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

[MarcelNL]

I have dabbled with DXD a long time ago, but the system I'm now listening to is completely different so that experience is not valid IMO..

At the time DXD seemed to have its merits.

[seeteeyou]

I2S in and of itself ain't a problem at all since it's able to reach 32FS for sure, though something else was definitely holding it back

 

https://audiophilestyle.com/forums/topic/30376-a-novel-way-to-massively-improve-the-sq-of-computer-audio-streaming/page/574/?tab=comments#comment-967495

On 6/22/2019 at 12:14 PM, numlog said:

The CM8888 PCIe to I2S chip only supports 192kHz.

 

 

Metrum Amethyst is only capable of accepting 8FS via USB while CM8888 inside Pink Faun I2S Bridge is limited to 4FS, did you realize what's going on behind the scenes?

 

Upsampling to 8FS (PGGB)

↓

Downsampling to 4FS (CM8888 inside Pink Faun I2S Bridge)

[MarcelNL]

it's pretty much why the first batch of files was PGGBed ending up as 176 OR I am mistaken and got the interpretation of the file details wrong. I am quite familiar with the 192 Khz limit.

 

[Vee]

Hopefully not off topic, but for those of us looking to apply a best match profile for streaming Qobuz... is there a set of equivalent filters in HQP that can be applied to mimic PGGB? 

 

I will be trying PGGB for my offline files too.

[Zaphod Beeblebrox]

I just released version 2.1.05 of PGGB . Thanks to continued support and feedback, I have  updated noise shapers used by PGGB (to further reduce the noise floor in the audible range) and I have also added an option to 'Dither only' which will be helpful for those who wish to further process gargle-blasted tracks and noise shaping may not be ideal. Please see here for the full release notes. 

 

For those who are already using PGGB, here is a cheat sheet that may help you:

• If you are playing at 16fs output at 24bits - changes affect your 44.1k and 48k albums
• If you are playing at 16fs output to 32bits - changes  your entire library
• If you are playing at 32fs output at 24bits or higher -  changes affect your 44.1k and 48k albums
• If you are playing at 8fs output at 32bits - changes your 44.1k and 48k albums
• If you are setting your HF Noise filter to 'Full', please revisit Full/Moderate/Minimal to check if your preference has changed

 

[ASRMichael]

2 hours ago, Zaphod Beeblebrox said:

I just released version 2.1.05 of PGGB . Thanks to continued support and feedback, I have  updated noise shapers used by PGGB (to further reduce the noise floor in the audible range) and I have also added an option to 'Dither only' which will be helpful for those who wish to further process gargle-blasted tracks and noise shaping may not be ideal. Please see here for the full release notes. 

 

For those who are already using PGGB, here is a cheat sheet that may help you:

• If you are playing at 16fs output at 24bits - changes affect your 44.1k and 48k albums
• If you are playing at 16fs output to 32bits - changes  your entire library
• If you are playing at 32fs output at 24bits or higher -  changes affect your 44.1k and 48k albums
• If you are playing at 8fs output at 32bits - changes your 44.1k and 48k albums
• If you are setting your HF Noise filter to 'Full', please revisit Full/Moderate/Minimal to check if your preference has changed

 

Thanks ZB. What about 8fs @ 24bit? 

[Zaphod Beeblebrox]

11 minutes ago, ASRMichael said:

Thanks ZB. What about 8fs @ 24bit? 

No change.

 

The reason is there is not much leeway for relatively low sample rates to push the noise floor down without incurring increased out of band HF noise that could affect SQ.

 

[edwardsean]

Okay, the big question:

 

If changes effect my use-case (16fs/24bit from 48K for me) is it worth it to re-gargle-blast my entire library?

[Zaphod Beeblebrox]

15 minutes ago, edwardsean said:

Okay, the big question:

 

If changes effect my use-case (16fs/24bit from 48K for me) is it worth it to re-gargle-blast my entire library?

 

Only you can answer that question, I plan to redo it on my CD/48k albums, good news is those fly fast compared to DSD and higher rates.

 

To be clear - for 16fS/24bits, the changes affect only CD and 48k rates.

[ASRMichael]

4 hours ago, Zaphod Beeblebrox said:

No change.

 

The reason is there is not much leeway for relatively low sample rates to push the noise floor down without incurring increased out of band HF noise that could affect SQ.

 

So effectively one would need to convert at 32bit to achieve lower floor noise? Albeit I would be back to truncating 32bit to 24bit, which is not recommended. But off course one could try? 

[Zaphod Beeblebrox]

38 minutes ago, ASRMichael said:

So effectively one would need to convert at 32bit to achieve lower floor noise? Albeit I would be back to truncating 32bit to 24bit, which is not recommended. But off course one could try? 

