saturation/distortion with excellent anti-aliasing

[Gam456]

I found that 2x or 4x minimal phase oversampling typically works best. And if you are working in a high sampling rate (eg 88.2 khz and up) you will probably not need oversampling.

Indeed. A good exemple the distortion on the Gforce Bass Station.
Runing my Motu at 48khz, aliasing is pretty bad.
Between 96khz and 192 khz everything is gone.

It's a trade-off. Working at 96khz and up has the issue of a build up of ultrasonics/IMD. Pick your poison.

You can have intermodulation distortion or not even at 48khz, depend on the plugin and the way you use the distortion, insert, parallel, bus etc...

[_leras]

.. and here in lies the crux. Even if you can't hear the effects of aliasing when doing a/b comparisons with a single plugin on a single track, it's like a really nasty form of noise that gets compounded and concentrated at the frequencies where our ears are the most sensitive.

No oversampling

Oversampling

The issue isn't, and never was, if you can "hear it" with one single plugin. The issue is when it is compounded. You very easily end up with a lot of extra uncorrelated "false" harmonic information which just happens to live happily in the 3 to 12kHz region, where our ear absolutely hates that kind of shit.

I'm not sure what you are doing here, but yes you've achieved some extra, harsher distortion, on an already distorted sound.

In the real world do you have this problem?

I use tons of distortion, from different vendors, and it never does that. I don't get a harsh edge that I have noticed.

I'm also not hearing that in the tons of music I listen to. I just don't think it a real world problem.

[_leras]

Another tip: use saturation very sparingly on high frequency material. I use it mostly on low to mid frequency instruments (kicks, bass, sometimes rimshots or even sometimes snares; not so much on hihats, cymbals or high pitch synths)

Nah, that's just a taste thing. Many saturators will also roll of the top end anyway, softening these types of sounds while adding distortion.

[bmanic]

.. and here in lies the crux. Even if you can't hear the effects of aliasing when doing a/b comparisons with a single plugin on a single track, it's like a really nasty form of noise that gets compounded and concentrated at the frequencies where our ears are the most sensitive.

No oversampling

Oversampling

The issue isn't, and never was, if you can "hear it" with one single plugin. The issue is when it is compounded. You very easily end up with a lot of extra uncorrelated "false" harmonic information which just happens to live happily in the 3 to 12kHz region, where our ear absolutely hates that kind of shit.

I'm not sure what you are doing here, but yes you've achieved some extra, harsher distortion, on an already distorted sound.

In the real world do you have this problem?

I use tons of distortion, from different vendors, and it never does that. I don't get a harsh edge that I have noticed.

I'm also not hearing that in the tons of music I listen to. I just don't think it a real world problem.

Listen to the tonality, the "frequency response" differences in that example. There are zero eq changes in those two files, yet one is much brighter/harsher. That tonality and subtle transient change is caused 100% by aliasing alone.

THAT was the point. It matters when it is compounded.

[El°HYM]

I found that 2x or 4x minimal phase oversampling typically works best. And if you are working in a high sampling rate (eg 88.2 khz and up) you will probably not need oversampling.

Indeed. A good exemple the distortion on the Gforce Bass Station.
Runing my Motu at 48khz, aliasing is pretty bad.
Between 96khz and 192 khz everything is gone.

It's a trade-off. Working at 96khz and up has the issue of a build up of ultrasonics/IMD. Pick your poison.

You can have intermodulation distortion or not even at 48khz, depend on the plugin and the way you use the distortion, insert, parallel, bus etc...

This is where TDR - Ultrasonic can help.
https://www.tokyodawn.net/tdr-ultrasonic

In the end of the day, there is actually no perfect solution for all of this, as too many variables involved. What works well for one thing, wont necessarily work as good on the next one.

[Ah_Dziz]

Melda waveshapper can go up to 16X resampling

single band version is free, multiband version is 40$ but often go on sale for 20$

https://www.meldaproduction.com/MWaveSh ... m0VF07fbCY

All the melda stuff has oversampling, the freebies have OS and a few other features turned off.

[kraster]

The absolute best quality distortion pedal plugin is, without question, The Scream.

It's hard to overstate how good the scream is.

[Ah_Dziz]

I've always found that simply not using distortion willy nilly all over my mixes is the best move. Trying to stear clear of things that try to sound "analog" while doing something rather simple like EQ or compression by adding distortion can add up to a mess of unwanted frequencies caused by aliasing once you have a ton of tracks going.

On the other hand, if I'm going for a distorted sound, I'm happy to lean into the aliasing if the processing creates it. There are very nice anti aliasing algorithms implemented in various ways. Unfortunately, to get all the aliasing up out of the audible range generally takes a good bit of OS. People have the "use heavy OS on export" option, but I generally try and get my stuff to sound how I want it while I'm actively working on it so I don't want things to just sound different on export.

[El°HYM]

Some very interesting tests with different sampling rates and filter - slopes etc. - might be a bit mind-boggling for some, how high you actually need to go to fully push aliasing down into the in-audible.

https://dibiquadroaudio.com/oversampling-vs-aliasing

[_leras]

Listen to the tonality, the "frequency response" differences in that example. There are zero eq changes in those two files, yet one is much brighter/harsher. That tonality and subtle transient change is caused 100% by aliasing alone.

THAT was the point. It matters when it is compounded.

Yeah, I heard it. I'm just not sure what you had to do to get to that harshness though.

