We have talked about sampling rates and bit depths in an earlier article within this series. There we also pointed out that while higher is better, there are also limits to what we can still perceive. In this article, we are going to talk about up- and oversampling. What is the difference? And why should you apply it?
The terms upsampling and oversampling are pretty often mixed up. And this is understandable, because they are – in part – similar things with an equivalent purpose. Let’s try to define what upsampling and oversampling do.
Upsampling
If we have to define upsampling, it is the process of increasing the sampling rate of an audio signal. Often this is also accompanied by an increase in the number of bits per sample.
An example of upsampling is increasing a digital signal from 16 bit / 44.1 KHz to 24 bit / 192 KHz. The observant reader can see that we are not talking about an exact doubling of the sampling rate ánd that the number of bits has gone from 16 to 24.
Now, you cán upsample from 44,1 to 88,2 or 176,4 KHz and you don’t need to increase the amount of bits. It’s just a possibility that upsampling has to offer. .
Upsampling in principle does not affect the musical information. However, a lot of calculations have to be made and if it is not an exact doubling, like from 44.1 KHz to 88.2 KHz, then it has to be rounded off and the algorithm has to guess sometimes.
Oversampling
Oversampling works a little differently. Oversampling simply increases the number of samples using interpolation. So it calculates the values between samples and based on that value it adds samples.
Oversampling also does not increase the number of bits. Moreover, oversampling will only increase the number of samples in multiples. So, for example, we go from 44.1 KHz to 88.2, 176.4 or, say, 352.8 KHz. You cannot oversample from 44.1 KHz to 96 KHz. Then we are talking about upsampling again.
Internal and external
Now it is true that many dacs already oversample internally. Unless we are talking about NOS dacs. So in a lot of cases, you can’t choose at all about whether you want to oversample or not, since the converter decides that for itself.
Upsampling is often a choice, though. At least: if we look at it very black/white (or binary :-)) for a moment. This is because many soft- and hardware give the choice to set the digital output to a certain sampling rate. What sounds best is totally dependent on the quality of the source as well as the dac.
Gray
You now know the – subtle – differences between over- and upsampling. And you now know that many d/a converters actually always oversample. But why do they do that? And why does upsampling even exist?
Basically, oversampling is used to reduce quantization noise. The more samples and the more data, the less rounding errors occur. At least that’s the theory. Also, it has filtering advantages. High sampling rates require less steep (analog) filters and reduce – indirectly – aliasing.
If we look at the usefulness of upsampling, we are indirectly talking about the same things. Higher sampling rates allow less steep filters. Ok… but oversampling in the dac already fixes that…Right?
Yes, it does. But the question is, how well does the dac perform that task? Some dacs really benefit from getting higher sampling rates, because then they do not have to work as hard on the data. And that is audible. In short: by making smart use of upsampling, you can improve the performance of the dac.
The only way to find out what is optimal for your dac is to go out and listen. Grab a track with which you can observe differences well and incrementally increase the sampling rate (and corresponding bit size). You may find that at some point it sounds best.
Next Episode
In the next installment, we’ll look at digital interfaces. Which ones are out there? And which one is best (if any)?