Really bad resampling of 44.1kHz/16bits files with Fixed Resoultion SPDIF setting

[erazortt]

Dear all,

I have tested the output quality of the resampling which is being performed when Fixed Resoultion is selected for SPDIF Output Resolution. The Wiim Ultra is connected to a DAC via Optical and I measured the RCA output of the DAC. I tested the maximum setting of 192kHz/24bits with four files of the CCIF 19kHz & 20kHz dual tone, where these files differed only by sampling rate (44,1 vs 48 kHz) and bitdepth (16 vs 24). I played the files using the Wiim App.
The files with 44.1kHz/24bit, 48kHz/16bit and 48kHz/24bit are resampled correctly, however the file with 44.1kHz/16bit is not!
See the measurements below. (I am using the newest Firmware: 5.2.704665)

Greetings!


44.1kHz/16bits file: not ok


44.1kHz/24bits file: ok


48kHz/16bits file: ok


48kHz/24bits file: ok

 

[EddNog]

Impressive first post!!!

-Ed

 

[Burnside]

Dare I ask, can you hear the difference?

 

[canard]

It's annoying because it corresponds to the most common case "cd resolution"... such distortions on just one ccif...

 

[canard]

resampling should be avoided for the moment if this is confirmed...
A long time ago onlyoneme had already observed the strange impact of resampling...
if confirmed....it's up to wiim to investigate

 

[merifon]

Maybe there are dithering problems, passing from 16 to 24 bit and for not exact multiple sample rate.
Could you test even more combinations?

 

[Burnside]

Have you submitted your findings to WiiM via the more/feedback section in the app so they can inspect your logs and comment on your findings? Again, is this “bad resampling” audible?

 

[onlyoneme]

What happens if you reduce test signal RMS level slightly, let's say by 3 dB?

 

[erazortt]

Dare I ask, can you hear the difference?

Have you submitted your findings to WiiM via the more/feedback section in the app so they can inspect your logs and comment on your findings? Again, is this “bad resampling” audible?

Yes, this can be very clearly heared! The original file has tones only at 19kHz and 20kHz which due to my age I cannot hear, however what I hear in the case of 44.1kHz/16bit are all the other tones below 16kHz which are produced by the lacking resampling. Especally that group of four tones at around 10kHz are really obvious (look at the first spectrum in my first post).
Of course this is a constructed test case because I wanted to make the issue as clear as possible, however, this test is really a standard test for all music equipment. And think of what happens if you have not only two tones in the input, but a recording of a whole orchestra...

What happens if you reduce test signal RMS level slightly, let's say by 3 dB?

When the singnal is reduced, so are the artefacts, by the same amount.

Maybe there are dithering problems, passing from 16 to 24 bit and for not exact multiple sample rate.
Could you test even more combinations?

I can do more tests tomorrow, now it's already way too late to put my test setup back into place.

 

[erazortt]

Also, I want to make clear that the low frequency peaks at 60Hz, 120Hz, 180Hz, and 240Hz are not due to the Wiim but only due to my crappy equipment. I have just used a soundcard to record this. Nevertheless, the crapy soundcard does not invalidate my findings, since I also show the results where the resampling works as expected.

 

[spoidyjan]

This resampling issue with 44.1kHz/16bit files is really concerning, especially since CD-quality audio is so common. Fixing this could make a big difference for music lovers. Did you try testing with a different DAC to see if it changes the results?

 

[canard]

This resampling issue with 44.1kHz/16bit files is really concerning, especially since CD-quality audio is so common. Fixing this could make a big difference for music lovers. Did you try testing with a different DAC to see if it changes the results?

I don't know if the function is used that much...but if it is confirmed, you just need to first disengage it and strongly encourage Wiim to quickly fix the problem.

 

[Smartplug]

Is it even audible? Measurement can show this information but what matters if it’s audible. We get excited when we see thd+noise .011 vs .0011. It’s nice to see many zero before number. Many in blind test wound not able to tell difference unless your ear is precision audio.

