[Currawong]

On Chord DACs I’m a fan of using TPDF for dither. Jussi, the dev of HQP, says LNS dither options are for r2r DACs.

Also I think sinc-long is def worth an audition.

That sounds backwards. Have you got a link to an exact quote?

 

[taipan254]

Hi Amos / @Currawong - big fan of your channel.

Here's a back and forth between the PGGB Developer (Zaphod Beeblebrox) and Jussi (Miska) on AudiophileStyle.com.
https://audiophilestyle.com/forums/...ew-of-math-and-magic/page/62/#comment-1239278


I've pasted the relevant parts below:

On 4/16/2023 at 8:46 PM, Zaphod Beeblebrox said:


If that is the case why even provide Noise shaping options like LNS15? 24bit dither should be good enough?

On 4/16/2023 at 9:10 PM, Miska said:





It is designed to linearise R2R ladders. In such cases word lengths are much lower than 24-bit. And curve is designed such way that in-band noise floor stays below analog noise floor and out of band noise floor is also as low as possible. But it still removes the ladder non-linearity distortions.




Like here, Holo Spring 3 at 24-bit TPDF:






And with LNS15 20-bit:






Noise floor is limited by analog domain, but the distortion got removed.




If the output is to a delta-sigma DAC like most DAC chips, Chord DACs, etc, then TPDF or Gauss1 are best choices.

On 4/16/2023 at 9:24 PM, taipan254 said:





Interesting - I'll try this. I use HQP for all streaming on a Gustard X26 Pro. I just followed the manual's implication that LNS15 was designed for 16x fs and not for R2R DACs:









And, FWIW, I've been enjoying the conversation! I am at least learning! Also, I think the tenor of the conversation has been quite civil. A testament to both @Zaphod Beeblebrox's patient, academic approach and @Miska's openness to talking shop.

On 4/16/2023 at 9:45 PM, Miska said:





Assumption is that you wouldn't send PCM to an SDM DAC in first place.




There are various R2R DACs with 16x or 32x fs input capability.




Preference is of course that the DAC operates as a bit-prefect D/A converter without any further DSP.




X26 Pro is ESS based, and it is never bit-perfect. You get a bit shorter path through the DSP with DSD though. (fifth order modulators being optimal from objective point of view)

 

[sajunky]

I use PGGB as well, actually. Yes, I can imagine the issue of storage, but it's a minor inconvenience compared to all the bulk of Dan's gear driven by a laptop, in my opinion.

There is a free PGGB online extension for Foobar 2000. You need powerful PC with a lot of RAM to process millions of taps. I only tried off-line samples created by the author, they are excellent.

Still, with R2R NOS DAC which has no opamps in the audio path and the entire gear is made of non-feedback class A amplification, there is no real benefit of upsampling. This is result of my tests, with your equipment result can be different.

The current push for HQPlayer is in result of using opamps in the large signal audio path, or even a nested opamps architecture. Dynamic response of such design can be tuned in computer simulators to avoid overshots (overall harshness or temporary artefacts - sibilants), but it require a smoothed, well prepared source to avoid internal stages entering saturation. It is why you need upsampling, to avoid sterile black background during fast transients. I would rather search for a DAC + amp which do not benefit from upsampling.

In practice it means no DAC from the recommended list on the HQPlayer website.

 

[GoldenSound]

which has no opamps in the audio path and the entire gear is made of non-feedback class A amplification, there is no real benefit of upsampling

None of that is relevant to upsampling.

Whether or not that product/design is preferred or good/bad, those do not affect the same things as upsampling

 

[GoldenSound]

The current push for HQPlayer is in result of using opamps in the large signal audio path, or even a nested opamps architecture

It has absolutely nothing to do with this

 

[kennyb123]

I think @kennyb123 uses poly with pggb files for his on the go set up. I’d imagine storage becomes an issue. Instead of cds you’ll have sd cards!

Yes I do. This helps:

10 Slots Micro SD Card Case Holder Storage Organizer, Ultra Slim Credit Card Size Lightweight Portable TF MSD Memory Card Storage https://a.co/d/czcnJVk

 

[kennyb123]

I've pasted the relevant parts below

Thanks for sharing. It sure would be great if HQPlayer had its own website and helpful tips like what you shared were posted there.

(sarcasm intended)

 

[taipan254]

Thanks for sharing. It sure would be great if HQPlayer had its own website and helpful tips like what you shared were posted there.

(sarcasm intended)

Agreed. Getting useful info from Jussi can be like pulling teeth. And it really shouldn’t be that way. His program is incredible and clearly Jussi is extremely talented. A basic user guide that explains the circumstances under which you use what settings would be killer.

 

[kennyb123]

Getting useful info from Jussi can be like pulling teeth. And it really shouldn’t be that way.

It’s almost like it’s intentionally designed so no one can come up with the optimal settings on their own. Here’s what the help for Embedded HQP 4.35.0 says about LNS15:

15th order linear noise-shaping. Smooth noise shaping curve designed specifically for 16x rates (705.6/768 kHz). Can be also used at 8x rates.

