slartibartfast
Major Contributor
- Joined
- Apr 18, 2024
- Messages
- 4,219
Does that really make a difference though? Anyway 0.004% distortion is OK for a TV input
Noise shaping and the Ultra DAC
- Thread starter Musician
- Start date
-
- Tags
- wiim ultra
No one has been here for a while, so I thought I was just saying something strange...
In other words, is my understanding of post #4 correct?
Did Amirm use 2 test signals with different sample rates (48kHz/96kHz) in this measurement?
In the first place, nowhere does it say that the 48 kHz signal is 16 bits. And isn't the default output resolution for WiiM 44.1 kHz?
In other words, is my understanding of post #4 correct?
Did Amirm use 2 test signals with different sample rates (48kHz/96kHz) in this measurement?
In the first place, nowhere does it say that the 48 kHz signal is 16 bits. And isn't the default output resolution for WiiM 44.1 kHz?
Probably.
@WiiM Team , @WiiM Support , @RyanWithWiiM - can this be confirmed as still being an "issue" or has a fix been applied in firmware? I fear for anyone coming to the forum and seeing this thread title, trying to understand what's being said and wondering whether they should do something which to me seems rather obtuse. Thanks
Last edited:
Thank you
I interpreted this as only the person who took the measurement would know.
But it's the gospel from the Blessed Amir, so it must be importantThank you
I interpreted this as only the person who took the measurement would know.
???
Machiavelli
Member
- Joined
- Jul 15, 2025
- Messages
- 11
Blasphemy!
slartibartfast
Major Contributor
- Joined
- Apr 18, 2024
- Messages
- 4,219
He just presents the graphs without explaining how they are produced. I guess the acolytes are expected to knowBut it's the gospel from the Blessed Amir, so it must be important
As far as I understand it, this specific measurement is done into the WiiM Ultra optical input, and measured at the analog/line output.
Two different sample rate are sent to the optical input:
a) first test with 48kHz sample rate (meaning usable audio signal is 0-24kHz; everything else is out-of-band), and
b) second test with 96kHz sample rate input signal (meaning usable audio signal is 0-48kHz; everything else is out-of-band).
But in both cases the measurement bandwidth at the Audio Precision analog input was 90kHz (which means that AP ADC uses some sample rate higher than 180kHz for this test). As such, this is a wideband measurement especially designed to capture any artifacts that are out of the operating range of the DAC.
The fact that the Ultra toslink 48kHz test shows higher (and relatively constant) THD-N level indicates that there is some kind of noise or spurious tones in this measurement, most likely between 24kHz and 90kHz (i.e. out of the audible band), when using this sample rate. As already mentioned, usually a wideband measurement that looks like this is indicative of noise shaping, which is a very common practice in DAC design to increase SNR in the DAC operating range by pushing its noise out of the usable (and audible) band.
EDIT: Just checked the ASR review and Amir actually explicitly shows this noise shaping in a separate measurement - look at the noise increase above 30kHz:
Here's one example from a different DAC review illustrating this, as well as how just decreasing the AP measurement bandwidth in the same test from 90kHz to 45kHz (for the same 48kHz sample rate) can decrease the measured wideband THD-N level:
I.e., IMHO this kind of measurement result should not be a problem in itself - it just indicates what kind of DAC implementation is used in WiiM Ultra. Some DACs do better in this test because they manage low noise in their operating range without noise-shaping - that's great from a technical perspective, but I doubt it has much audible benefit.
Note also that if you increased the measurement bandwidth at the Audio Precision to a higher frequency (e.g. 200kHz) many other DACs that measure 'well' with 90kHz measurement bandwidth might potentially measure significantly worse. Likewise, if you decrease it to 25kHz most would measure better. It just depends what it is you want to measure, and 90kHz is a bit of a random figure to use - I don't see any reason why it would be especially relevant from psychoacoustics standpoint.
So in summary I don't really see this measurement to be indicative of any real problem or bug in the WiiM Ultra; IMHO you can use any sample rate without any fear of degradation in sound quality.
Just my 2 cents!
Last edited:
By the way, 0.004% THD+N with 48kHz optical input and a 90kHz measurement bandwidth indicates that the noise above 24kHz is about -88dB below the fundamental.
So to put things into perspective - we're talking about ultrasonic noise (i.e. noise above the maximum frequency young humans with healthy hearing can hear) that is almost 90dB below 0dBFS in level.
So to put things into perspective - we're talking about ultrasonic noise (i.e. noise above the maximum frequency young humans with healthy hearing can hear) that is almost 90dB below 0dBFS in level.
