PGGB and HQPlayer Discussion

[Mista Lova Lova]

1 minute ago, Zaphod Beeblebrox said:

I don't think you said anything wrong. There are tradeoffs, and preferences play a huge role. Since DSD and PCM do not take exactly the same path, it is really apples to oranges.   It does not tell a whole lot about the quality of the reconstruction algorithms used.

 

But when you switched to DSD, it was a more level playing field.

 

While from your preferences perspective, DSD was always your choice, there are many who preferred PCM over DSD on the same DAC. And similarly, may or may not prefer PGGB on your DAC.

 

In summary, my approach has not changed, but the path through which it is delivered has changed for you.

I was under the impression that PGGB used to initially adopt a different upsamling approach (simply a very, very long filter) which at some point changed, hence your previous insistence that describing it as simply an ultra-long filter wasn't accurate/wasn't doing it justice. I was referring to this in the context of the above exchange of arguments - HQP to my ears offers filters which have particular strengths and as such it is very easy to understand why different people may prefer a different filter. Your product is the first one which has provided me with an experience where I don't feel like I'm sacrificing X for Y - I feel like I'm getting everything at once, cranked up to 11. 

 

And yes, like you say - getting it in DSD was for me personally the last, but also a very needed step towards being able to enjoy it without any ifs or buts.

 

But in the context of the aforementioned exchange, I was expressing my supposition that there's something special about PGGB's approach as it delivers a new level of audio quality (and, as mentioned above, it's not just a matter of the filter being a very long one).

[ray-dude]

2 minutes ago, Mista Lova Lova said:

But in the context of the aforementioned exchange, I was expressing my supposition that there's something special about PGGB's approach as it delivers a new level of audio quality (and, as mentioned above, it's not just a matter of the filter being a very long one).

 

I think this is more about aligning to the processing (or lack of processing) in the DAC.   Experience seems to vary from DAC to DAC, but there is a consistent correlation between SQ heard vs DAC doing as little processing as possible.  I was very very surprised to actually enjoy DSD over PCM (Vinnie Rossi L2 DAC).  It was a first for me, regardless of DSD source (native, or converted with HQP or other source).  That being said, it has been years since I've given DSD a serious shot in my system, so my historical impressions are likely stale

[Zaphod Beeblebrox]

2 minutes ago, Mista Lova Lova said:

I was under the impression that PGGB used to initially adopt a different upsamling approach (simply a very, very long filter) which at some point changed, hence your previous insistence that describing it as simply an ultra-long filter wasn't accurate/wasn't doing it justice. I was referring to this in the context of the above exchange of arguments

I guess it depends on when you last heard PGGB PCM. Previously I did use Windowed sinc based filters that were almost as long as the track. This was with v3, but all that changed beginning of 2023 when I released v5 and the reconstruction algorithms have not changed much since then.

[Mista Lova Lova]

9 minutes ago, Zaphod Beeblebrox said:

I guess it depends on when you last heard PGGB PCM. Previously I did use Windowed sinc based filters that were almost as long as the track. This was with v3, but all that changed beginning of 2023 when I released v5 and the reconstruction algorithms have not changed much since then.

Gotcha. It would have been a while ago so I probably heard the pre-2023 PGGB first. The current version in its DSD form is like nothing else out there, to the best of my knowledge.

[Zaphod Beeblebrox]

21 minutes ago, StreamFidelity said:

But it makes it more difficult to believe what they write. Do we agree on this?

I think it is just easy to listen and decide, instead of believing or not believing a single word I say. 

 

If your belief does not align with what I say, I am perfectly fine with that. I understand the pessimism given PGGB has existed only for the past 4 years or so, compared to other softwares out there. That does not mean I graduated from college 4 years back, had a Eureka moment, and decided to do audio signal processing.

[kennyb123]

42 minutes ago, StreamFidelity said:

But it makes it more difficult to believe what they write. Do we agree on this?

No, I don't agree that it makes it more difficult to believe what they write.  We should judge the strength of a position by the strength of the arguments put forth in favor of that position, not by who it was that put forth the arguments.  

[Zaphod Beeblebrox]

19 minutes ago, Mista Lova Lova said:

Gotcha. It would have been a while ago so I probably heard the pre-2023 PGGB first. The current version in its DSD form is like nothing else out there, to the best of my knowledge.