Is it multiple decapitators? For me decapitator usually rolls of the top end, though I'm not normally pushing the highs with... Their other distortions devil loc and radiator also don't seem particularly harsh. So how you get to it I'm not sure.

I don't get that kind of harshness in any mixes, and I don't particularly pay that much attention to what oversampling settings a plug I used. Hence my question of, is this really a real world issue?

[Andreya_Autumn]

I've always found that simply not using distortion willy nilly all over my mixes is the best move. Trying to stear clear of things that try to sound "analog" while doing something rather simple like EQ or compression by adding distortion can add up to a mess of unwanted frequencies caused by aliasing once you have a ton of tracks going.

Good advice. Like every one-sentence mix wisdom, there's exceptions. But yeah, using linear processing for your baseline EQ/compression and such (and then using a dedicated plugin for any distortion/saturation needs) lessens the compounding problem that Bmanic is talking about.

[_leras]

Yeah, I just don't think this is a real world problem.

Are we tried to say saturation is 'more' prone to this than other types of FX?

I will admit I'm not usually stacking up multiple distortions on a single channel, I don't generally have it on buses, but I do use it on the master buss, sometimes in parallel. I just don't get this harshness that's being discussed. I do get a lot of lovely mainly saturated, and sometime distorted sounds. But it's only harsh if thats what I've specifically dialled in and the plug in gives me.

I would say perhaps I'm lucky and my saturation plug ins are all all over sampled but I'm not sure, and I do use sound toys fairly regularly..

[Gam456]

Are we tried to say saturation is 'more' prone to this than other types of FX?

It's not only about saturation. Any gain stage introduce distortion such eq, compressor

[bmanic]

Listen to the tonality, the "frequency response" differences in that example. There are zero eq changes in those two files, yet one is much brighter/harsher. That tonality and subtle transient change is caused 100% by aliasing alone.

THAT was the point. It matters when it is compounded.

Yeah, I heard it. I'm just not sure what you had to do to get to that harshness though.

Is it multiple decapitators? For me decapitator usually rolls of the top end, though I'm not normally pushing the highs with... Their other distortions devil loc and radiator also don't seem particularly harsh. So how you get to it I'm not sure.

I don't get that kind of harshness in any mixes, and I don't particularly pay that much attention to what oversampling settings a plug I used. Hence my question of, is this really a real world issue?

If I remember correctly it was one instance of decapitator, one SieQ, one Devil Loc and one Radiator. All moderate to hard processing. The point was to show how much simple 4x oversampling can help in mitigating aliasing, to show the way the tonality completely changes when you have multiple badly aliasing DSP processes in series.

It is not at all uncommon to end up with at least 4 or 5 non-linear plugins in series, on multiple tracks. Just by having a quite common chain of Track -> Bus -> Mixbus, you already likely end up with 3 plugins in series. If you now add mastering to this you have 4. If you had a "parallel crush/compression" bus then you easily end up with a whole lot more (depending on where you tap the send).

I'm trying to show with this example just how quickly the nasty stuff adds together.. and once you hear that stuff, it's going to be hard to unhear it. Any kind of modern mix with 70 to 100 tracks, especially if it's a pop mix with tons of vocals, will immediately be very noticeable if you go from 44.1kHz to higher sampling rates if you use plugins that don't have built in oversampling. The difference is not at all subtle. These differences will be further highlighted once you do final processing on the audio.

Having said that, if the mix was already perfect at 44.1kHz then the change of sampling rate (or switching on oversampling) can actually be worse as you will be losing a lot of brightness and "aliasing glue" (it is sort of uncorrelated noise after all).

So yes, people who know what they are doing and are good at mixing don't have to really worry about this because they will auto-magically mitigate it by using/abusing frequency masking (knowingly or unknowingly!). Thus ironically, people who don't know what they are doing and don't have any training in critical listening are the ones that would benefit the most as it gives the final mastering session much more leeway. A mix that is chock full of aliasing is extremely difficult to get good sounding at the mastering stage. This is why really muddy and seemingly bad analogue mixes can be almost completely rescued in the mastering stage, with minimal compromise, whereas the same is pretty much impossible to pull of if the mix is entirely digital and contains tons of aliasing.

Are we tried to say saturation is 'more' prone to this than other types of FX?

Any gain stage introduce distortion such eq, compressor

Only a gain stage that is being changed per sample.. so if the gain change is "moving". If it is 100% static, like an ordinary static digital EQ, then it causes zero harmonics.

[bmanic]

Are we tried to say saturation is 'more' prone to this than other types of FX?

No, it's ALL non-linear processes. Dynamics, Saturation, all "emulation" plugins etc. etc. Pretty much every single plugin that causes some form of movement to audio. So almost everything in today's world of emulation crazyness, including your emulated choruses, flangers, phasers, delays, reverbs, auto-panners etc. It's also not always harmonic distortion. It can be inharmonic too (intermodulation distortion for instance) and it'll still fold back and cause added foldback noise in the critical 3 to 12kHz region. It just won't be as annoying as the inharmonic saturation is quite nasty to begin with.

Saturation, harmonic or inharmonic is everywhere. This is the big advantage of analogue.

The only absolutely 100% sure not to have any aliasing at all are completely digital EQ plugins when they are static. As soon as you use a dynamic EQ you get tons of added harmonics.

Heck, even automating the volume in your DAW, especially if done very quickly over large values causes harmonic distortion which in turn can cause aliasing (this also very much depends on how well the DAW is smoothing the data).