A toast to PGGB, a heady brew of math and magic

Dumb question so forgive me. Does output stage matter for a dsd DAC? 
 

The heavy feedback sterility of topping products would likely still deter me from considering this product. But maybe my assumption is incorrect here. 

On 3/31/2023 at 1:13 AM, ajm said:

All this was done with 256 bit processing and I found that, in particular, choral and piano renderings sounded more tonally natural/ realistic with dither only. With noise shaping, I had a feeling that something had been shaved from the sound.

I'm starting to lean towards 256+dither camp more and more as well. I've been going back and forth, but somehow I always come back to dither only. I can easily see why people like noise shaping more. It's cleaner and somehow it keeps colors better. Everything is just... right. But there is something. It's almost like it's too "right", unnatural.

Noise shaping tends to do the same effect for me what TA-ZH1ES did when I had it. That clarity was amazing and it represented space so beautifully. But I had a problem with it: I never tapped my toe. It feels like noise shaping and TA-ZH1ES both shave off the dirt, but baby tends to go with the bath water. Musical presentation becomes more coherent and clean, but with dither only there is some kind of incisiveness in small details. Yes, with dither I lose some trailing edge information in decay, which affects the depth perception, but especially with 32bit on Mojo, that digital noise floor already gives a lot of room. I also prefer dither with 24bit on May.

To me it's a trade-off between transient micro information vs. depth of the void / coherence of the background. The depth of the void is something I really, really want, but that transient micro information is something that is needed to keep me interested and focused. I'm also aware that the transient micro information may well be just about small inaccuracies, which make transients artificially edgy. However, I still like it.

Also something to note: I do not like TPDF dither in Roon nor in HQP. However, I do like gaussian dither in HQP and I've understood that PGGB also uses gaussian dither.

8 hours ago, taipan254 said:

Dumb question so forgive me. Does output stage matter for a dsd DAC? 
 

The heavy feedback sterility of topping products would likely still deter me from considering this product. But maybe my assumption is incorrect here. 

My response was purely from digital a signal processing perspective on whether the DAC will benifit from DSD upsampling. Every aspect of the DAC design matters, so we have to wait to see if it actually sounds good.

5 hours ago, rayon said:

I'm starting to lean towards 256+dither camp more and more as well. I've been going back and forth, but somehow I always come back to dither only. I can easily see why people like noise shaping more. It's cleaner and somehow it keeps colors better. Everything is just... right. But there is something. It's almost like it's too "right", unnatural.

If you like 256bit + dither, you should try 64bit + dither and also 128bit + dither. Without noise shaping, there is very little justification to do higher precision.

With R2R DACs the reason to do noise shaping is two folds. 1. To linearize the last few bits 2. To take advantage of high precision computing IMHO you will be giving up even more.

2 hours ago, Zaphod Beeblebrox said:

If you like 256bit + dither, you should try 64bit + dither and also 128bit + dither. Without noise shaping, there is very little justification to do higher precision.

Yes, I've been thinking about that and done some tests. With 32bit PCM the difference between 256 and 64 is there for Mojo 2 (though it wasn't proper blind test, but me always knowing which one I listen). 256 sounds more colorful, textured and lively. I could do 128bit comparisons as well, but as I would already take that huge penalty when going from 64 to 128, I've been using 256 for peace of mind. Speed wise 64 is in it's own league (due to obvious reasons). Currently I do 64 first as that can be done in seconds and then replace with 256.

2 hours ago, Zaphod Beeblebrox said:

With R2R DACs the reason to do noise shaping is two folds. 1. To linearize the last few bits 2. To take advantage of high precision computing IMHO you will be giving up even more.

Yes, I've noticed that with May the difference is less noticeable. I also enjoy May more @ 24 bits, of which the last bits are not very linear. But I've also noticed that with May I enjoy most 4fs. It's punchy and dynamic, but cleans up higher frequencies a bit. 4fs with dither doesn't take very long anyway, even with 256. I should do proper blind testing with May to see if it makes much sense to do anything else than 64 processing going forward when doing PCM. Based on the little testing I've done, I've even felt a few times that I actually prefer 64 over 256 on dithered 4fs.

24 minutes ago, Zaphod Beeblebrox said:

I plan Is to introduce a 108 bit option for PCM that would be nearly as fast as 64bits, my hope is for the Foobar plugin to also get that upgrade for real-time PCM upsampling. No ETA though.

 

The change will coincide with a major overhaul to PGGB UI, I will be moving away from Matlab to a more modern cross platform interface.
 

I will post some preview screenshots when it is more mature.

 

This move, In the long term, would allow me to incorporate real-time upsampling for streamed content.

That's amazing news! Have to do some 128 testing to get a rough feeling on where that lands with dither SQ wise. It may well be that 108 would be close to a point where it "doesn't matter" anymore. Not that the speed would be a problem anymore anyway for processing locally (or then my patience has been trained into steel when processing DSD1024 with one pass), but streaming would be sooo nice, especially if that could be used with Roon some day. Currently mixing PGGB'd files bit perfect and upsampling only streamed content (with headroom) is mission impossible.