I had to move away from the windowed sinc based approach as it always felt like a tradeoff between time domain reconstruction accuracy and frequency domain attenuation, the former improved transparency at the cost of being musical and the later improved density but was at the cost of transparency. 

 

Ofcoures the above is my subjective interpretation (and also what I heard from those who tried), but the two variable I was able to control was the width of the window where I kept the original sinc coefficients intact and the tapper of the window. For better reconstruction accuracy I needed to keep more sinc coefficients untouched and for better frequency attenuation I had to make the window more tapered. PGGB v3 used to even have a knob (which we then dropped) to where you can control and reach your own set of compromises.

 

I was happy to move away from making such compromises with the current approach.

 

 

[austinpop]

5 hours ago, Mista Lova Lova said:

A very interesting exchange of arguments (and some sarcastic comments).

 

The most interesting part to me was @Miska's claim that his focus is on restoring the signal the way it was before it entered an ADC, thus to my understanding criticising PGGB's approach as being inherently flawed in that no matter how close to perfection its reconstruction may be, it will inevitably contain and amplify the errors contained within the ADC-processed signal (if my understanding of what's been said is correct).

 

Prior to the release of PGGB DSD I found the above explanation convincing as the Gaussian filters did to my ears offer a more natural/smooth sound than the other HQP filters (albeit not as full-bodied as say Sinc-Mx). I used to always refer to my choice of filter as "my preferred trade-off", which implied a slight disappointment that I couldn't have it all at once. This has changed now as PGGB has better transparency than Sinc-L, similar or better tactility than Sinc-Mx and it manages to be very smooth, as in natural, without needing to give up any of the full-bodiedness in the process, like the Gaussian filters would do. And all of this happens at the same time which I've never encountered before.

 

Which makes me wonder if PGGB's new reconstruction/upsampling approach simply escapes the previous definitions and allows for minimal trade-offs to be made at the cost of a huge amount of processing being required per track.

 

It's also occurred to me that both @Miska and @Zaphod Beeblebrox could perhaps be right at the same time, to some extent anyway, in that there are perhaps inherent flaws in any signal that's been put through an ADC but not as much of it is audible/relevant as @Miska's approach would suggest. Perhaps a ringing analogy would be adequate here - if in fact ringing is mostly a non-issue, is it worth creating shorter filters which reconstruct a smaller portion of the frequency spectrum in order to prevent potential ringing where in reality, on average, we may be throwing away more than we're gaining? Same with apodising errors - perhaps they are audible in an irritating way in certain songs, but there's also a loss of transparency involved when switching from Sinc-MG to Sinc-MG(a). Are we doing more good or more damage in this process?

 

It seems to be hard to find an answer here that would convince everyone, so the topic is definitely complex. But it's my personal opinion that circumstances have now changed as PGGB DSD's sound quality is so outstanding that it can defend itself against otherwise purely theoretical claims.

 

Which gives @Zaphod Beeblebrox a very strong argument; not a theoretical one but a very practical one - the results are simply fantastic. This may change the conversation from "What is likely to give us a better result" to "What are the actual limits of digital audio reconstruction and how far can we push it".

 

You raise some good points, MLL.

 

These seem to be fundamentally different approaches, one (HQP) based on correction/repair of ADC and very recently, DAC imperfections, while the other (PGGB) focusses on near-perfect reconstruction using Shannon-Whittaker. This does not even get into the other fundamental difference of real-time vs. ahead-of-time.

 

The problem that seems to keep erupting is when folks want to ascribe "right' or "wrong" to these approaches. This makes for some -- entertaining for a few, uncomfortable for everyone else -- exchange of posts, after which no minds are really changed.

 

Folks, this is audio. This is a field where almost every component type has competing vendors who all claim theirs is the right path. Speakers, amps, DACs, servers, switches, power supplies, clocks, you name it. I gave up a long time ago about worrying too much about which approach is the "correct" one, and just used my ears.

 

The fact of the matter is that most, if not all, PGGB users have tried other upsampling methods before. I suspect I am typical in moving from being an upsampling skeptic to a fan of the Chord M-Scaler, then HQPlayer, and finally PGGB.

 

If current HQP users don't even want to try PGGB because Jussi says it's incorrect, rings all over the place, blurs transients, they are giving up their own agency and going by faith alone.

 

Listen before judging.

[Schafheide]

Could someone explain the appearance of Metering, whilst playing a PGGB file via HQPlayer Desktop v5 series ??

The spectrogram displayed seems much more complicated in appearance, compared to that of a file processed by HQP.