No, you should not do that. Since your DAC uses only 24bits, noise shaping to 32bits and then letting your DAC  truncate back to 24 will be a step backwards  compared to just using noise shaped output to 24bits. You will not get any benifit of the new 32bit noise shaper.

 

Edit: Noise shaping affects the last few bits; truncating to 24bits you already lost anything that has been done at 32bits.

[edwardsean]

Okay, after some A/B comparisons, I’m re-converting my library in batches—using 2.1.05. 

 

To my ears, yeah, it’s worth it. It takes the things I love about PGGB that one step further. 

 

I guess I really am going to get used to periodically reconverting my whole library. 

[austinpop]

It's a definite uptick in SQ. I've redone all my 1FS albums to 24/16FS using v2.1.05. 

 

Luckily, PGGB is blazingly fast for 1FS files, so it goes relatively quickly.

[HeeBroG]

@austinpop

Hi Rajiv. I am new to PGGB on my Win10 PC with 64GB DD4 2666MHz RAM.

I have configured the Virtual RAM with a further 200GB as you suggested in your post Jun 8.

For a 1FS album at 24/705.6, what is considered blazingly fast so I have a reference. 

 

Thanks

 

Geoff

[atxkyle]

I concur with @edwardsean and @austinpop that its worth the trouble to re-gargle-blast with the new rev. Remastering with this latest release is like removing another thin veil you didn’t realize was there.  

[Zaphod Beeblebrox]

5 hours ago, HeeBroG said:

@austinpop

Hi Rajiv. I am new to PGGB on my Win10 PC with 64GB DD4 2666MHz RAM.

I have configured the Virtual RAM with a further 200GB as you suggested in your post Jun 8.

For a 1FS album at 24/705.6, what is considered blazingly fast so I have a reference. 

 

Thanks

 

Geoff

It depends on the length of the track on CDs but, if you have at least 32GB of RAM and a fast drive for virtual memory, you can expect 3x or higher rate. So a typical CD album may take 15 - 20 minutes to process. PGGB creates a csv file inside your output album folder (pggb_album_analysis_v5.csv),  If you look at the column 'Time ratio', it gives you an idea of how fast it was processed. For example, a number 4 implies, the track got processed at 4x speed (i.e, a 4 minute track took one minute to process).

[austinpop]

@HeeBroG

 

Just as @Zaphod Beeblebrox explained above, my 1FS albums process at rates between 2.5x to 3.5x, depending on track lengths.

[happybob]

On 8/9/2021 at 6:08 AM, Zaphod Beeblebrox said:

I just released version 2.1.05 of PGGB . Thanks to continued support and feedback, I have  updated noise shapers used by PGGB (to further reduce the noise floor in the audible range) and I have also added an option to 'Dither only' which will be helpful for those who wish to further process gargle-blasted tracks and noise shaping may not be ideal. Please see here for the full release notes. 

 

For those who are already using PGGB, here is a cheat sheet that may help you:

• If you are playing at 16fs output at 24bits - changes affect your 44.1k and 48k albums
• If you are playing at 16fs output to 32bits - changes  your entire library
• If you are playing at 32fs output at 24bits or higher -  changes affect your 44.1k and 48k albums
• If you are playing at 8fs output at 32bits - changes your 44.1k and 48k albums
• If you are setting your HF Noise filter to 'Full', please revisit Full/Moderate/Minimal to check if your preference has changed

 

Cool!

 

Question: If I'm going from server > SRC-DX (via USB) > MScaler (via dual BNC) > Dave (via Dual BNC), should I enable dithering with this new 2.1.05 PGGB or not? Earlier you had clarified that the best option for this (temporary for me) setup is to do 32 bit 16FS upsampling with NO noise shaping (given that the MScaler always noise shapes to 24 bits no matter what it's sent as an input even if it doesn't upsample). But I just wanted to clarify whether or not to do dithering only. Likely nothing at 32 bits matters given the 24 bit truncation of the SRC-DX.

 

Also, since I'm using my M1 Macbook Pro, should I still stick with 1 worker or does is this new version more optimal with extra workers? 

Thanks for clarifying!

[Zaphod Beeblebrox]

10 minutes ago, happybob said:

Cool!

 

Question: If I'm going from server > SRC-DX (via USB) > MScaler (via dual BNC) > Dave (via Dual BNC), should I enable dithering with this new 2.1.05 PGGB or not? Earlier you had clarified that the best option for this (temporary for me) setup is to do 32 bit 16FS upsampling with NO noise shaping (given that the MScaler always noise shapes to 24 bits no matter what it's sent as an input even if it doesn't upsample). But I just wanted to clarify whether or not to do dithering only. Likely nothing at 32 bits matters given the 24 bit truncation of the SRC-DX.

