[kennyb123]

It just takes time and experience to figure out what it is that we're after. When I look back at when I started in this hobby at the beginning of 2020, I had a very basic idea of what sound consisted of. And I'm so glad that I progressed gradually because I don't think I would have been able to fully appreciate what my gear is capable of if I hadn't spent all those hours comparing different headphones and pieces of gear and in the process training my ears to be able to detect all those nuances, instead of experiencing a sensory overload which can easily happen with this level of detail being thrown at the listener

Your entire post was spot on. I especially like what you say in what I quoted. It's definitely a journey. I'm thankful to those ahead of me on this journey for sharing their learnings. I hope I can be as helpful to others as they were to me.

 

[Mista Lova Lova]

Your entire post was spot on. I especially like what you say in what I quoted. It's definitely a journey. I'm thankful to those ahead of me on this journey for sharing their learnings. I hope I can be as helpful to others as they were to me.

Thanks, I now also appreciate the reviews and opinions which steered me in the right direction and together with my growing experience allowed me to make educated choices, or even take educated risks at times all of which have recently paid off which shows that I'm progressing on this journey. At the same time, as they say - the more you know the more you're aware of how little you know. And I can now fully appreciate that there are "levels to this game" and that there are both things that I haven't tried yet as well as better implementations of the things I have tried and enjoyed.

The most important thing is that I'm still enjoying my music and new upgrades/discoveries are still capable of sending shivers down my spine. I'll need to visit one of the CanJams at some point - I'd be like a kid in a candy store

 

[kennyb123]

The most important thing is that I'm still enjoying my music and new upgrades/discoveries are still capable of sending shivers down my spine. I'll need to visit one of the CanJams at some point - I'd be like a kid in a candy store

You should definitely do that. Like a candy store, there will be things you like as well as things you don’t like,

 

[taipan254]

ChatGPT is surprisingly pretty good at explaining the difference between Gaussian and sinc based filters in D to A conversion. Pretty cool!

Gaussian interpolation and sinc filter-based interpolation are both methods used in digital to analog conversion (DAC) to reconstruct a continuous signal from its digital samples.

Gaussian interpolation uses a Gaussian function to fill in the gaps between the digital samples. The Gaussian function is centered on the known samples and its width controls the amount of smoothing applied to the signal. Gaussian interpolation is known for its ability to effectively reduce noise and smooth out discontinuities in the signal. However, it may introduce some blurring to the reconstructed signal and may not accurately represent high-frequency components of the signal.

Sinc filter-based interpolation, on the other hand, uses a sinc function to interpolate the signal between the digital samples. The sinc function is the ideal low-pass filter and is used to reconstruct a continuous signal that accurately represents the high-frequency components of the signal. Sinc filter-based interpolation is often used in high-precision DACs where preserving the high-frequency content of the signal is important. However, it can also result in ringing artifacts near the edges of the signal if the cutoff frequency is not carefully chosen.

In summary, Gaussian interpolation is better suited for reducing noise and smoothing the signal, while sinc filter-based interpolation is better for preserving the high-frequency content of the signal. The choice between the two methods depends on the specific requirements of the application and the trade-off between smoothness and high-frequency accuracy in the reconstructed signal.

 

[kennyb123]

ChatGPT is surprisingly pretty good at explaining the difference between Gaussian and sinc based filters in D to A conversion. Pretty cool!

@Mista Lova Lova prefers what's called a "Gaussian polyphase sinc" filter, so it may not be as bad as ChatGPT makes it out to me.

I'm thankful that the designer of PGGB made the decision of what filter to use while also allowing me to adjust a few parameters to tune more to my liking.

 

[kennyb123]

Sure, that's the first setting that I've been so happy with that I've stopped testing others.

Thanks for sharing your settings. I was able to match these except my server couldn’t handle DSD256, so I had to listen at P128. It sounded excellent. I can see why you were happy with it.

I will try to do some comparisons this wee.

 

[taipan254]

@Mista Lova Lova prefers what's called a "Gaussian polyphase sinc" filter, so it may not be as bad as ChatGPT makes it out to me.

I'm thankful that the designer of PGGB made the decision of what filter to use while also allowing me to adjust a few parameters to tune more to my liking.

Question to ChatGPT: What is a gaussian polyphase sinc filter?

A Gaussian polyphase sinc filter is a filter that combines the characteristics of Gaussian interpolation and sinc-based interpolation in a single filter design.

A Gaussian filter uses a Gaussian function to fill in the gaps between digital samples and can effectively reduce noise and smooth out discontinuities in the signal. A sinc filter, on the other hand, uses a sinc function to reconstruct a continuous signal that accurately represents the high-frequency components of the signal.

A Gaussian polyphase sinc filter is a filter that combines these two approaches by using a Gaussian function to interpolate the low-frequency components of the signal and a sinc function to interpolate the high-frequency components of the signal. The result is a filter that provides a smooth and accurate representation of the signal over the entire frequency range.

This type of filter is often used in digital-to-analog conversion (DAC) to achieve a high-quality reconstruction of the analog signal from its digital samples. The Gaussian polyphase sinc filter can provide a good balance between smoothness and high-frequency accuracy, making it a useful tool for many digital audio applications.

What a cool tool!!

 

[Mista Lova Lova]

Thanks for sharing your settings. I was able to match these except my server couldn’t handle DSD256, so I had to listen at P128. It sounded excellent. I can see why you were happy with it.

I will try to do some comparisons this wee.