[Crwilli57]

I am lazy.  I have tried but failed  to learn the answer to perhaps a fundamental question re: PGGB.

 

Do I run it on my entire library and thus create a new library or does it need a PC to run it in real time?

 

I use an Antipodes Kala K50 with ~4 TB of internal media.   
 

Does PGGN rewrite that media into a higher sample rate?

 

Sorry for perhaps an obvious question.

[taipan254]

I always try to remember this academic study of the "Pepsi Challenge" and the power of a brand.
 

Quote

In 2004, Baylor College of Medicine performed a version of the Pepsi Challenge with subjects hooked up to a functional magnetic resonance imaging (fMRI) machine. The results were interesting. In blind taste tests, most people preferred Pepsi, just like the Pepsi Challenge. The Pepsi drinks lit up activity in an area of the brain known as the ventral putamen, which helps us evaluate different flavors. It would seem that Pepsi does taste better! However, when the subjects were told which beverage they were sampling before they tasted it, they decided Coca-Cola tasted better. The fMRI scans showed the sips of Coca-Cola increased activity in the medial prefrontal cortex. This part of the brain is more involved in decision making. When the subjects knew they were drinking Coca-Cola, they weren’t simply evaluating flavor, they were considering memories and experiences. The study’s conclusion was that this prefrontal activity was associating the soda with the brand, in effect, overriding the taste buds. When choosing Coca-Cola, consumers are choosing the brand they know and love. Building a successful brand is more than a good product, it’s building a comprehensive experience.

 

Bias is ever-present! Just try the two approaches for yourself and see what you like! It may not be good as a double-blind test, but it's a start and a great exercise in building confidence in your own ability to listen and discern music that pleases you and your tastes.

[The Computer Audiophile]

I cleaned up the sideshow. Back to the actual apps. 

[Miska]

22 minutes ago, The Computer Audiophile said:

I cleaned up the sideshow. Back to the actual apps. 

 

You deleted the posts without addressing the legal aspects?

 

[The Computer Audiophile]

12 minutes ago, Miska said:

 

You deleted the posts without addressing the legal aspects?

 

An audio site isn’t the best place to discuss international business law. Similar topics arise every so often. Given the number of jurisdictions, cultural customs, etc… it’s always a dead end discussion. 
 

For example, as a US based company I ignore GDPR just like I ignore Sharia law. 

[MarkusBarkus]

4 hours ago, StreamFidelity said:

If someone wants to remain anonymous, I respect that. But it makes it more difficult to believe what they write. Do we agree on this?

...I don't agree with this at all. Admittedly, at first I was off-put by the nom de plum, but interestingly, the net effect was that I had to focus exclusively on the results. And a wee bit later, the technical discourse. What I was unable to do was to judge the results as wrong, poor quality, etc. because I didn't know the pedigree, or didn't know I had an aversion to names that end in: vowels, consonants, sound foreign, sound southern, sound like my ex-wife's, etc. which BTW are stupid biases that we may not even know we have. There are many reasons folks use pseudonyms online. Happy Listening---

Markus "Not Really" Barkus

(believe it or not).

[The Computer Audiophile]

4 hours ago, StreamFidelity said:

If someone wants to remain anonymous, I respect that. But it makes it more difficult to believe what they write. Do we agree on this?

This is the argument that MQA used against me during the RMAF debacle. When it comes to objective data, the identity of someone is far less important because the conclusions and data can be dis/proven by anyone with the right skills. 

[Miska]

7 minutes ago, The Computer Audiophile said:

This is the argument that MQA used against me during the RMAF debacle. When it comes to objective data, the identity of someone is far less important because the conclusions and data can be dis/proven by anyone with the right skills. 

 

What would be reason for you to try stay anonymous?

 

MQA was trying to bundle a lossy DRM protected delivery format with 16fs oversampling filters. Delivered in FLAC, but that encryption ruined FLAC's native compression.

 

IMO, one would want to buy lossless content in standard container without DRM. Just like you buy CD's or such. And then you use what ever playback means you like, This may change over time, the but the content is a separate entity. Not something like SACD, DVD-A, HD-DVD or Blu-ray. Like the HD-DVD's I have, which are not pretty much useless...

 

[The Computer Audiophile]

6 minutes ago, Miska said:

 

What would be reason for you to try stay anonymous?

 

MQA was trying to bundle a lossy DRM protected delivery format with 16fs oversampling filters. Delivered in FLAC, but that encryption ruined FLAC's native compression.

