[gregorio]

This is referring specifically to entirely periodic content, this is not the same for transient content

A transient is made up of “entirely periodic content”, so it is the same! Again, why don’t you try it? Instead of using a duplicated 1kHz sine wave as per my suggestion, use a transient and duplicate that instead. I’m sure you can find a snare drum hit or some other sharp transient online.

Can't see the graph there accurately enough to verify. I've ordered one to test though as I appear to have lost mine.

The graph does appear to agree with the measurements he states though.

Additionally it's still attenuating by a notable amount before 20khz iirc

“Notable amount” in what sense, that you can see it on a graph, that you could design a test signal and whack the level up to a point that it would blow your ear drums out if it were a piece of music? You wouldn’t be able to note a 0.15dB in the most sensitive part of your hearing let alone way outside it, at 19.2kHz!

Do you have examples of contrary studies you feel should have been included?

No, that’s the whole point, there aren’t any. All the numerous studies over the decades since higher than 44.1/48kHz has been available all show results that are no better than random chance, 50% or roughly a percent either side. To get to 52% by adding a bunch of studies together, all you have to do is eliminate a single study with enough test subjects which showed say a 49% (statistically insignificant) success rate.

G

 

[gregorio]

My dac is a holo spring 3 and doesn’t offer over sampling. I really like the sound of NOS but also like the upsampled sound from HQPlayer.

Sorry, I don’t really understand. You’ve got a non oversampling DAC which you’re then effectively turning into an oversampling DAC, why not just buy a typical oversampling DAC? If you want, you could apply EQ to the oversampling DAC that removes roughly the same frequencies that your NOS DAC fails to reproduce, with the added advantage that you can bypass the EQ whenever you want.

G

 

[Scarydog]

Sorry, I don’t really understand. You’ve got a non oversampling DAC which you’re then effectively turning into an oversampling DAC, why not just buy a typical oversampling DAC? If you want, you could apply EQ to the oversampling DAC that removes roughly the same frequencies that your NOS DAC fails to reproduce, with the added advantage that you can bypass the EQ whenever you want.

G

Good question. I suppose my thinking was: I can add OS to a NOS DAC but can't remove it from a DAC that does OS internally. So I can have both. I like how it sounds both ways.

 

[gregorio]

as: I can add OS to a NOS DAC but can't remove it from a DAC that does OS internally. So I can have both. I like how it sounds both ways.

Yes, it’s kind of counterintuitive until you understand that a NOS DAC is not reproducing some of the high frequency content which is in the recording, that an OS DAC will reproduce. So to turn a OS DAC into a NOS sounding DAC it’s effectively just an act of reducing some of the content that the OS DAC is reproducing, which is easier than trying to add content that’s been removed.

It might be worth a try if you can borrow or trial an OS DAC, get yourself a graphic EQ with a lot of bands and try a very gradual roll-off starting around 2-3kHz and going down to around -6dB by 20kHz. Starting there and playing around, you should find something to your taste and then of course you can bypass the EQ and get the OS DAC back.

G

 

[Raketen]

My dac is a holo spring 3 and doesn’t offer over sampling. I really like the sound of NOS but also like the upsampled sound from HQPlayer.

I just find HQPlayer itself clumsy and awkward. I also don’t really like having a computer in my audio chain as it nice to get away from them sometimes . (Even running a headless Linux box I find myself having to futz with it semi regularly)

Interesting discussion, I need to do some reading as I couldn’t completely follow it

IDK if I understand the specific request here and haven't used these features myself but I believe Neutron can output to DSD and network renderers?, works on android/iOS if you want to at least get away from a desktop OS - not sure if such devices actually output DSD (though probably some DAPs have workarounds for Android limitations).

Can only attest to the Android version myself, just used as a normal media player (plus the internal upsampling engine for EQ effects)- the interface takes some getting used to but is also customizable.

 

[Ragnar-BY]

I've tried all major software players and had Chord M-Scaler for a couple years...

1. Nothing sounds better than HQPlayer. Even if you do not take into account DSD upsampling, which is much better than PCM if you have right DAC.
2. Nothing beats Roon interface. An interface that brings our listening experience as close as possible to that provided by physical media and at the same time gives access to modern streaming services with their endless supply of music.

