Upsampling and Aliasing.

[charleski]

Here's a figure from Ken Pohlman's book that shows what's actually happening with oversampling, note that filtering out the unwanted ultrasonic images is an integral part of the process. After the processing the data is then downsampled in a similar fashion, again applying the proper filter to get rid of the unwanted images, along with the ultrasonic junk caused by the processing.

 

[AnalogSteph]

In the context of digital effects, upsampling can be useful since running them at a higher sample rate will make nonlinear effects (e.g. compressors or amp simulators) behave more like their analog counterparts. A prime example would be brickwall limiters, which in the olden days (before about 2012) were quite notorious for introducing nasty aliasing artifacts. Basically, nonlinear operations in the digital domain are not inherently band-limited, and frequency components generated past fs/2 will "wrap around" and end up somewhere in the audible spectrum. The higher fs, the more the out-of-band components will already have decayed (as they generally do go down with frequency) and the more they are spread out. Even a factor of 2 tends to help a lot.

 

[PaperBoat]

In the context of digital effects, upsampling can be useful since running them at a higher sample rate will make nonlinear effects (e.g. compressors or amp simulators) behave more like their analog counterpart

That's why I was trying to upsample the signal before EQing but the implementation of SOX resampler is linear in JRiver which is not what I'm prefer.

A prime example would be brickwall limiters, which in the olden days (before about 2012) were quite notorious for introducing nasty aliasing artifacts.

I've setup Fabfilter Pro L2 as a brickwall limiter.

 

[restorer-john]

In the context of digital effects, upsampling can be useful since running them at a higher sample rate will make nonlinear effects (e.g. compressors or amp simulators) behave more like their analog counterparts. A prime example would be brickwall limiters, which in the olden days (before about 2012) were quite notorious for introducing nasty aliasing artifacts. Basically, nonlinear operations in the digital domain are not inherently band-limited, and frequency components generated past fs/2 will "wrap around" and end up somewhere in the audible spectrum. The higher fs, the more the out-of-band components will already have decayed (as they generally do go down with frequency) and the more they are spread out. Even a factor of 2 tends to help a lot.

Yes, but please clarify a brickwall limiter is not a brickwall filter (LPF).

 

[PaperBoat]

Here's a figure from Ken Pohlman's book that shows what's actually happening with oversampling, note that filtering out the unwanted ultrasonic images is an integral part of the process. After the processing the data is then downsampled in a similar fashion, again applying the proper filter to get rid of the unwanted images, along with the ultrasonic junk caused by the processing.
View attachment 295173

My grey cells are now busy learning it...

 

[AnalogSteph]

That's why I was trying to upsample the signal before EQing but the implementation of SOX resampler is linear in JRiver which is not what I'm prefer.

What's wrong with that? I can't imagine that this is a realtime application where the higher latency of FIR filters might matter?

 

[Hayabusa]

Why? It's destroying the bit-perfect recording that you probably started with. Run your DAC in exclusive, bit-perfect mode instead.

One reason could be the DAC has a not optimal upsampling itself and doing it in software could improve that..

 

[danadam]

What's wrong with that?

The boogeyman of ringing, I'm guessing

 

[charleski]

the implementation of SOX resampler is linear in JRiver which is not what I'm prefer.

It makes sense to use minimum-phase for filters in the audible band, though the ringing only really becomes audible if using a steep filter slope. But anti-imaging filters are happening right up at 22kHz (the ringing is actually a chirp with the max amplitude at the filter frequency) and the ringing is effectively inaudible. This is why all the hoo-haa over different DAC filters is largely a waste of time (as long as they attenuate properly).

In fact, I once ended up using a maximum phase filter (with all the ringing preceeding the impulse) because I hadn't read the SOX documentation properly. I couldn't hear the difference when I switched it to linear- or minimum-phase.

 

[krabapple]

[nm, redundant delete] You all drank my milkshake. ;>

 

[voodooless]

The boogeyman of ringing, I'm guessing

Ding, ding, ding!

Clearly he skipped my link to Archimago

 

[krabapple]

I hear ringing all the time! It's called tinnitus.

I blame my DAC.

 

[fieldcar]

Aliasing is an ADC thing, imagining is a DAC thing.

https://sdg-master.com/lesestoff/aes97ny.pdf

 

[voodooless]

https://sdg-master.com/lesestoff/aes97ny.pdf

View attachment 295259

That’s just wrong, see https://xiengineering.com/sampling-frequency-audio-aliasing/

 

[krabapple]

ooh, I can pull up an AES paper too! By Julian Dunn no less. From 1998.

https://www.nanophon.com/audio/antialia.pdf

Anti-alias and anti-image filtering: The benefits of 96kHz sampling rate formats for those who cannot hear above 20kHz.

Anti-alias filters are used to limit the bandwidth of a signal being input to a sampled
system, such as a digital audio system, so that it is less than one half the system
sample rate.
.
.
.
In the reverse process, anti-imaging (or reconstruction) filters are used at the output
of a sampled system to eliminate spectral images of the desired signal.

Your dCS guy is using lazy terminology. You've been thoroughly corrected on this. I suggest accepting that and moving on.

 

[PaperBoat]

Wow! Look... what I've found...

what-is-the-point-of-upsampling.

 

[voodooless]

Wow! Look... what I've found...

what-is-the-point-of-upsampling.

So? What conclusion do you think you can derive from this?

 

[fieldcar]

That’s just wrong

I disagree. What happens with a Sigma-Delta DAC without a reconstruction filter? You get an aliased voltage * time domain signal resembling stair-steps. Thus, an anti-aliasing filter is still correct. I do not dispute that it's called an anti-imaging or reconstruction filter, but it's all the same thing.

 

[voodooless]

I do not dispute that it's called an anti-imaging or reconstruction filter, but it's all the same thing.

No, it’s not the same thing. They are reverses of each other. Two sides of the same coin, but still different sides.

As also detailed by wiki:

Anti-aliasing filters are used at the input of an analog-to-digital converter. Similar filters are used as reconstruction filters at the output of a digital-to-analog converter. In the latter case, the filter prevents imaging, the reverse process of aliasing where in-band frequencies are mirrored out of band.

Emphasis is mine..

 

[fieldcar]

Two sides of the same coin, but still different sides.

Heh. I was just going to say the same thing.

Again. How is a stair-stepped ZOH signal from a DAC not the exact definition of an aliased signal?