 

IMO, one would want to buy lossless content in standard container without DRM. Just like you buy CD's or such. And then you use what ever playback means you like, This may change over time, the but the content is a separate entity. Not something like SACD, DVD-A, HD-DVD or Blu-ray. Like the HD-DVD's I have, which are not pretty much useless...

 

People have many reasons for retaining their anonymity. I don’t judge them, but I’ll judge the product or opinions. 
 

MQA, the company tried to say Archimago’s work exposing MQA was invalid because he was anonymous. I think we all agree that’s senseless. 

[Miska]

14 minutes ago, The Computer Audiophile said:

MQA, the company tried to say Archimago’s work exposing MQA was invalid because he was anonymous. I think we all agree that’s senseless. 

 

IMO, it would be rather fair to do with that with your own face publicly. Like I did. Mostly those were leveraging other people's work anonymously. So I would say it would be fair to state your sources and present that publicly without anonymity, with your own face. That way you show you are personally standing behind what you are saying.

 

And I personally value such things as Microsoft AuthentiCode signatures and notarized and signed macOS applications. I generally don't install anything unsigned elsewhere than sandboxed test environments. Not least given current world situation. It is damn about $200/year for Windows and $100/year for macOS to have your proper cryptographic signature on the software! Cannot be too much to ask.

 

[jelt2359]

Since we're using analogies, nobody should ever buy, sell, use, or support Bitcoin because we don't know who Satoshi is.

[The Computer Audiophile]

10 hours ago, Miska said:

 

IMO, it would be rather fair to do with that with your own face publicly. Like I did. Mostly those were leveraging other people's work anonymously. So I would say it would be fair to state your sources and present that publicly without anonymity, with your own face. That way you show you are personally standing behind what you are saying.

 

And I personally value such things as Microsoft AuthentiCode signatures and notarized and signed macOS applications. I generally don't install anything unsigned elsewhere than sandboxed test environments. Not least given current world situation. It is damn about $200/year for Windows and $100/year for macOS to have your proper cryptographic signature on the software! Cannot be too much to ask.

 

The great thing is that we as consumers can decide what to purchase or not and evaluate risk/reward pros/cons. Fairness and what should be done is in the eyes of the beholders.

 

There shouldn’t be unnecessary gatekeepers for audio software. 

[bogi]

22 hours ago, Zaphod Beeblebrox said:

OK thanks for establishing you can hear past 20kHz. I sure can't.

 

Miska spoke about audibility of aliasing artifacts. These fold from the area above Nyquist to area below Nyquist, so they may fold into audible range (and with enough low sample rate like 44.1 or 48k most probably they will do).

[Miska]

On that FAQ page, there's also funny way described how long filters cause transient softening/smear/blur:

Quote

A short filter has to dissipate the residual energy past its corner frequency within its time domain footprint, the shorter it is, the more it has to dissipate within the short duration.

 

Yes, short filter keeps the transient's energy concentration as close as possible to the transient without smearing the energy over time.

 

Quote

The peak pre-ringing for PGGB can be less, though the duration can be longer, with a lower magnitude.

 

So the transient's energy gets spread over longer time period...

 

There are also funny copy-paste statements from the internet that are incorrect and so clearly person writing this FAQ didn't understand what he is copy-pasting:

Quote

So as the filter's cutoff frequency increases, the ringing will become more localized in time

 

The whole FAQ is full of all kinds of errors and misconceptions. But I don't have time or motivation to go through all of that.

 

[Zaphod Beeblebrox]

28 minutes ago, bogi said:

 

Miska spoke about audibility of aliasing artifacts. These fold from the area above Nyquist to area below Nyquist, so they may fold into audible range (and with enough low sample rate like 44.1 or 48k most probably they will do).

If they do fold into the audible range, the only way to remove those aliasing artifacts would be to use an apodising filter whose cut-off frequency is lower and well into the audible range. I don't think anyone in their right mind would do that.

 

To 'apodize ' means to 'chop off the leg'. But if the filter starts cutting off within the audible band, it is more akin to chopping off the head for fear of a headache.

 

[Zaphod Beeblebrox]

2 minutes ago, Miska said:

But I don't have time or motivation to go through all of that.

You can dance around it all you want, the fact remains, it is all beyond the audible range, and your whole design hinges on being able to hear above the audible range, I have wasted enough time myself entertaining this.