Thanks for letting us know, we ended up complimenting each other's settings which shows we both have open minds about this - and in my opinion that's how audio discussions should be had

By now I'm way beyond the point where I could be influenced by someone else's opinion contradicting the facts as I see them (there's always the placebo effect but it will disappear sooner or later, that's why I tend to have a few listening sessions before sharing my opinions) so even if you hadn't noticed any difference it wouldn't have changed how much I like that setting. But I'm so glad that you did. I remember back when I had the D90, switching to the Slow filter (knowing about its imperfections) and whilst it was "interesting", it had at least as many weaknesses as strengths about it. The Gustard's factory filter (Vivid) is nicely balanced, Roon's upsampling gives extra air and space but takes away some body. Sinc-M gives a lot of "meat" but it's quite in-your-face. Sinc-L is similar but appears more refined and balanced (but then it's not apodising). I was going to just settle for switching between M and L (M for tracks with lots of apodising errors) when I came across this Gauss and it was different enough that it took me by surprise. And then switching back and forth proved to me that it doesn't really have any obvious weaknesses, it's just what I've been after.

And it also happens to be rather light on my system, compared to Sinc-L which in DSD is a disaster in terms of loading times. So a win-win.

I wouldn't think that 128 vs 256 would make much of an audible difference. Admittedly, 256 is considered to be the "sweet spot" in terms of performance vs usage of processing power. The Gus will not run anything higher than 256 in DSD over PCM mode but I don't think I'm missing out on anything - DSD has much more noise but at 128, let alone 256 it's already so high above human hearing that surely it must be "good enough".

 

[taipan254]

The Gus will not run anything higher than 256 in DSD over PCM mode but I don't think I'm missing out on anything - DSD has much more noise but at 128, let alone 256 it's already so high above human hearing that surely it must be "good enough".

Why not just go direct and avoid DoP? I can def run the x26 Pro on x512 DSD using simple filters and modulation. Similar to you, my computer cannot handle hardcore DSD upsampling.

 

[Mista Lova Lova]

@Mista Lova Lova prefers what's called a "Gaussian polyphase sinc" filter, so it may not be as bad as ChatGPT makes it out to me.

I actually read that excerpt as in line with my (previously mentioned) understanding of some of the filters (the brick-wall ones) being more accurate in terms of frequency response whilst making some sacrifices in the time domain (introducing ringing etc.).

But the thing is, upsampling allows us to have a well-designed and complex filter which will make small sacrifices in the frequency domain (and bear in mind we're talking about the frequency area close to 22.05kHz whereas we, if we're lucky, can only hear up to 20kHz) whilst doing a much better job at reconstructing depth, space, layering etc.

So if one is after a "textbook" reconstruction of the frequency domain alone, then sure, a brick-wall filter will tick that box. But if we're looking at the whole issue of getting the most out of both the frequency and time domains, then it's a matter of compromise. Each DAC's internal filtering tries to reach some sort of compromise but thanks to upsampling and complex software like this we can get much closer to the point where any trade-offs are so insignificant that we're actually hearing something resembling live music.

The theory is just theory, though, I don't know if this would've convinced me on paper, but the results of the filter I'm using speak for themselves. Transformational on my system.

 

[Mista Lova Lova]

Why not just go direct and avoid DoP? I can def run the x26 Pro on x512 DSD using simple filters and modulation. Similar to you, my computer cannot handle hardcore DSD upsampling.

I don't think I ever managed to get the HQPlayer to work with my Gus other than via the DoP. Maybe something was wrong about the settings I was using.

 

[taipan254]

I don't think I ever managed to get the HQPlayer to work with my Gus other than via the DoP. Maybe something was wrong about the settings I was using.

That's a bummer. Are you up-to-date on drivers (if using windows) and firmware? Happy to help via DM as well.

 

[Mista Lova Lova]

That's a bummer. Are you up-to-date on drivers (if using windows) and firmware? Happy to help via DM as well.

Yeah, I'm pretty sure I am. I'll try it again and will DM you if it's still not working. Thanks!

 

[Exekuhtor]

I have listed my X26 in the classified (EU, excellent condition, under 1k€), i enjoyed my time with it. I'm making the jump to a more expensive R2R DAC (probably Spring 3 KTE). Let's just hope i can hear an improvement, lol.

 

[kennyb123]

But the thing is, upsampling allows us to have a well-designed and complex filter which will make small sacrifices in the frequency domain (and bear in mind we're talking about the frequency area close to 22.05kHz whereas we, if we're lucky, can only hear up to 20kHz) whilst doing a much better job at reconstructing depth, space, layering etc.

I would caution you about making assumptions like this. You seem to take generalized explanations about these things as if though are gospel but things are much more complicated than this in the real world. Our ears are far more sensitive to things than would normally seem possible. I’d encourage you to try to follow more of what Rob Watts has posted about his journey developing his WTA filter. He actually conducts listening tests and these often challenge the conventional wisdom that you are espousing. The devil is in the details with this stuff when it comes to playback on high end gear.

What happens around 20 kHz and just above it, despite us not hearing that high, can definitely be audible. I have been a beta tester for PGGB and have heard what just small differences in that can make. Anyone can hear try to hear these for themselves by using the current version and toggling between HF:aggressive and HF:minimal when processing a high res file. Even more confounding is the differences between what is called “deprecated”. One might think they are an expert in such things just by espousing textbook definitions, but the real world has humbled me. I stay away from regurgitating the textbook theories on this stuff because I understand that the real world isn’t that simple. We can’t just hand wave over “making small sacrifices” unless we understand the practical implications of what that will have on the music. Those aren’t accounted for in the boiler plate.

If this stuff was easy then we wouldn’t need applications like PGGB or HQPlayer because all our DACs would have built in filters that can’t be beat. Few, if any, have achieved that.