Luckily, these two programs work great together. In my opinion, right now Roon + HQPlayer is absolute best that audiophile can have. Comparable alternatives at the moment simply do not exist.

 

[VNandor]

This is referring specifically to entirely periodic content, this is not the same for transient content

The example does not rely on periodicity and you could try to do the same with a kick drum as well. See pictures below of the beginning of a kick:

The left channel is delayed by one sample (1/192000s) yet this delay is accurately captured in the 48kHz signal as well. The accuracy depends on the SRC algorithm.

Here are the files so anyone who cares enough could verify that a 48kHz PCM file is perfectly capable to properly represent timing differences below ~5us even with "transient" content.

Counterintuitively, (as usual for digital audio), the minimum timing difference that can be represented by digital audio is actually depending on the bit depth, not the sample rate. See some explanation here. It would be great if I could explain properly why that is but unfortunately that would take more than just a couple of paragraphs (on a thread where it would be off topic as well). The gist is that the samples must shift on the "Y" direction to capture the tiny shift in time, not the "X" direction.

 

[KMann]

Counterintuitively, (as usual for digital audio), the minimum timing difference that can be represented by digital audio is actually depending on the bit depth, not the sample rate. See some explanation here. It would be great if I could explain properly why that is but unfortunately that would take more than just a couple of paragraphs (on a thread where it would be off topic as well). The gist is that the samples must shift on the "Y" direction to capture the tiny shift in time, not the "X" direction.

I don't think one can dismiss the importance of higher sample rate that easily. Depending on the type of processing, the effective bit-depth can be influenced by the sample rate. DSD is single bit, and can only represent two levels 0 or 1, and yet thanks to noise shaping (done at 64x CD rate), depending on the modulator used, one can expect at least the equivalent of 16 bits resolution. Similar arguments can be made for higher PCM rates too. The true limit is the DAC's noise floor.

 

[gregorio]

I don't think one can dismiss the importance of higher sample rate that easily.

No one is dismissing the importance of sample rate, it’s vitally important as it defines the audio bandwidth that can be captured/reproduced. What we’re dismissing is the false claim that higher sampling rates affect audible timing accuracy.

Similar arguments can be made for higher PCM rates too.

We’re not arguing against the benefit of oversampling during conversion.

The true limit is the DAC's noise floor.

The true limit is the noise floor of whatever recording we’re converting.

G

 

[KMann]

No one is dismissing the importance of sample rate, it’s vitally important as it defines the audio bandwidth that can be captured/reproduced. What we’re dismissing is the false claim that higher sampling rates affect audible timing accuracy.

We’re not arguing against the benefit of oversampling during conversion.

The true limit is the noise floor of whatever recording we’re converting.

G

The previous poster implied that the timing accuracy was affected by bit-depth and not the sample rate, while that may look like it is true at first glance, all I am doing is pointing out that higher sample rates can help timing accuracy because noise shaping will allow a higher effective bit depth at higher sample rates as one can push the quantization noise from a lower bit depth to the larger bandwidth that will be available at a higher sample rate and this noise will be placed beyond the audible range.

 

[gregorio]

all I am doing is pointing out that higher sample rates can help timing accuracy because noise shaping will allow a higher effective bit depth at higher sample rates as one can push the quantization noise from a lower bit depth to the larger bandwidth that will be available at a higher sample rate and this noise will be placed beyond the audible range.

That doesn’t help timing accuracy though. You cannot push the quantisation noise (or any other sort of noise) in the recording to the larger bandwidth available with higher sample rates. The only noise you can push beyond the audible range (with a higher sample rate) is any dither that you add to the recording. Obviously you cannot get more timing accuracy when converting a recording than the recording actually contains to start with.

None of this makes the slightest bit of difference as we’re talking about audible timing accuracy and as already quoted, the threshold of audibility is about 5 micro-secs, which is very roughly 50,000 times longer than the timing accuracy of even 44/16 with a timing accuracy down to about 110 pico-secs!

