What is the point of upsampling?

[Lambda]

There is no value in upsampling.

Upsampling/ Oversampling, Tomato Tomahto.

Oversampling - Wikipedia

en.wikipedia.org

Almost all good DACs oversample internally. so it not only has value it is essential for the function. (of digital filtering)

An idealized electronic filter with full transmission in the pass band, complete attenuation in the stop band, and abrupt transitions is known colloquially as a "brick-wall filter" (in reference to the shape of the transfer function). The sinc filter is a brick-wall low-pass filter,

It is an "ideal" low-pass filter in the frequency sense, perfectly passing low frequencies, perfectly cutting high frequencies; and thus may be considered to be a brick-wall filter.

Real-time filters can only approximate this ideal, since an ideal sinc filter (a.k.a. rectangular filter) is non-causal and has an infinite delay

As the sinc filter has infinite impulse response in both positive and negative time directions, it must be approximated for real-world (non-abstract) applications; a windowed sinc filter is often used instead. Windowing and truncating a sinc filter kernel in order to use it on any practical real world data set reduces its ideal properties.

Brick-wall filters that run in realtime are not physically realizable as they have infinite latency (i.e., its compact support in the frequency domain forces its time response not to have compact support meaning that it is ever-lasting) and infinite order

So there is an "infinite delay" but who has time to wait for infinite in real time...
if up sampling happens in the PC the real time limitation is no longer relevant better and user definable filter can be used.
You don't have to care about the internal filter options from your DAC you can effectively apply them in software

If not for some synthetic Aliasing and inter sample over test signal you will not hear a difference!
But you you will be able to measure an improvement.

And if "we" care about finde detilese like this:

Why don't we control them...

 

[Sokel]

The point I see in upsampling is DAC-specific and doesn't have to do much at how high but to what family of freqs we upsample (44.1 or 48Khz).
I see all the time that some DACs don't handle each "family" the same (same for DSD,they measure VERY differently) and they are better on one of them (call me clock).
So it's nice to know where it performs better and set it accordingly.
Needless to say that the ones that handle the 44.1Khz family better are the ones to prefer,as all music is there.

 

[DonH56]

Upsampling is converting a lower sampling rate to a higher rate. Oversampling is sampling a signal at a rate greater than Nyquist (twice the signal bandwidth). They are not the same thing. Both schemes have the potential for better image rejection but there are fundamental differences between them.

 

[Lambda]

Waht stops you from calling Upsampling Oversampling, upsamplig is also sampling a signal at a rate greater than Nyquist.
I don't see how they are fundamental different? They are essentially the same?

Essentially New sampling points get generated and digital filter makes sure they are below the below the Nyquist frequency.

Edit: ah you mean Oversampling is only for recording?
In this case it is semantics.

So the DAC is (internally for signal processing and filtering) Upsampling the output signal or is it Oversampling the input signal?
for me its the same

https://www.analog.com/media/en/training-seminars/tutorials/mt-017.pdf

They also call it oversampling and interpolatio in the context of DACs

The basic concept of an oversampling/interpolating DAC is shown in Figure 2. The N-bit words
of input data are received at a rate of fc. The digital interpolation filter is clocked at an
oversampling frequency of Kf c, and inserts the extra data points. The effects on the output
frequency spectrum are shown in Figure 2. In the Nyquist case (A), the requirements on the
analog anti-imaging filter can be quite severe. By oversampling and interpolating, the
requirements on the filter are greatly relaxed as shown in (B).

So oversampling/interpolating ( a digital) signal results in an upsampled signal?

 

[Julf]

Edit: ah you mean Oversampling is only for recording?
In this case it is semantics.

No, it is not just semantics. In upsampling you are restricted to the information you have in the original waveform - you just generate more data points by interpolation, but theyu contain no new information, and you don't increase precision. In oversampling you gather more information than you need (so at a higher time resolution).

 

[Lambda]

In upsampling you are restricted to the information you have in the original waveform

Same is true for a "oversampling" a digital signal
You say yourselves upsampling and "waveform"
the digital signal is a theoretical "mathematical" waveform this "waveform" can digitally/mathematically reconstructed and Oversampled.

he concept of oversampling and interpolation can be used in a similar
manner with a reconstruction DAC. For instance, oversampling is common in digital audio CD
players, where the basic update rate of the data from the CD is 44.1 kSPS. Early CD players used
traditional binary DACs and inserted "zeros"

https://www.analog.com/media/en/training-seminars/tutorials/mt-017.pdf

They make DACs and they call it oversampling?

 

[Julf]

Same is true for a "oversampling" a digital signal

No, it isn't.

You say yourselves upsampling and "waveform"
the digital signal is a theoretical "mathematical" waveform this "waveform" can digitally/mathematically reconstructed and Oversampled.

But what you are "oversampling" isn't a higher resolution waveform. It doesn't contain any new information - unlike the case of real oversampling.

 

[voodooless]

But what you are "oversampling" isn't a higher resolution waveform. It doesn't contain any new information - unlike the case of real oversampling.

