Purpose of Upsampling

Can someone explain the purpose of upsampling?

Have a music streamer/server that has HQPlayer embedded and I’ve tried upsampling MP3 and flac files to DSD64 and DSD128 but don’t notice any difference. Know upsampling is also available in Roon.

Seems like trying to get more resolution than exists in the source file. Like converting a mp3 file to flac.

What am I missing?

Converting a digital (sampled) signal to a continuous analogue waveform requires interpolation to produce the values between sample points. Doing part of this interpolation digitally (upsampling) simplifies the analogue circuitry and gives better results.

So before anyone thinks I’m smart, I found the above on the web. However another point to your question, upsampling a MP3 is just taking a crappy file and adding more crap. It won’t get any better. Let’s see if someone a lot smarter than me replies.

1 Like

Converting an MP3 to a flac will not improve the quality, you are effectively doing what playback would do only in advance (and then wrapping it in a flac container). It could be argued that this might reduce the noise levels on playback as playback of flac is “easier” for a processor to do, but this difference would be tiny.

Upsampling, though, is useful as it allows the reconstruction filter used in the DAC process to
be simpler, hence having less imperfections in the audio band.
If “perfect” theoretical filters were practical to make, standard redbook 44.1 kHz would be enough for perfect reproduction up to 20 kHz. Upsampling allows the practical implementation to approach theoretical performance more nearly.

Some of the above may make sense, it’s half past two in the morning here :slight_smile:

1 Like

In the PCM realm the purpose of upsampling is to move the cutoff of the brick-wall filter further above the 20 kHz end of the audio band. The filter is needed to suppress aliasing artifacts. Look at any of JAs lab reports in Stereophile where cause and effect due to higher sample rates in the post-reconstruction frequency domain is obvious. These high order filters are anything but phase coherent. I’ve done some filter work in the past. High order filters are a pain to design. The idea is to keep these less desirable filter side effects as far away from the audio band as possible. The point of ‘upsampling’ flac in the DSD realm escapes me.

1 Like

Just curious are all mp3 music files crappy?

Best wishes

If you rip a CD to mp3 then rip the same CD to wav or flac, you could clearly hear the difference. But, I would not call it crap. It’s a format that has its own use and users. It is still the most popular format for internet radio stations, and is sold by Spotify and Tidal streaming services. All those subscribers have chosen a format that suits their budgets, playback gear, and probably tastes.

Hey Serhan…

My earlier question arose from this statement…and made me wonder…

When I ripped the Gabriel Mervin “Say Somethin” to my thumb drive
all the versions made it to thumb drive
DSD 64, DSDDirect Mastered 192k 24-bit, 96k 24-bit, 44.1k 24-bit

The high res DSD Direct Mastered played through my Oppo 205 shows
off the amazing sound captured …playing back the 192k, 96k and 44.1k
there is a noticeable lessening of the sq…

Most of my music is in mp3 format and in no way would could they be considered

But did make me wonder though…

Thanks Serhan and everyone…

Best wishes ya’ll

1 Like

Good info. Wonder why upsampling flac to dsd is a questionable option. Still in my system at least can really discern any audible difference/improvement.

Doesn’t the DSD Sr. already upsample everything by some huge amount (seems like I recall 20x).

Guess with age my memory as well as my hearing is going!

For the DSD, incoming PCM is first converted to 352.8 or 384 PCM (depending on the sample rate of the input) then to 56.448MHz (20 times the DSD rate), then to quad rate DSD (DSD256).

(I see Ted responding. I hope I got it right).