On 9/10/2024 at 6:05 AM, rayon said:

Btw @Zaphod Beeblebrox I may have found a "bug". I have a problem when playing PGGBd wav files using MPD with Chord Poly. When I play flac files from SD card, everything works just fine, but if I play wav files and try to seek within a track into a different time point, MPD throws me "not seekable" -error.

I just attempted this myself.  I don't get an error thrown, but I found it not possible to try to skip to a different time point when playing PGGB/WAV files.  

On 9/13/2024 at 4:56 PM, kennyb123 said:

I just attempted this myself.  I don't get an error thrown, but I found it not possible to try to skip to a different time point when playing PGGB/WAV files.  

Was this the case with any wav files or just PGGB wav files? Also wonder how much this a function of size of the file is.

2 hours ago, Zaphod Beeblebrox said:

Was this the case with any wav files or just PGGB wav files? Also wonder how much this a function of size of the file is.

I haven't yet tried with other wavs than PGGB generated personally, but at least with PGGBd files the size didn't matter. I had the same problem with 192/24 and 768/32 (and some other options in between).

On 9/16/2024 at 7:34 AM, Zaphod Beeblebrox said:

Was this the case with any wav files or just PGGB wav files? Also wonder how much this a function of size of the file is.

Only PGGB-generated WAV files on my SD cards.  I will try to remember to load some other WAV files onto one so I can try with Poly.  I also suspect size may have something to do with it.

Just received the newly-released Analogue Productions SACD of Stone Temple Pilots' classic Core. I also have a 24 / 96 version. I've ripped and gargle-blasted both to 24 / 16fs with PGGB at 128 bits of precision to my Mojo 2 / Preamp / Head Amp / ZMF Caldera. 

I remain SO IMPRESSED with DSD source material vs. PCM source material of the same recordings after gargle-blasting to PCM with PGGB. There's some level matching issues, and I'm not doing a blind A / B / X, so take my opinion with a grain of salt, but the DSD source material sounds so much more relaxed, nuanced, textured, harmonically correct, and unharsh / "less edgy" compared to the PCM source material. 

Another fun DSD vs. Hi Res PCM comparison is Soundgarden's Superunknown. My observations are consistent between both of these recordings. 

I clearly do not want to make a blanket statement on DSD Source Material vs. PCM Source Material, but I am glad I am investing in SACD ripping. 

I look forward to eventually getting a solid, discrete DSD DAC so I can experience the full DSD 512 PGGB experience (I haven't even hit PCM at 256 bit precision yet!). 

In that same vein - has anyone experimented with the Holo Audio Cyan 2 and evaluated its PCM and DSD performance (subjectively or objectively)? 

18 hours ago, Zaphod Beeblebrox said:

Still work in progress

 

will this version be native on apple silicon?

8 minutes ago, pavi said:

 

will this version be native on apple silicon?

Not on release but it could. It will depend on the demand for native version, as porting the high-performance multi-precision library to ARM is significant effort, the current library for DSD on ARM even in emulation already closes the gap (between native and emulation) and only about 25 - 30% gain can be expected. On the other hand, for PCM the expected improvement is about 50%.

Wow, realtime upsampling for streaming content would be amazing! I'm on the edge of my seat waiting for this feature!!

Any new update for PGGB-RT cooking?

On 9/23/2024 at 11:58 PM, taipan254 said:

Just received the newly-released Analogue Productions SACD of Stone Temple Pilots' classic Core. I also have a 24 / 96 version. I've ripped and gargle-blasted both to 24 / 16fs with PGGB at 128 bits of precision to my Mojo 2 / Preamp / Head Amp / ZMF Caldera. 

 

I remain SO IMPRESSED with DSD source material vs. PCM source material of the same recordings after gargle-blasting to PCM with PGGB. There's some level matching issues, and I'm not doing a blind A / B / X, so take my opinion with a grain of salt, but the DSD source material sounds so much more relaxed, nuanced, textured, harmonically correct, and unharsh / "less edgy" compared to the PCM source material. 

 

Another fun DSD vs. Hi Res PCM comparison is Soundgarden's Superunknown. My observations are consistent between both of these recordings. 

 

I clearly do not want to make a blanket statement on DSD Source Material vs. PCM Source Material, but I am glad I am investing in SACD ripping. 

 

I look forward to eventually getting a solid, discrete DSD DAC so I can experience the full DSD 512 PGGB experience (I haven't even hit PCM at 256 bit precision yet!). 

 

In that same vein - has anyone experimented with the Holo Audio Cyan 2 and evaluated its PCM and DSD performance (subjectively or objectively)? 

I have the Cyan 2 and did some initial listening with PCM1.5 and DSD512 & 1024. All sounded great of course. I found myself preferring often preferring PCM for some reason, but this should be taken as a first impression only. Between 512 and 1024 I couldn't really choose.

Recently @GoldenOne, who has been doing some testing of the Cyan 2, informed me that the DAC performs better from a technical standpoint with DSD512 than with 1024.