Hi all

I want to know something simple:

For sampling (cd) 44.1 kHz the nyquist is fs/2.
If I Ask audirvana to up-sampling for example at 88.2 kHz…does it means that Nyquist increase to 44.1 kHz ?
That also means bandwitch is closed to 44khz?

Thanks for your help

When a signal is sampled at 44.1 anything above 22.05 kHz is filtered and gone. This can not be restored by upsampling.

Hi Jacob

No any harmonics from 22.05?

And if the original files IS a HD files like 96khz 24 bits.
The bandwidth IS 48 kHz…
Can i Can limit this Large bandwidth to 22khz?
What s happens into my tweeter ?
This IS energy for nothing from my amplifier.


The upsampling does create new information/energy. But this is not information recovery.

There is not much energy in the harmonics above the hearing threshold. I wouldn’t worry about that.

Yes… the Nyquist of the new file is increased to approximately 44.14kHz, however, since the 16/44.14kHz A/D encoding has a brick-wall filter at approximately 22kHz there is little to no high-frequency energy beyond the Nyquist cut-off, depending on the nature of the filter applied in the encoding process. (The up-sampling process inserts zero’s (0’s) to create the dynamic-range and resolution, but does not interpolate new data-points. (synthesize new code). Unless an interpolation algorithm is applied that does contextually synthesize new code between the original sample-points.

The salient reason for up-sampling is to increase the dynamic-range of the original file and to increase the resolution of the digital signal so it appears more like a pure analog signal being presented to the DAC output circuitry… The increased dynamic-range exposes low-level energy that was captured in the encoding process that is masked by the 16/44.14kHz encoding noise.

:notes: :eye: :headphones: :eye: :notes:

1 Like

Nothing is going to happen to your tweeter unless you don’t filter the low-frequencies below it’s resonant frequency… The nature of most high-frequency transducers today, is that they will roll-off high-frequency energy quite steeply beyond their designed response capabilities… If you can get 30kHz from your tweeter at a decent level, you are doing pretty well… :wink:

1 Like

One reason to upsample is to take the burden of upsampling from the DAC which may or may not improve the sound quality… Another reason is that ultrasonics beyond the human hearing range may be felt by the body in other ways and ultrasonics interact with sound that is within the hearing range… These sensitivities are often described as air and sparkle… And there are plenty of high-frequency transducers that will go beyond 22kHz…

1 Like


Thanks a lot

Upsampling does not create new information. The reason for upsampling, which started back before there were even such things as separate DACs, is to make it easier to design good filters.

With all filters there is a tradeoff between a couple of different kinds of distortion. One is when the filter does not cut quickly enough, so ultrasonics leak through. For a 44.1k sample rate, these ultrasonics aren’t music, they’re distortion, and they can intermodulate with other frequencies to become audible noise. Another type of distortion is when the filter cuts too quickly, which causes what’s commonly called “ringing.” (Really this is the Gibbs Effect, which you can read about in Wikipedia if you like.) Without upsampling, the fine line between cutting too quickly and not cutting quickly enough is very difficult to achieve. Upsampling moves the noise into a higher ultrasonic range, making it easier to cut off this noise without causing a lot of ringing.


Might be an interesting paper on the subject… https://www.tnt-audio.com/casse/life_above_20khz.pdf

Hi all

Someone Try to have a look one a 1 kHz square wave on an oscilloscope to see the influence of up-sampling settings (sox or r8brain)?

It is not an answer to the above question, but it is interesting to see the results of various Sample Rate Converters. In this case from 96kHz to 44.1kHz. Various tests and SRCs can be selected.


“1 kHz square wave” is a label that contains a lot more information than it appears to.

What’s that mean? It means that unlike the familiar sine waves, square waves contain loads of very high frequency harmonics, well into the ultrasonic range. (That’s what causes the rise and fall times of the wave to be so fast, making it look squared off.) So what you’re seeing is performance pretty far up into ultrasonics, not just at 1 kHz.


I look an square wave at 1khz ans 10khz (96khz 44.1) I was surprise to see the Bad quality of the signal.
I Try many UP sampling settings with audirvana.
I Never see a perfect signal on the output of my Teac UD 505 X.

You don’t mean a sine wave?

Square waves won’t start to look good until you’re at least double the frequency of the highest harmonic component. So let’s say we’re talking about 13th harmonic of 20kHz because square waves incorporate virtually all audible frequencies and harmonics. That’s 260kHz, so you may need a rate above 520kHz for good looking square waves, at a very rough guess.

Edit: Oh, and also - The analog reconstruction filter in your DAC will filter frequencies above anywhere from 50kHz to 150kHz, so square waves at the output won’t look great. You want to see the results of oversampling at the computer output instead.

Thanks jud
It os more clear now

Thats correct Jud.

You want to see the results of oversampling at the computer output instead…

But how is it possible ?

Sorry to say I’ve never bothered to get test equipment myself, so I can’t tell you off the top of my head. I guess some reading and research is necessary, though I don’t think it will be too difficult.