequency, it can sound a bit like distortion.
Nicely done 👍
I got all of them incorrect lol. So I knew there was a difference. I just assumed the brighter one was the aliasing lol.
what are some of the best plugins to use that have oversampling under the hood or as an option?
BB Tubes by Waves just came out (see my latest video). Blackbox by Brainworx (I think). Some other Waves plugins also have it built in but not accessible to turn on or off. Most of the Fab Filter stuff. Most likely Sonible stuff to some degree. Metaplugin can turn anything into an oversampled plugin though and it's cheap. Remember it's only relevant to plugins that are nonlinear and produce extra harmonics.
@@SPLMixing Wouldn’t running your session at 96K have the same affect as over sampling?
It would, if you were 2x upsampling a 48kHz session, but the strain on cpu resources increases significantly, plus your session requires double the hard drive space. Consider also that linear processes like digital eq do not produce any harmonics, so upsampling the whole session has no real benefit in those cases. Better to oversample on a per plugin basis instead IMHO. You probably won't see major differences until you're oversampling 4 or 8 times. At 48kHz, that's an equivalent session sample rate of 192kHz or 384kHz respectively. Try running a full mix session at that! 🙂
Relevant also for equalizers and filters where cramping happens near Nyquist. For modern and HQ tools this problem mostly solved
Don t really know how well oversampling works in practice, but won´t using very high sampling rates solve these fold back problems?
Yes, but I believe it also increases quantisation errors because of the math involved with crunching the larger file sizes. Larger files from increased sample sizes are impractical and tax your CPU and memory. Since not all plugins are non-linear and don't need oversampling, you're better off doing it only on what needs it on a case by case basis instead.
@@SPLMixing right, i get you . Thank you for the clarification ;)
The oversampled ones sound so much more open. Like there’s room to put your feet up in the mix and stretch out. A, D and X sound more cluttered. Not bad but I definitely prefer B C and Y. Great tests!
Killer video
I only got the last one correct 😅 I personally never cared much about aliasing as long as it's barely audible (~80dbFS). Softube and PA, for example, rarely have OS option. But if any plugins have OS, I tend to use 2x for track / channel and 4x on mixbus.
Metaplugin by DDMF does a good job of this up to 64x offline OS
A/ C and Y
Is it possible to phase reverse both signals to hear the aliasing at work?
I believe it's possible, but it may be more complicated than a simple null test, so I'll have to get back to you on that one.
@@SPLMixing thanks for getting back to me! Really interesting stuff!
@@OfficialGOTCHA33 Okay, after some investigation, I'm going to say "No", you cannot properly null test this, as every oversampling processor I've used appears to phase-shift the extreme lows and highs, but also (critically) slightly across all frequencies. This means a null test is out of the question if you want to completely extract only the aliasing portion of frequencies from the original. You can, however, run standard sine sweep from 20Hz to 20kHz through a plugin (like I did in the video) and hear the effect.
Hi. I don’t understand what’s the point of enabling oversampling if the project is still rendered at 44.1 khz
If you enable oversampling, the upper range content can be filtered out before it producing aliasing. Oversampling is not looking to add more highs. it's looking to reduce aliasing significantly.
Thank you very much for the explanation, the best I have seen on the subject to date. There you talk about mixing at a higher sample rate. Do you recommend recording at 44,100 and oversampling the plugins/session, or recording at a higher sampling rate?
Or would this bring no benefit and just more processing and storage usage?
Thanks!
It really depends on the strength of your cpu and storage capacity. I'm partial to recording at 88.2 kHz and working at that sample rate. My PC can handle it, so long as I don't get too crazy with plugins. As far as investigating the potential invreased benefits of working with higher sample rates versus the equivalent oversampled audio, this is something that I plan to test out in a future video. No doubt it negates some of the aliasing artefacts, but if you lack cpu power and space, then oversampling the plugins that need it most is your best strategy.
Before the reveal, I hear distortion on the low end on an and d. On AirPod
my initial prediction was
A- oversampled
D- oversampled
X- oversampled
curious to see if my guesses are right or not but in all of my answers for oversampled the transients felt slightly softer I would assume from the FIR linear phase oversampling that DDMF uses, and in the mixes i guessed were not oversampled I thought I could hear some aliasing in the guitar in particular.... time to see if i'm full of shit or not!
funny i could tell the difference but picked the non-over sampled ones!
Are the differences between X & Y lower cause i got the first 2 right but the last example seemed to be subtler to me
The differences may be more/less subtle on different speakers. The aliasing is subject to multiple factors, such as what is happening at any one time.
What do you think about UAD plugins, do they use high oversampling to better preserve all harmonics?! Hugs and beautiful video!
UAD sound great, but having only used them in other studios, I'd suspect that they might include some form of oversampling. Then again, I've not formally tested them to know for sure. Many plugins still sound great and still exhibit aliasing when put under the microscope. It's the really badly designed older plugins that you want to watch out for. Many of the Soundtoys plugins need a good update, for example.
I got every one backwards (wrong)
B,C,Y?
Bcy. But I can only BARELY tell in the cymbals
would the difference be apparent if Im listening on my laptop speakers?
Difficult to say, as I don't know what your laptop speakers sound like. You most likely will not hear a difference without first hearing it on a decent set of headphones or monitor speakers.