G

 

[KMann]

You cannot push the quantisation noise (or any other sort of noise) in the recording to the larger bandwidth available with higher sample rates. The only noise you can push beyond the audible range (with a higher sample rate) is any dither that you add to the recording

Noise shaping was invented for a reason, it is also part of the A/D process. When a signal is processed (like in an upsampler or the DAW), you add more noise due to requantization. Say you take the 44.1kHz 16bit signal to equalize it and do the math in 32 bit or 64 bit float and then convert it back to say 24bits. The quantization noise can be reduced during processing by pushing it out to higher bandwidth if you have a higher sample rate. Of course, simple dither is an option to decorrelate quantization noise, but noise shaping will be far more effective.

 

[VNandor]

The previous poster implied that the timing accuracy was affected by bit-depth and not the sample rate, while that may look like it is true at first glance, all I am doing is pointing out that higher sample rates can help timing accuracy because noise shaping will allow a higher effective bit depth at higher sample rates as one can push the quantization noise from a lower bit depth to the larger bandwidth that will be available at a higher sample rate and this noise will be placed beyond the audible range.

I agree that higher sampling rates can help with timing accuracy as long as it is used to increase the "effective bit depth". I was specifically addressing the following points (and the nonsense comment about periodic and transient signals):

several studies have found that we can discern time domain differences of around 10uS, which would require a bandwidth of about 80khz to describe. Some studies have even showed audibility of time domain differences down to about 6uS which would require an almost 150khz bandwidth to describe.

Easy, why don’t you try it yourself, I presume you have some sound editor software or a DAW? Create a say 1kHz sine wave, duplicate it to the other channel, move it by say 6uS and record the stereo output down. If your software doesn’t allow movement by a fraction of a sample then move it one sample using a sample rate of say 176.4k or higher and down sample it to 44.1k or even 32k if you want, no difference.

This misunderstanding probably stems from confusing frequency with the reciprocal of time delay and not having a good fundamental understanding of the Fourier-transform. Describing a 10us or even 6us timing difference does not require more bandwidth than what 48kHz PCM audio already has. In fact, the minimum timing difference does not depend on the available bandwidth. Ideally (with "infinite" bit depth/SNR), as long as the recorded analog signal does not exceed the band limit imposed by sampling, any sub-sample time delay will be captured and described perfectly which is the opposite of what GoldenOne implies.

The reason DSD can represent such timing differences even though it's only 1 bit is exactly due to what you already touched on: DSD64 actually has around the same dynamic range as a 16bit PCM file as long as the massive amounts of noise is filtered away properly.

 

[gregorio]

Noise shaping was invented for a reason, it is also part of the A/D process.

Indeed it was invented for a reason and as the Wikipedia article explains, Noise Shaping changes the spectral shape of the resultant noise during dithering, hence why it’s called “Noise Shaped Dither”. However, that is the only noise that Noise-Shaping affects. So if for example a mastering engineer applies dither or noise shaped dither, that noise is now part of the recording. A subsequent noise-shaped dither process, say in the DAC, only shapes the noise of the dither the DAC is applying, it does not noise-shape the dither applied by the mastering engineer or the dither applied during AD conversion or indeed any other noise that’s already part of the recording.

G

 

[Scarydog]

I've tried all major software players and had Chord M-Scaler for a couple years...

1. Nothing sounds better than HQPlayer. Even if you do not take into account DSD upsampling, which is much better than PCM if you have right DAC.
2. Nothing beats Roon interface. An interface that brings our listening experience as close as possible to that provided by physical media and at the same time gives access to modern streaming services with their endless supply of music.

Luckily, these two programs work great together. In my opinion, right now Roon + HQPlayer is absolute best that audiophile can have. Comparable alternatives at the moment simply do not exist.

Thanks! I have decided to give HQPLayer another try. It was working with Holo Audio Red as the NAA endpoint.

I tried to use the demo of version 5 but that didn't work so I rolled back to my version 4 but that had now broken by installing the demo of 5. So I started from scratch with version 4 and can't get that working at all.

Now I have installed HQPlayer Embedded (on a different computer) and still can't get any sound out. In the config for HQPlayer it can see my Red as the naa renderer but I cant get it to play.

This is why I wanted a hardware alternative, I want to listen to music not futz around with computers!

Feeling a bit lost!

If anyone can help here is some more info. I have attached files for my config and how everything looks. Roon says its playing but sits at 0:00.

If I switch from NAA to ALSA as the backend it works fine

HQplayer says its playing