I use HQP for 16x rates so this is why I used LNS15.

Here’s what it says for TPDF:

Triangular Probability Density Function. This is the industry standard simple dither mechanism. Suitable for any rate and recommended if playback rate is 44.1/48 kHz. Recommended for general purpose use.

No way that one seeking to scale to 16FS could possibly arrive at TPDF is the best choice for their non-R2R DAC.

 

[taipan254]

It’s almost like it’s intentionally designed so no one can come up with the optimal settings on their own. Here’s what the help for Embedded HQP 4.35.0 says about LNS15:

15th order linear noise-shaping. Smooth noise shaping curve designed specifically for 16x rates (705.6/768 kHz). Can be also used at 8x rates.

I use HQP for 16x rates so this is why I used LNS15.

Here’s what it says for TPDF:

Triangular Probability Density Function. This is the industry standard simple dither mechanism. Suitable for any rate and recommended if playback rate is 44.1/48 kHz. Recommended for general purpose use.

No way that one seeking to scale to 16FS could possibly arrive at TPDF is the best choice for their non-R2R DAC.

Yeah. He does make the implicit assumption that ESS AKM and Chord DAC owners are using dsd.

I was surprised he even decided to design pcm filters for Ferrum given his all out preference for dsd!

 

[gregorio]

It’s almost like it’s intentionally designed so no one can come up with the optimal settings on their own.

That’s pretty much the whole point of it. There would be no point in creating software with all these options if the “optimal settings” were the goal. The goal/point of the software is to provide consumers with a wide range of non-optimal settings to play with.

Here’s what it says for TPDF:
Triangular Probability Density Function. This is the industry standard simple dither mechanism. Suitable for any rate and recommended if playback rate is 44.1/48 kHz. Recommended for general purpose use.

Except that the industry doesn’t use a “simple dither mechanism” so it hasn’t been a “standard” for 30 years or so. And, as the playback rate of DACs is almost never 44.1/48kHz, how can it be “recommended for general purpose use”?

G

 

[Atriya]

deleted

 

[kennyb123]

I just started using Poly with PGGB files. How do you create the MPD playlists that the Chord "gofigure" app needs? I noticed that the app can't show the playlist contents: so navigating to a track takes multiple presses of the "next track" button.

I don't use GoFigure to play the tracks. I use either Rigelian or JPlay as these just allow me to browse and then play what's on the card.

I'm guessing you're using WAV

Yes. Occasionally the aforementioned apps have trouble reading metadata. The solution that seems to work every time is to use a tag editor and remove embedded artwork (after first saving it in the same folder as the songs with the name cover.jpg). It's an odd quirk of both MPD and Squeeze that it is occasionally necessary to do this in order to have the tags displayed correctly.

 

[Atriya]

I don't use GoFigure to play the tracks. I use either Rigelian or JPlay as these just allow me to browse and then play what's on the card.


Yes. Occasionally the aforementioned apps have trouble reading metadata. The solution that seems to work every time is to use a tag editor and remove embedded artwork (after first saving it in the same folder as the songs with the name cover.jpg). It's an odd quirk of both MPD and Squeeze that it is occasionally necessary to do this in order to have the tags displayed correctly.

I realized that MPD supports Wavpack. Using 768/32 .wv files I don't seem to be getting the tag issues I had with .wav. PGGB encodes .wv directly when the option is enabled. Files are smaller too.

 

[Redflamingolingo]

I recently got a used Holo May KTE with my coming from a Chord Qutest, Chord TT2 with M Scaler, and eventually Chord DAVE with M Scaler.

My chain is QoBuz -> Roon -> USB AudioQuest Carbon Cable -> Holo May in NOS mode with PLL -> Auris Nirvana tube rolled -> Susvara.

I've been messing with HQ Player in my chain as well: 1.5 mil upscaling PCM, DSD 1024 upscaling, multiple decoders and filters based on reviews of the Holo May and suggested settings. My computer is very capable as well with 32gb DDR5 5000, i9-12900k, and a 4090 GPU.

Nothing, and I mean nothing, that HQ Player does sounds as good as the Holo May running NOS in exclusive mode through Roon. The HQ Player upscaling, regardless of decoder or filter or sample rate, always introduces a metallic sheen to the music and softens the impact of every single recording, even whippy electronic or rap. It's as though an artificial coating is tossed on the recording, like Vaseline, that smothers and smears fine details. Separation becomes muddy instead of concise and airy. Bass loses its punch . The music becomes... anemic.

It's interesting to hear so many people praise HQ Player for making their music sound better. I suspect the Chord DACs could have some improvement or maybe other chip-based DACs. Perhaps it's because the Holo May processes true NOS and is R2R. But many people who reviewed the May preferred it in HQ Player.

I just do not hear it. At. All.

NOS without any upsampling is superior to me, and it isn't even close.

Has anyone else had this experience?