 

[slartibartfast]

Is it even audible? Measurement can show this information but what matters if it’s audible. We get excited when we see thd+noise .011 vs .0011. It’s nice to see many zero before number. Many in blind test wound not able to tell difference unless your ear is precision audio.

Yes it is audible.

E

Post in thread 'Really bad resampling of 44.1kHz/16bits files with Fixed Resoultion SPDIF setting'

Jan 17, 2025

Dare I ask, can you hear the difference?

Have you submitted your findings to WiiM via the more/feedback section in the app so they can inspect your logs and comment on your findings? Again, is this “bad resampling” audible?

Yes, this can be very clearly heared! The original file has tones only at 19kHz and 20kHz which due to my age I cannot hear, however what I hear in the case of 44.1kHz/16bit are all the other tones below 16kHz which are produced by the lacking resampling. Especally that group of four tones at around 10kHz are really obvious (look at the first spectrum in...

• erazortt

• 

 

[Valentino]

@WiiM Team @RyanWithWiim @WiiM Support
Heads up please.

 

[Burnside]

@WiiM Team @RyanWithWiim @WiiM Support
Heads up please.

Maybe once the OP submits a ticket with their findings and WiiM have analysed the logs and reproduced the issue, they’ll be in a better position to respond.

 

[Burnside]

Given the prevalence of CD quality files in streaming services (over 90% in some estimations), I’ll play devil’s advocate and ask why you wouldn’t set the fixed resolution to 16/44.1 to avoid sample rate change clicking on your external DAC until this is resolved, unless that too had unwanted artifacts?

 

[Valentino]

Certainly. I can not see from the thread that a ticket has been raised. Therefore the heads-up.
In the meantime my Ultra shall not do resampling but let the following MiniDSP Flex Eight handle it.

 

[canard]

Given the prevalence of CD quality files in streaming services (over 90% in some estimations), I’ll play devil’s advocate and ask why you wouldn’t set the fixed resolution to 16/44.1 to avoid sample rate change clicking on your external DAC until this is resolved, unless that too had unwanted artifacts?

This is verified in terms of impact...but in the meantime the idea seems to be a good one for those who need resampling...

 

[erazortt]

Certainly. I can not see from the thread that a ticket has been raised. Therefore the heads-up.
In the meantime my Ultra shall not do resampling but let the following MiniDSP Flex Eight handle it.

Have you submitted your findings to WiiM via the more/feedback section in the app so they can inspect your logs and comment on your findings? Again, is this “bad resampling” audible?

Yes, I have issued a report, with a link to this thread.

Given the prevalence of CD quality files in streaming services (over 90% in some estimations), I’ll play devil’s advocate and ask why you wouldn’t set the fixed resolution to 16/44.1 to avoid sample rate change clicking on your external DAC until this is resolved, unless that too had unwanted artifacts?

The problem is that I also have my TV connected, which sends out 48kHz/16. And the problem is symmetrical, if I set the resampling to 44.1/16 then the TV will have a bad resampling! And since the sampling rate of 44.1kHz is so much lower then 192, the resampler has much less space to spread its artefact tones, and all tones land in the range of 0-22.05 kHz range instead of the 0-96kHz range, and thus worsening the issue in the audible range very badly. Basically increaseing the tonal energy of the wrong tones in the audible range by a factor of 4. So I opted for the better of both evils and settled to the maximal sampling rate.

So here is the picture when the Fixed Resolution setting is set to 44kHz and 16bits (setting it to 24 bits produces exaclty the same) while playing back the 48kHz/16bit file:



Playing back the 48kHz/24bits file on the same Fixed Output Resoluition setting again produces a nice result:


So the bad resampling is triggered by the 16bit input bitdepth, and is independent on the output bitdepth.
Thus at 24 bit inputs the resampling is always good, independently on the input and output samlping rate.
While at 16 bit inputs, non-integer conversions of the input sampling rate to the output sampling rate will produce a bad result.