@@SPLMixing Very basic macbook air speakers, probably cut off at around 100-120 Hertz. Great video anyways!!
I failed miserably! Thank you for the vid!!
Yes
It's very situational dependant tbh. Some times aliasing can be so insignificant nobody would be able to hear it. Where as in another instance it might be so obvious literally everyone could hear it.
Possibly. I think with training, it becomes easier to hear, too.
@@SPLMixing Yeah for sure. I was mixing the other day and noticed aliasing from the saturation in the channel strip plug in I had recently started using. I hadn't noticed it up until then as I hadn't pushed the saturation so hard. Going back with it in mind to listen for aliasing specifically, I picked it up subtly on a few other tracks too, ones with less saturation. However there were still a few you just couldn't really hear anything on as the saturation applied was really low.
With the one that I initially noticed it on I was also using subtle frequency modulation on (as it was a really long dead bass note that needed some living up). This set off all the funkiness with intermodulation effects that aliasing can have when things are moving around. This made it so much more apparent it was there, although I would of heard it anyway as it was bad regardless, just it made it stick out even more.
It's a shame the plugin doesn't give options for OS. It was a really really handy plug in otherwise, and actually has really nice saturation. ;(
I haven't tested it in 96Khz projects yet, this one happened to be sent to me in 44.1K, so I was working in that, I normally work in 96K, so maybe it's a plug in I can use with a higher sample rate, I need to test it.
DDMF meta plugin is a good workaround when OS is not an option on the original plugin
@@SPLMixing Yeah I've heard about this before a few times. I've actually been meaning to check it out, but have not gotten round to it. It's either that or move to reaper, but I'm not quite ready for that kind of commitment just yet (throwing away 20 years of experiance using cubase) so using DDMF meta seems like a good solution. 👍
A,d and y sounded oversampled to me, i believed i could hear non musical harmonics in snares fx etc had more of a fuzzy noise layer over it where as the saturation in the d and y versions especially, sounded like harmonics were integrated and musical and therefore clearer mixes... Im being honest here cos i think it would be hilarious if im completely wrong lol ....... didnt do too well in the end lol 😂 its my ears or maybe i can get away from this bruising by blaming my headphones lol 😆
It's a sound we're already accustomed to hearing, I guess. It takes training to hear the differences. Highly recommend DDMF metaplugin
@@SPLMixing thats what i thought too thanks, your video sent me on a learning binge to try and minimise the build up of aliasing in my mixes, a lot of the plugins i use often seem to have oversampling built in so i use that, hopefully until my ears can hear it better at least i can keep aliasing as low as possible. Thanks for the reply 👍
I got it all correct!
It's really easy to hear even through youtube, you can hear more 3d depth and fullness of bass and more natural less harsh sound
Excellent work 👍
Wow....I picked them out with my phone audio....no joke
Not through your phone speakers. That would be a lie. Or you just got immensely lucky
@@INeedsMoneys I picked all 3 correctly very easily from my phone speakers as well. Not a lie.
Seems like the hi hat really gives it away. In the non oversampled it sounds a bit washy and resonant.
I couldn't pay attention because I was listening to the riff in The background
B,C,Y
Edit: yay!
B/C/Y
1) B.
2) C.
3) Y.
Wow, I got them all right! :D
I guess all those months comparing plugins that have oversampling with different ammounts of it really paid off (being a mastering engineer helped too, my ears are sensitive to the highs).
Btw, just like how you said, limiting and saturation are greatly affected by Aliasing. But there's one kind of plugin you didn't mentione that distorts like crazy: The clippers.
If you've tried different clippers, you know that some of them have options for crazy ammounts of oversampling (like 256x).
Yeah, right, interesting! The closest thing I use to clipping is the analogue setting on the Waves API 2500 Compressor. Thanks for the tip!
Also, nice work getting all answers right
the over-sampled has more "life" in it.
Yes, I agree, also seems to have a space around each instrument too. Metaplugin just got updated to include 16x oversampling and a whopping 64x oversampling in offline bounce mode!
@@SPLMixing Is that a fact? I've wanted to buy that for a while but didn't want to spend the $50. Now I'm thinking about getting it soon. I heard a guy say that he uses AirWindows tape plugins and runs it through Meta with a lot of oversampling and he says it will blow your mind if you try it. I demoed meta and did this but didn't hear much difference other than a spike in CPU usage so I was not impressed with that. I think it's really only feasible to use it on tracks you can immediately bounce and commit to. I have some plugins that just running them on 2x oversampling uses up a ton of CPU. I don't have a lot of CPU but I have a decent amount.
Bruh , I chose A, D, X 😭
what about set a project to 96khz to avoid this problem without using oversampling?.....thank god i use analog hardware most of the time
The issue with this is not all plugins have a non-linear behaviour and therefore require oversampling, so you're upsampling the whole mix for a small percentage of plugins that do. Second, you're only giving your mix the equivalent of 2 x oversampling (@48 kHz sample rate), whereas upsampling goes a lot higher than this. Third, the increase in file sizes for all files could tax your cpu, and potentially lead to computational errors, halt your system, and take up loads more space than necessary.
Rip I got 0/3