Upsampling and Aliasing.

[PaperBoat]

I have not used Fabfilter enough to know all the nitty gritty details of its use. I would think it will decide correctly when it needs to upsample a signal and then downsample it back to the original rate to do its processing. Unless the manuals or tutorials for it suggest you do upsampling, you probably are less likely to do audible harm by using its DSP for whatever EQ effects you wish while letting it decide how to get that DSP done. Pre-empively upsampling everything was once a needed thing for some processing, but that day has long passed for a good product like Fabfilter. As I recall you do get to choose linear or minimum phase in Fabfilter and which you choose will depend upon what you are doing. Just dump the upsampling on your part and life is simpler(and no less good).

Professional plugins like Fabfilter is a complex but extremely effective eq solution for consumer environment... That can't be achieved by regular EQ options. I must thank my friend who gave me the Fabfilter bundle. Yes... Upsampling isn't always a right thing and I'm not fond of heavy colouring in audio at all. But there is something exciting when you hear some good added air in your music feel the tight bass without loosing openness of the sound... Fabfilter is there for it.
I'm feeding a linear phase (high quality) Fabfilter Pro Q3 plugin to add a bit of airiness and Fabfilter Pro MB working in Dynamic Phase to clean up the low end and Fabfilter Pro L2 all without oversampling or true peak enabled. Yes, settle up those demands attention and time (I've spent approximately a year on it).

Plugins stage: Linear > Minimum > Minimum = No pre-ringing, minimal phase shift but high latency (playback doesn't care about that).

 

[fieldcar]

I think the confusion is "reconstruction or anti-aliasing filter." A reconstruction filter is an anti-imaging filter while anti-aliasing fits on the ADC end. Perhaps you meant both, but the way it read you appeared to be using two names for the same filter at the output of a DAC.

I see what you mean. Few people refer to the DAC's reconstruction filter as an anti-aliasing filter.

anti-imaging filter, I think you mean

Ah. Sorry I dismissed your last post. You're right. Imaging and aliasing are the same thing, but you had the more proper term for the artifact.

It's all the same thing though. The filter accomplishes many goals. Reconstruction and anti-imaging filter is accomplished with a filter called an anti-aliasing filter. Even in the dac stage. So I understand why I was creating confusion. My bad.

 

[restorer-john]

I asked ChatGPT the following question:

And that's where I stopped reading.

 

[PaperBoat]

And that's where I stopped reading.

But ultimately you couldn't resist yourself to replying...

 

[Blumlein 88]

I see what you mean. Few people refer to the DAC's reconstruction filter as an anti-aliasing filter.


Ah. Sorry I dismissed your last post. You're right. Imaging and aliasing are the same thing, but you had the more proper term for the artifact.

It's all the same thing though. The filter accomplishes many goals. Reconstruction and anti-imaging filter is accomplished with a filter called an anti-aliasing filter. Even in the dac stage. So I understand why I was creating confusion. My bad.

View attachment 295155

Respectfully, no. Not the same thing. The artifacts the filters are trying to avoid are different.

The issue going into an ADC is not to let frequencies above Nyquist alias downward into the below 20 khz range. Once that happens you cannot remove it. You are polluting the audible range with ultrasonic artifacts you wouldn't hear otherwise.

Anti-imaging is when sounds in the below 20 khz range image upward into the ultrasonic range coming out of the DAC. You might not hear it, but it could cause issues for gear like tweeters that may then intermodulate distortion back down into the audible range. With good anti-image filtering that will not occur.

So opposite ends of the ADC to DAC flow. Aliasing is an ADC thing, imagining is a DAC thing.

Maybe look at these results. I feel I'm not explaining things well.

Aliasing, imaging and upsampling a real example.

A couple of clarifications. Aliasing is when content above the Nyquist frequency (half the sampling rate) is folded back down into the recording by an ADC. Imaging is when content below the Nyquist frequency shows up mirror imaged above the Nyquist frequency during playback by a DAC. So what...

www.audiosciencereview.com

 

[charleski]

I'm upsampling my audio signal in JRiver to 384kHz before applying some minimum phase DSP (Linear Fabfilter Pro Q3, Dynamic/Minimum phase Fabfilter Pro MB and no oversampling Fabfilter Pro L2) and sending the signal to my IFI Zen V2 (7.4C GTO filter firmware) to avoid dithering noise generation in the DAC. The DAC is connected with a pair of JBL 305p MKii monitor. I know the phase is getting altered but my goal is to achieve pre-ringing and distortion free sound. Is it a right approach?

If you were just running straight EQ with the Q3 (i.e. not using any ‘dynamic’ EQ settings) then you wouldn’t gain any benefit by upsampling* as the transform would be linear (as @KSTR mentioned above). But the compressor and limiter will introduce harmonics and you want to make sure these don’t fold down into the audible range, so upsampling will be useful. But the Fabfilter plugins are technically pretty good and actually incorporate oversampling within their architecture (unlike some other plugins which can cause aliasing). So you can either oversample before the plugins and turn off their internal oversampling, or just leave the signal as-is and let the plugin handle the oversampling and filtering. Personally, I’d tend to go for the latter, as it involves less useless data being thrown around between plugins in your chain.

