• WANTED: Happy members who like to discuss audio and other topics related to our interest. Desire to learn and share knowledge of science required. There are many reviews of audio hardware and expert members to help answer your questions. Click here to have your audio equipment measured for free!

Does sample rate conversion affect sound quality?

starfly

Senior Member
Forum Donor
Joined
Jun 6, 2019
Messages
370
Likes
295
I've been wondering if sample rate conversion will affect sound quality. For example, if you're using a Denon receiver with Audyssey enabled, any digital music being streamed through the Denon with a sample rate > 48 khz will be down sampled to 48 khz. Will this affect how it sounds?

Also, downsampling from 96 khz > 48 khz should be "clean", as in, they're factors of each other. But how about from 88.1 khz to 48 khz?

Is there any good data on this?
 
You're correct, any integer sample rate conversion should be more or less lossless. I think floating point conversions should also be undetectable if done with good enough quality settings. Sometimes you have to care about inter-sample overs, so decreasing gain slightly could be a good idea.

Archimago has written quite a lot about resampling over the years:

https://archimago.blogspot.com/2015/07/the-linear-vs-minimum-phase-upsampling.html

https://archimago.blogspot.com/2018/01/musings-more-fun-with-digital-filters.html

https://archimago.blogspot.com/2018/01/audiophile-myth-260-detestable-digital.html

https://archimago.blogspot.com/2020/07/summer-musings-post-hi-res-audio-why-hi.html

I generally run my DACs at 48khz and resample everything in software (with SoX, you can look at its quality the infinitvewave site).
 
Last edited:
Windows has a pretty shite resampling system. And I'd say even that is not too audible. I believe our ears are quite forgiving, which makes super-over-sampling the like a Chord M-Scaler does even more ridiculous. That's just my view on the matter, at least...
 
Last edited:
Windows has a pretty shite resampling system. And I'd say even that is not too audible. I believe our ears are quite forgiving, which makes super-over-sampling the like a Chord M-Scaler does even more ridiculous. That's just my view on the matter, at least...


their resampler is there at https://src.infinitewave.ca/
artifacts are all below -90dB; uness you can hear abouve 20000Hz
 
their resampler is there at https://src.infinitewave.ca/
artifacts are all below -90dB; uness you can hear abouve 20000Hz

Any idea why I can hear obvious resampling artifacts on my Windows 10 PC when playing a 48 kHz sweep with the mixer set to 44.1 kHz (or vice versa)?

Steps to reproduce:

1. Open Windows mixer settings, set to 24 bit, 44.1 kHz.

2. Open REW, set sample rate to 48 kHz.

3. Play a sweep from 10 kHz to 20 kHz.

There are many spurious tones, ascending and descending, throughout the sweep. Sounds like an alien spaceship.

Files are attached so you can hear the difference. These were recorded using a loopback in Audacity. You can see there are problems just by looking at the waveform.

Not resampled:

1596614543513.png

Resampled:

1596614952706.png

Maybe REW uses it's own resampling algorithm that's really, really bad? I'll try using a different program to generate the sweep.
 

Attachments

  • windows_10_resampling_artifacts.zip
    3.4 MB · Views: 129
Any idea why I can hear obvious resampling artifacts on my Windows 10 PC when playing a 48 kHz sweep with the mixer set to 44.1 kHz (or vice versa)?
In order to exclude any REW-specific issues (BTW - Java or ASIO sound interface?), please try the Tone Generator (2) in Audacity after setting project frequency. Amplitude 0.3 or 0.9 or 0.95 or 0.999. There is a limiter object (APO) that may start taking effect at about 0.985.

The only time I've had major issues like this was when Windows output interface sample rate was set higher than hardware sample rate on Asus Xonar cards; apparently they or their drivers can only upsample but not downsample and will resort to linear interpolation for the latter. It's also how I found that D1 and D2 drivers >build 1800 (at least Win7 ones) will not init the hardware correctly.

You can try using ASIO e.g. via ASIO4All in REW and see how that goes. Checking your music playback chain is obviously advised.

I only have a Win7 machine at hand now, but I did try the generator in REW 5.19 at 48 kHz (Java) with the (lowly HDA) output device set up at 44.1 kHz and the 10k-20k @-12dBFS sweep didn't show anything out of the ordinary.
 
@T3RIAD I can't realy add to this, since I actualy don't use Windows. I just looked at the graph at that webpage.
But I can imagine that in your testcase the java-sound engine made the conversion once it saw that the sound server requested 44.1
 
Back
Top Bottom