 

Also, since I'm using my M1 Macbook Pro, should I still stick with 1 worker or does is this new version more optimal with extra workers? 

Thanks for clarifying!

Because v2.1.05 now have the dithering option, I would suggest using 24bit 'Dither Only'  via SRC-DX->HMS. You could use the same file with Roon too (if you want to have just a single copy). 

[happybob]

4 minutes ago, Zaphod Beeblebrox said:

Because v2.1.05 now have the dithering option, I would suggest using 24bit 'Dither Only'  via SRC-DX->HMS. You could use the same file with Roon too (if you want to have just a single copy). 

So does the MScaler processing (24 bit noise shaping and maybe dithering?) not interfere with the PGGB 24 bit dithering?

 

Also, I do use these files to playback on other systems via Roon and unfortunately am forced to convert everything to 192K sample rate. I was using 32 bit no noise shaping PGGB upsampling so that the extra bits could be used by Roon for optimal sample rate conversion to 192K (sometimes not power of 2 conversions), but perhaps 24 bit dithering only is close to the same as 32 bit no dithering or noise sampling even for Roon? I thought that Roon would also do dithering on sample rate changes...  

 

I know these are likely small changes but if I'm going to reprocess everything it would be nice to be optimal SQ for all my system flows.

 

Thanks!

[Zaphod Beeblebrox]

28 minutes ago, happybob said:

So does the MScaler processing (24 bit noise shaping and maybe dithering?) not interfere with the PGGB 24 bit dithering?

Because you are using SRC-DX that is limited to 24 bits you have three choices:

1. Do not dither or noise shape, just truncate to 24bits - truncation is never a good idea and the only reason I suggested it previously is because PGGB did not have a dither only option for 16FS.
2. Noise shape to 24bits - If playing direct to DAVE this would have been a good option, but for further processing by HMS, it seems to degrade SQ (but do try this  yourself and let us know)
3. Dither to 24bits: Though this will have a higher noise floor compared to noise shaping, it is more friendly for further processing by HMS.

You do have a fourth option which is dither to 32bits, but in this case SRC-DX will truncate it 24bits. I will not recommend  this because I assume you would want the best quality out of SRC-DX->HMS->DAVE over Roon downsampling.

 

Without having two sets of files, there is no way to optimize this for all your setups (when using PGGB offline upsampling).

 

 

[Peti]

I have been using the demo version of PGGB for a few days now so here's a couple of comments/observations. In medias res - it does indeed work in my case/to my ears. Get this: for fun, I started to run some cherished redbook audio material of mine through PGGB at 24/176.4 b/c I have an ancient Fiio X3 (yup, the very first Fiio dap) that handles audio up to 24/192. Compared to the redbook, untreated tracks, the PGGB version sounds more open and it feels like an audio magnifying glass was put right in front of the music - details galore. while keeping coherence within the layears of the track. Keep in mind, I am talking about listening straight from the Fiio dap into a pair of Sennheiser IE800S earphones. So I'm pretty much sold on this software.

 

Soon I will start to PGGB treat some albums to be listened on my main headphone rig that entails a Chord Mojo dac. In that case I should set 32 bits 768000hz? (See my full rig

 

I am using not the newest version of this software but the one before and I seldom mess with the filter, etc. settings. Not sure if I can get my hands on the newest rendition of this software while on my trial period...?

 

Also, as an audio OCD, I keep thinking about how to make my system to sound even better, and the thought occurred to me, that what if I first treated an album with the PGGB 32/3072 setting for maximum results and then run the same file again in PGGB, but this time at the 32/768 setting for the Chord Mojo?

 

One suggestion: when doing the 24/192 conversions I set flac as the output file format and it'd be useful if one could set the compression level of it. I believe there are options from 0 to 8, zero being practically uncompressed. I would use setting 8, to save even more space.

 

Ceterum Censeo, I find this software a major discovery in my audio journey!

[kennyb123]

15 minutes ago, Peti said:

Soon I will start to PGGB treat some albums to be listened on my main headphone rig that entails a Chord Mojo dac. In that case I should set 32 bits 768000hz?

Will you be using a Poly?  

[Peti]

No, please see my full rig in my profile.

[kennyb123]

2 minutes ago, Peti said:

No, please see my full rig in my profile.

Ah sorry.  When I clicked over there I didn’t see the Audio System tab as I was holding my iPad in portrait mode.  Yeah 32 bit and 16 FS should be the way to go.  Be sure to set HQPlayer to passthrough (none for 1x, Nx and dither).