*I know there was a long debate in another thread on whether there’s a technical difference between ‘upsampling’ and ‘oversampling’, but I switched off when trying to follow it, sorry.

 

[PaperBoat]

plugin handle the oversampling

As far as I know, if I let Fabfilter Pro MB and Pro L2 to handle the oversampling then they oversample the signal before applying modification then downsamples it again which is still a confusing thing for me.

 

[pma]

Reconstruction and anti-imaging filter is accomplished with a filter called an anti-aliasing filter.

Sometimes we may see it named this way, however, it confuses the issue. We should say anti-aliasing filter for A/D conversion and anti-imaging filter for D/A conversion.

 

[charleski]

they oversample the signal before applying modification then downsample again

Yes, that’s what they do. It means if you start out with a 44.1kHz signal that’s what you end up with. But the internal oversampling and filtering means that all the garbage ultrasonic harmonics caused by the compression have been safely thrown away. Which is what you want.

 

[PaperBoat]

Yes, that’s what they do. It means if you start out with a 44.1kHz signal that’s what you end up with. But the internal oversampling and filtering means that all the garbage ultrasonic harmonics caused by the compression have been safely thrown away. Which is what you want.

I still don't understand. To avoid aliasing they oversample the signal that is understandable but when the signal gets downsampled it's actually inviting distortion which is undesirable.

 

[pma]

but when the signal gets downsampled it's actually inviting distortion

No. Why do you think so, what is your reasoning? Any links, literature?

 

[PaperBoat]

No. Why do you think so, what is your reasoning? Any links, literature?

downsampling adds distortion via a reduction in the sampling rate of your audio. While you technically lose audio quality, you gain a ringing, edgy distorted sound.

 

[restorer-john]

@PaperBoat I suggest buying a few books on digital audio. Actual published, physical books. Stay off Google and ChatGPT- go straight to the horse's mouth- the guys who actually wrote the books...

These are two I keep referring back to in my collection:

 

[voodooless]

downsampling adds distortion via a reduction in the sampling rate of your audio. While you technically lose audio quality, you gain a ringing, edgy distorted sound.

I'd say: citation needed!

 

[restorer-john]

downsampling adds distortion via a reduction in the sampling rate of your audio. While you technically lose audio quality, you gain a ringing, edgy distorted sound.

The link you are showing is for deliberate distorted effects, in that case by reducing the available data to the point it distorts... Nothing to see here.

 

[Blumlein 88]

downsampling adds distortion via a reduction in the sampling rate of your audio. While you technically lose audio quality, you gain a ringing, edgy distorted sound.

As said, this is intentional distortion. They are downsampling (probably to very low rates) to intentionally distort sound and to make filter ringing occur at a frequency we can clearly here. In playback without downsampling below 44.1 khz you are not getting that. Very much I would suggest letting Fabfilter take care of any up or downsampling. You don't need to worry about it and simply need to EQ as suits you letting the software handle it all for you.

 

[SSS]

@PaperBoat I suggest buying a few books on digital audio. Actual published, physical books. Stay off Google and ChatGPT- go straight to the horse's mouth- the guys who actually wrote the books...

These are two I keep referring back to in my collection:View attachment 295164View attachment 295165View attachment 295166View attachment 295167View attachment 295168View attachment 295169

Well done. I appreciate it. All this was already described since the CD was developed. DSP exists since around 1960 for military and medical applications. Many electronic professional measurement instruments use DSP since 1970ies. And for this there exist technical books describing Nyquist/Shannon theorem, LP-filter, Sample&Hold, and the AD-converter. Nothing new today except that the processor performance is now able to do realtime filtering and manipulations which was limited to low audio frequencies in the early days.

 

[PaperBoat]

@PaperBoat I suggest buying a few books on digital audio. Actual published, physical books. Stay off Google and ChatGPT- go straight to the horse's mouth- the guys who actually wrote the books...

These are two I keep referring back to in my collection:View attachment 295164View attachment 295165View attachment 295166View attachment 295167View attachment 295168View attachment 295169

That's truely helpful. Thanks...

The link you are showing is for deliberate distorted effects, in that case by reducing the available data to the point it distorts... Nothing to see here.

In my poor understanding: 48kHz upsampled to 384kHz downsampled to 48kHz = distortion. Am I right? Please correct me.

 

[PaperBoat]

As said, this is intentional distortion. They are downsampling (probably to very low rates) to intentionally distort sound and to make filter ringing occur at a frequency we can clearly here. In playback without downsampling below 44.1 khz you are not getting that. Very much I would suggest letting Fabfilter take care of any up or downsampling. You don't need to worry about it and simply need to EQ as suits you letting the software handle it all for you.

I'll try Fabfilter Pro MB 4x and Pro L2 8x oversampling.

 

[Blumlein 88]

That's truely helpful. Thanks...




In my poor understanding: 48kHz upsampled to 384kHz downsampled to 48kHz = distortion. Am I right? Please correct me.

No that is not correct. If done perfectly, there is no difference. What real world difference there is will be so low as to be meaningless. However, the software will do it and present you with a good 48 khz version. For instance some processing software for some DSP might sample to 512 khz, do the DSP and downsample. The main takaway is don't try and make those decisions on a case by case basis. Just use the software and quit obsessing over this.