Upsampling - Yes or No

Volumio offers a up sampling feature and allows you to control the degree of it. I just turned it on and have sampled some local files and some Qobuz files and it appears to be working on both. I say that as although the songs or 44/16 my SMSL DAC is showing 192/24.

Folks, do you use it, if so what degree and have you noticed SQ improvements?

it’s impossible to get improvements with upsampling only, best case scenario you will be on par

There will Absolut be no improvement. With your example it will even be worse.
You added non existing bits in your audio. as 192 is not a duplicate of 44.1
If you up sample it needs to be a “by 2” to avoid introducing non existing bits, like:
44,1 - 88,2 - 176,4 - 352,8.
But still it will be the same audio signal. Just like giving a kid candy. you can break a candy bar in 2, still it’s 1 candy bar.

1 Like

Daphile music server has an option to upsample to the highest sync rate. So if your DAC is capable of 192KHz, 48K and 96K will upsample to 192K, but 44.1K and 88.2K will upsample to 176.4. This is independent of DSD playback which will continue to play at the native rates. In this, Daphile is superior, but Volumio still sounds better.

1 Like

hey,

that is starting an endless discussion.

for me it turned out it is very much depending on what you‘re listening to on what system and sofore does not match with why or not volumio - which is streamer software with certain cool options

br
Josh

1 Like

The only benefit I see is gapless playback. But thats about it.

Wonderful, again and again the academic discussion. Usually the technicians say “you can’t hear that”. But the technicians usually hear nothing anyway. With a high-resolving system - Lyngdorf TDAI1120, Buchardt S400mkII, Primo 1 as a pure bridge and very good cables - you can clearly hear a difference whether upsampling is switched on or not. Whether better or worse is a matter of taste. Not day and night, nobody gets blown out of the water. With UPSamp. the depth layering and individual imaging of instruments is clearly better. Without upsampling. the music is a little snappier. Anyway, I enjoy the music with Upsamling.

I haven’t tried upsampling before reading this thread as I figured the native sample rate will be best. I play local wave files form a harddrive, some are cd rip and some are higher resolution files up to 32 bit. My connection is coax RCA and limited to 24/192. When setting Volumio to resample the files to 24/192 the playback of 16/44 seem slightly more spacious and controlled, but playback of 32/44 is far better overall and alot smoother.
My DAC is a Denafrips Ares II which goes up to 24 bit. In both cases the DAC only receives 24 bit because of the limitation of the connection but may the resampled data make the DAC perform at a higher level because 24/192 is the “optimal” resolution for the DAC?

1 Like

No.

The 44.1Khz playback sample rate chosen for CD is not an accident of technology but rather a product of human hearing physiology and digital sampling theorem. The reproduction chain that stretches from the signal to the speakers is optimised for the 20hz - 22Khz analog audio band, with a Nyquist digital sampling frequency around 44.1 to 48Khz.

In terms as simple as I can make them. Between the samples is an undesirable aliasing frequency that is filtered out during recording and analogue reconstruction. What happens in the real world when you upsample beyond the Nyquist frequency is you raise the aliasing frequencies and create a potential for the analogue components, loudspeakers especially, to produce non-tonal harmonics within the audible band. (Alias Intermodulation Distortion.)

To summarise a very complicated area of digital sampling theorem.

  • Upsampling can be no better but may be worse.
  • Upsampling can not reproduce a signal any more accurately than originally recorded.
  • Upsampling may introduce non-tonal distortions.

I’ve just acquired the Rivo and have compared the up sampling to not up sampling and you’d be totally wrong with your statement. The difference is actually obvious and startling. In my system (dedicated, sound treated room) up sampling expanded the soundstage, separated instruments quite a bit and fleshed out more microdetails. So Mr Wheaten, making definitive statements without first hand experience is not helpful and makes your look foolish!

You second post and already start calling people foolish? You must be a delight…

We people can’t hear above 20kHz.
Suppose your playing a 44.1 kHz - 16 bit track, all you’re doing is moving the 20kHz filter to a higher frequency, which you might detect if you’re a person that can hear above 20kHz.
If that’s the case I would suggest you write to the Guinness book of records, as this makes you very special.

If you like the output of up sampling, do what ever you want. If you feel its all glory enjoy it.
I can’t argue on personal preferences.

@wheaten People who argue against hi res music often make this argument. But isn’t this (can’t hear above 20Khz) argument missing the point that higher sampling means more detail at all frequencies? Such is explained here:

Sample rate

Sample rate is the number of samples recorded per second. The higher the sample rate, the closer the recorded signal is to the original. Sample rate is measured in hertz.

If the samples recorded above were plotted on a graph, the resulting representation of the sound wave would not be too accurate:

A sound wave with a low sample rate
A sound wave plotted from 10 samples
However, if the sample rate is doubled - twice as many samples in the same time period - the resulting representation would be closer:

A sound wave with a doubled sample rate
A sound wave plotted from 20 samples
However, the higher the sample rate, the larger the resulting file. As a result, sound files are often a compromise between quality and size of file.

@backenst

You missed the topic,

I don’t argue Hi-res, if the master track is recorded and released in a higher sample rate, you’ll notice more details. No argue about this.

However stating that a master track of 44.1kHz reveals more details when up sampled to 88.2 kHz, that’s what I am talking about. It’s just doubling the same data. The only slight difference is that the filter is placed in a higher frequency range,
Even worse if you up sample to 44.1kHz to 48kHz. Here you might hear a sound difference, simply because non existing bits are added.

Keep in mind as you refer to soundwaves, up sample is in the digital domain, not analog.

Newb type question…

My understanding and please correct me if I’m wrong is that in the simplest terms, upsampling algorythms try to estimate the in between sample by taking the first sample and comparing the next sample (at the original sample rate) then inserting the new sample at exactly the mid point of the two.

Eg. S1 = 128, S2 = 512 … insert new sample =256 between S1 and S2.

I would assume the algorythm is more complex than my example, but isn’t this the basic gist of what’s happening?

If so, then the resulting upsampled file would more closely aproximate the analog waveform when the digital to analog conversion takes place, yes?

Cheers

@dmonds

If you up sample by factor 2 like 44.1 => 88.2 your just double the info without adding none existing bits.

If the factor is not by a factor 2, you need to interpolate the missing bits. We can assume it might be following the original signal better this way, but we and neither the algorithm can tell, as these additional bits have no reference to the original waveform.
By adding these none existing bits, your music can loose sharpness which are hearable by string equipment such as Guitars.

As you can see below, upsampling from 44.1 to 48, will mess up the audio signal.
If you look closely (blue original signal, red the 48k resample points) Only 3 have an match with the original wave, all others have no datapoint. All these need to be interpolated to “a value”, which are not in line with the original wave.

Here is an interesting read on the matter, with a clear explanation:
https://kevinboone.me/sample48.html?i=1

Ok, thanks for trying to explain and thanks for the link, which I’ll read later this evening (NZ Time).

I can easily imagine issues trying to upsample from 44.1 to 48, however, surely a 2x upsample from 44.1 to 88.2 (or a direct multiple of 44.1) might be a simpler matter considering a computer can buffer thousands of sequential samples and have an algorithm which estimates the new, inserted samples based on the following 2 (or 3) samples of the original data stream.

I’m not sure what you mean by…
“your just double the info without adding none existing bits”

Does this mean (very simplistic examples of 2x upsample for 16/44.1)…

If 16/44.1, 4 sample stream = 128, 512,128, 512…etc
Then you are saying that the upsampled 16/88.2 stream would look like…
128,128,512,512,128,128, 512,512…etc

and not…
128,256,512,256,128,256,512

?