So the claim is that oversampling does contain new information? How strange.

The way I always saw it this way: oversampling is a form of synchronous upsampling used in DACs to make analog reconstruction filters much simpler. Upsampling itself is just the conversion of one sample rate to a higher one. That can be either synchronous or asynchronous.

 

[Julf]

So the claim is that oversampling does contain new information? How strange.

The way I always saw it this way: oversampling is a form of synchronous upsampling used in DACs to make analog reconstruction filters much simpler. Upsampling itself is just the conversion of one sample rate to a higher one. That can be either synchronous or asynchronous.

Not sure what the difference between oversampling and upsampling is in your view, as you are talking of sample rate conversion in both cases.

Oversampling is sampling of the original signal at a sample rate that is higher than needed (so that it can be downsampled later, after filtering). Upsampling is oversamping an artificially created interpolation of an already sampled signal.

 

[pkane]

Here's my understanding:

Oversampling and Upsampling are similar but not the same operations.

Upsampling is an arbitrary increase in the sampling rate through interpolation (including fractional, non-integer factors, like 44.1k to 96k).

Oversampling is also an increase in the sample rate through interpolation that's made simpler by the use of an integer (usually a power-of two) factor. In effect, oversampling is a form of upsampling, but upsampling is not always oversampling Oversampling is easier to implement in hardware, which is why it's used in DACs and ADCs.

 

[voodooless]

Oversampling is sampling of the original signal at a sample rate that is higher than needed

If it's not needed, it wouldn't need to be done

(so that it can be downsampled later, after filtering).

What would be the point?

Upsampling is oversamping an artificially created interpolation of an already sampled signal.

I don't know what any of that means... What is a non-artificially created interpolation? All samples that you start with are always already sampled.

 

[tuga]

What would be the point?

Distribution (e.g. CD)?

 

[voodooless]

Distribution (e.g. CD)?

Yeah sure, but why would you upsample first, and then downsample again.

I can think of one one reason todo this: application of non-linear digital filers, but otherwise...

 

[Julf]

If it's not needed, it wouldn't need to be done

"Oversampling can make it easier to realize analog anti-aliasing filters. Without oversampling, it is very difficult to implement filters with the sharp cutoff necessary to maximize use of the available bandwidth without exceeding the Nyquist limit. By increasing the bandwidth of the sampling system, design constraints for the anti-aliasing filter may be relaxed. Once sampled, the signal can be digitally filtered and downsampled to the desired sampling frequency. In modern integrated circuit technology, the digital filter associated with this downsampling is easier to implement than a comparable analog filter required by a non-oversampled system."

I don't know what any of that means... What is a non-artificially created interpolation? All samples that you start with are always already sampled.

In a DAC, yes. Not in an ADC, or full chain.

 

[radix]

Although this might be a case of lipstick on a pig, one reason I see to upsample is some DACs have audible switching noise when going between sample rates. If you fix the sampling rate in your streamer (e.g. 48k or 96k) for everything, then you remove the switching noise. This might be up- or down-sampling, depending on where you fixed the rate.

Personally, I'd buy a DAC that does not have this problem.

 

[voodooless]

"Oversampling can make it easier to realize analog anti-aliasing filters.

That is "needing".

In a DAC, yes. Not in an ADC, or full chain.

This doesn't answer anything..

 

[Julf]

Yeah sure, but why would you upsample first, and then downsample again.

I can think of one one reason todo this: application of non-linear digital filers, but otherwise...

Sonarworks: Should I Be Oversampling?

 

[Julf]

That is "needing".

OK, let me change that for clarity reasons: "Higher sample rate than required by the Nyquist criteria".

 

[Lambda]

But what you are "oversampling" isn't a higher resolution waveform. It doesn't contain any new information - unlike the case of real oversampling.

Who says it needs to?
Why do they call it oversampling in the analog Paper?

What would be the point?

He is talking about ADCs and Anti aliasing

Oversampling is also an increase in the sample rate through interpolation that's made simpler by the use of an integer (usually a power-of two) factor. In effect, oversampling is a form of upsampling, but upsampling is not always oversampling Oversampling is easier to implement in hardware, which is why it's used in DACs and ADCs.

So your definition is oversampling = upsampling with integer value?
Could be but i think i have at least heard of digital Asynchronous or non integer oversampling.

 

[voodooless]

This doesn't answer anything..

Okay, so I think I'm getting somewhere now.

@Julf mentioned ADC's. Oversampling in an ADC is a different process than oversampling in a DAC. I think that is where the confusion comes from:

ADC: sample at a higher than the sample rate, then down convert to the target rate
DAC: up-convert to a higher than-needed sample rate so we can apply digital filtering

Both of these things are called oversampling, and they basically do the opposite...

From the wiki:

In signal processing, oversampling is the process of sampling a signal at a sampling frequency significantly higher than the Nyquist rate.

The term oversampling is also used to denote a process used in the reconstruction phase of digital-to-analog conversion, in which an intermediate high sampling rate is used between the digital input and the analog output