Explain what oversampling is to me

I don’t have a very good understanding of what oversampling is and would like for someone to clarify to me what exactly it is.

How do you tell whether a dac has oversampling or not, can you down sample, can you use programs for oversampling, and where does the file quality come into play?

Oversampling is interpolation process so it creates samples between the original samples. It’s a form of guessing.

NOS (Non-OverSampling) dacs don’t do oversampling.

Unlike increasing bit depth which is reversible without loss of quality, oversampling changes the music file. You can downsample, generally but doing back and forth results in quality loss.

You can use programs.

The main reason for oversampling is to make the analog low pass filter easier since you’d have a much larger bandwidth.

This sounds more like upsampling rather than oversampling, with oversampling in a dac design you aren’t making guesses I think, pretty sure you are just inserting empty samples or samples designed to help the dac in conversion, so not really trying to add more to the music. Upsampling a file or setting a program to upsample, which typically will try and guess I believe. But I don’t know for sure lol

@NoName in what context are you wanting to know, in terms of dac design or software/files?


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 analogue output. Here, digital interpolation is used to add additional samples between recorded samples, thereby converting the data to a higher sample rate, a form of upsampling. When the resulting higher-rate samples are converted to analog, a less complex and less expensive analog reconstruction filter is required. Essentially, this is a way to shift some of the complexity of reconstruction from analog to the digital domain. Oversampling in the ADC can achieve some of the same benefits as using a higher sample rate at the DAC.

1 Like

Was just going to post that exact same paragraph, lol.

1 Like

Yes, but that doesn’t involve “guesses” going off the wikipedia except, it’s either empty samples or specific samples designed to aid in da conversion, it’s not trying to add audible information to the audio, whereas upsampling is trying to make audible guesses

While the terms are closely related, they are not the same thing in the context of dacs

Definition of interpolation from deepai.org

Interpolation is making an educated guess with the information within a certain data set. It is a “best guess ” using the information you have at hand.

Here is an article explaining the differences between upsampling and oversampling in the context of dacs


Also nice little article from stereophile

At this point I’d be better off just leaving other people to explain it that are much more educated on this than I, I’ve just been taught that oversampling and upsampling are different things in the context of dacs, where one is designed mainly to aid in conversion and is essential to a sigma delta dac, where upsampling is designed to try and add information to a file that previously didn’t exist and is more pertaining to files and stuff like that where oversampling is mainly regarding the dac itself

1 Like

The use of the word Oversampling came from early DAC’s that couldn’t represent the entire bit depth of the signal instead they would use the extra signal to modulate the output at a higher sampling rate to add that “information” back in.
When modern DAC’s refer to NOS, it’s not about resolving existing information, it’s mathematically generating additional samples to make the filtering easier, this is pretty much mathematically identical to upsampling, but done at a different point in the chain.
Upsampling is exactly what it sounds like it’s adding additional samples, much like scaling up an image, to do it you make assumptions on what the original signal should have been, i.e. do you assume the pixels represent smooth gradients, or should you preserve edges in areas of high contrast.

Mathematically they are both going to end up being convolutions, I suspect Upsampling is going to be used to refer to more complex signal processing than oversampling.


Taking more datapoints from the incoming signal than required to satisfy the output bandwidth.
This is especially useful to push aliasing-errors out of the desired spectrum and improve SNR in Delta-Sigma DACs.

Ads datapoints to an incomming signal based on various formulas.
This is well known from image processing. Computerphile video on the subject

The biggest difference
Upsampling happens entirely in the digital domain, oversampling happens at the transition between D and A.


Okay, well I know I don’t understand everything being said here due to most of this being new terminology to me, but I think I get the general idea. Between upsampling and oversampling, which are necessary for the whole signal path to function, and which are done by either a dac or program or one or the other depending on which you choose? Can it depend on the dac you choose whether it is a necessary process, as in a choose by design whether it needs upsampling or oversampling to function?

In a multibit/R2R DAC neither are required.
In a DS DAC, you could consider the entire operation of the DAC to be an oversampling filter. You’re taking 1 bit (I know they don’t really work this way), and turning it on and off very quickly to emulate a continuous analog signal.
Some R2R/Multibit (old ladder chip) DAC’s are referred to NOS (None Oversampling), that means that you just get the input signal on the output.
Most DAC’s do some oversampling/upsampling, your smoothing out the signal, and potentially improving detail retrieval in doing so, but you’re not using information to do it. Those filters introduce ultrasonic artifacts in the time domain called “ringing”, so if you were to look at a square wave or an impulse response on a scope on an oversampling DAC, you would see unintentional artifacts on either side of each edge transition.
Filter design for DAC’s is complicated, there is no ideal filter, Chord for example have the mScaler which takes very large numbers of samples for upsampling and noise shaping.
Peoples preference for NOS/OS DAC’s vary, there is a distinct difference in how they sound.


Tldr both can sound great if well implemented and will faithfully reproduce your music for the most part lol. Personally I wouldn’t buy a dac on the fact that it’s oversampling or non oversampling because while they do have a bit of a specific sound to them, really it entirely depends on the rest of the dac for how good it actually is. There are excellent nos and excellent os dacs, and of course there are shit nos and shit os dacs, so it really comes down to tastes and everything else

1 Like

Just adding some info: oversampling and upsampling gets confusing because in DAC that’s the only way to have a higher sample rate. Oversampling in ADC, which is not common in audio, is important for getting higher resolution, as it helps let say “extract more information” from an analog signal.

Also, on Delta-Sigma, their main use is noise shaping. The move for higher frequencies also moves the noise, especially quantization noise. With this is easier to filter out this effects and you get a better signal.

So what they do is: they up the resolution by repeating bits, then they reduce the resolution beck to original with higher data rate. This way, you have the same data at a higher rate, which you can convert and have the noise shape benefits I talked about earlier.

Now that’s a simplified explanation but I hope it added some info here.