Infamy! Infamy! They’ve all got it infamy! (a very British pun/reference I do admit..)
Or pretend they do. Or maybe it’s a matter of faith ex cathedra?He just presents the graphs without explaining how they are produced. I guess the acolytes are expected to know
I agree!I.e., IMHO this kind of measurement result should not be a problem in itself - it just indicates what kind of DAC implementation is used in WiiM Ultra.
So in summary I don't really see this measurement to be indicative of any real problem or bug in the WiiM Ultra; IMHO you can use any sample rate without any fear of degradation in sound quality.
Just my 2 cents!
Sorry for taking up so much of your valuable time. Thank you so much
As far as I understand it, this specific measurement is done into the WiiM Ultra optical input, and measured at the analog/line output.
Two different sample rate are sent to the optical input:
a) first test with 48kHz sample rate (meaning usable audio signal is 0-24kHz; everything else is out-of-band), and
b) second test with 96kHz sample rate input signal (meaning usable audio signal is 0-48kHz; everything else is out-of-band).
But in both cases the measurement bandwidth at the Audio Precision analog input was 90kHz (which means that AP ADC uses some sample rate higher than 180kHz for this test). As such, this is a wideband measurement especially designed to capture any artifacts that are out of the operating range of the DAC.
The fact that the Ultra toslink 48kHz test shows higher (and relatively constant) THD-N level indicates that there is some kind of noise or spurious tones in this measurement, most likely between 24kHz and 90kHz (i.e. out of the audible band), when using this sample rate. As already mentioned, usually a wideband measurement that looks like this is indicative of noise shaping, which is a very common practice in DAC design to increase SNR in the DAC operating range by pushing its noise out of the usable (and audible) band.
EDIT: Just checked the ASR review and Amir actually explicitly shows this noise shaping in a separate measurement - look at the noise increase above 30kHz:
Here's one example from a different DAC review illustrating this, as well as how just decreasing the AP measurement bandwidth in the same test from 90kHz to 45kHz (for the same 48kHz sample rate) can decrease the measured wideband THD-N level:
I.e., IMHO this kind of measurement result should not be a problem in itself - it just indicates what kind of DAC implementation is used in WiiM Ultra. Some DACs do better in this test because they manage low noise in their operating range without noise-shaping - that's great from a technical perspective, but I doubt it has much audible benefit.
Note also that if you increased the measurement bandwidth at the Audio Precision to a higher frequency (e.g. 200kHz) many other DACs that measure 'well' with 90kHz measurement bandwidth might potentially measure significantly worse. Likewise, if you decrease it to 25kHz most would measure better. It just depends what it is you want to measure, and 90kHz is a bit of a random figure to use - I don't see any reason why it would be especially relevant from psychoacoustics standpoint.
So in summary I don't really see this measurement to be indicative of any real problem or bug in the WiiM Ultra; IMHO you can use any sample rate without any fear of degradation in sound quality.
Just my 2 cents!
Thanks for another of your usual excellent explanations.
Rightly or wrongly, I’ll naively take this not as an issue as such, but as positive evidence that WiiM is using noise shaping to push the “noise out of the usable (and audible) band” which is surely a good thing, and that mere mortals like me needn’t rush to make any changes to my WiiM app…
I see it the same and agree, in case that's any comfort.
I've just read the whole thread and I agree that it seems that OP misunderstood the ASR test.
There's no reason to believe that changing the WiiM Ultra optical output digital resolution would change the result measured at the WiiM Ultra RCA (DAC) analogue output. This is not what was tested in the ASR review.
In the ASR test it is IMHO clear that Amir tested two different sample rates of the incoming signal to the optical input to the WiiM Ultra, and then measured the resulting output of the Ultra DAC.
With a 48kHz sample rate input signal the DAC noise shaping was captured within the 90kHz test bandwidth, but not so with the 96kHz sample rate input signal. Note that even with 96kHz input there could still be noise shaping, just pushed to a higher frequency (and out of the 90kHz measurement bandwidth).
There's no reason to believe that changing the WiiM Ultra optical output digital resolution would change the result measured at the WiiM Ultra RCA (DAC) analogue output. This is not what was tested in the ASR review.
In the ASR test it is IMHO clear that Amir tested two different sample rates of the incoming signal to the optical input to the WiiM Ultra, and then measured the resulting output of the Ultra DAC.
With a 48kHz sample rate input signal the DAC noise shaping was captured within the 90kHz test bandwidth, but not so with the 96kHz sample rate input signal. Note that even with 96kHz input there could still be noise shaping, just pushed to a higher frequency (and out of the 90kHz measurement bandwidth).
Similar threads
