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

[ray-dude]

 

With the recent release of the upgraded PGGB, there are some new links

 

Please see here for PGGB 256 announcement:

https://audiophilestyle.com/forums/topic/62699-a-toast-to-pggb-a-heady-brew-of-math-and-magic/?do=findComment&comment=1227948

 

 

Here are updated PGGB 256 Links:

 

• Main site: http://www.remastero.com
• PGGB: http://www.remastero.com/pggb.html
• FAQ: http://www.remastero.com/faq-256.html
• Guide: http://www.remastero.com/guide-256.htm

 

 

 

*******************************************************

(original post)

 

 

With the embargo (finally!) lifted, wanted to start a topic to share impressions and feedback on the new PGGB music remastering/upsampling program.

 

• Main site: http://www.remastero.com
• PGGB: http://www.remastero.com/pggb.html
• FAQ: http://www.remastero.com/faq.html
• Guide: http://www.remastero.com/guide.html
• Onboarding: http://www.remastero.com/onboarding.html

 

The project started 6-7 months ago to do some experiments in upsampling music, and has grown into something WAAYY beyond anything any of us would have thought possible.  There are ~25 folks that have joined the journey from dev to alpha to private beta. With the impending public launch, very much looking forward to discussing PGGB as more folks are able to hear what has been hiding in their digital content all along.

 

Thank you @Zaphod Beeblebrox for pushing the math and sound quality to its ultimate limits!

 

(Disclaimer: although I'm delighted to have been helping on the project since basically day one, I have no financial interest in the project, I paid full price for my license, etc.)

[ray-dude]

12 minutes ago, ASRMichael said:

Is this effectively another HQPlayer? What are the benefits? 

 

I view it as complementary to HQPlayer.  While HQPlayer is fantastic for streaming content, PGGB is preprocessing (effectively remastering) music prior to playback.  That opens up a level of processing and compute that just isn't practical in a streaming context (note that HQPlayer Pro is capable of doing preprocessing of content, my comments are with respect to the more commonly used HQPlayer).  I have been and continue to be a delighted HQPlayer customer, even with PGGB.

 

The project originally started as some experiments to see how far you could push idealized filters for reconstruction (billions of taps vs millions).  The results were stunning, which led to a very long and VERY rewarding series of optimizations.  When used optimally, PGGB is remastering digital content to be aligned as much as possible to the DAC stage of your DAC (think NOS PCM DAC).  Results will vary from DAC to DAC, but with my Chord DAVE, the results are extraordinary and a complete game changer.

 

The website linked above has more info.  Hopefully the folks that have been in the private beta can share some first hand impressions

 

 

[ray-dude]

I can say 100% that Roy, Rajiv, and I have had zero compensation, and have zero financial interest of any kind in this project (for example, we all paid full price for our licenses).   I appreciate the sensitivities behind the question though.  I do not consider myself an ambassador, but more of a delighted early adopter and user.  Anything I post is representing me and my personal opinions, and not PGGB or any other vendor.  I am happy to share my experiences though.

 

As to the other questions @Zaphod Beeblebrox is on AS, so he's best to answer the other questions, but I believe the licensing scheme uses internet access (I'm not aware of any other use of internet access...ZB can verify)  

 

I know ZB is a strong advocate of trial licenses (he supported them during the private beta) so as the product formally launches, I'm pretty sure he will continue to do so.  Note that PGGB requires significant amounts of memory to run, so most folks don't have suitable hardware easily available to them.  Most PGGB users have been happy to help process sample tracks for folks.

 

 

 

[ray-dude]

 

Lot to unpack here.  A couple quick comments

 

* There is literally no engineer working in and around audio that I respect more than Rob Watts. I have owned 4 of his DACs, and will order his next DAC one day 1 preorders are allowed.

* The Whittaker-Shannon theory is as foundational as it gets for modern information theory.  It continues to be curious why more folks aren't leveraging it.  There is no IP/secret sauce here, just solid engineering and math fundamentals.

* Jussi clearly focuses on DSD more than PCM.  He has been very supportive of his PCM upsampling customers asking for various enhancements, and I'm very appreciative that he added the various sinc filters to HQP, and the LNS15 noise shaper (I happily use these for streaming).  He does tremendous work, and being a good engineer with a rrobust understanding of fundamentals and supporting your customers most definitely does not make you a copy cat.

* Having carefully compared Hugo mScaler (Rob's 1M tap upsampler, which I adore...see my review of mScaler here) with HQPlayer running sinc-M and LNS15, to my ear HMS is clearly better with Chord DAVE (I can not speak to other DACs).  There is more than applying 1M taps and a 15th order noise shaper. The details matter, and tuning to your target DAC matters.

* Having heard upwards of 1B tap processed files 6+ months ago, it was a huge step forward from 1M taps, but it can't even hold a candle to where PGGB is today. Rob is very open about using solid engineering fundamentals (1M taps based on Whittaker-Shannon theory), but I can say definitively with no equivocation that the secret sauce is NOT in the taps.  The taps are permission to play (and an easy thing to hang a marketing position on), but Rob's genius is his relentless tuning of the pipelines and filters so they sound the best they can possibly sound.  He does not publish that.  Having gone through this process for 6 months now with PGGB, it is grossly unfair to call what ZB has developed as copy cat.

 

With all that being said, I would ask that we keep things informational and share impressions.  Jussi/Rob/ZB have no need to defend themselves for solid engineering work, and if HQP or mScaler or PGGB or a stiff shot whiskey give you more enjoyment from music, that is worth sharing and celebrating.  Please, there is no need to impugn others, or ascribe intent or bad faith to people.  

[ray-dude]

13 minutes ago, Johnseye said:

@ray-dude, @romaz and @austinpop here I thought you guys were just kicking back and enjoying your Extremes.  Great to hear you're continuing to explore new improvements.  Thanks for all your efforts, and of course thanks to Zaphod.  I'm looking forward to hearing how this impacts the music.  I'll be checking this out with the Hugo TT2 and Omegas as well as with the new Holo May in my main listening environment.  I know you guys are huge DAVE fans, but if you ever get the chance to listen to the May, I'd love to know what you think.  I'm very impressed so far and it's just breaking in.

 

I now have the PGGP trial and have made a bunch of files for comparison of your different "preferences".  I had been enjoying real time upsampling with HQP's sinc-L filter and LNS15 shaper.

 

 

Between TT2+ Omegas at 16fs/32bit and the May at 32fs, that will be quite the shoot out!  Can't wait to hear what you hear!

[ray-dude]

My listening pattern has very much changed with PGGB.  I am back to my vinyl days.  I drop a PGGB album into the player, and disappear into it.  If I get a hankering for an album that I haven't processed, I drop it for processing into my PGGB machine, and by the time I finish with the album I'm listening to, it is ready to go

 

This is a wonderful way to take a random walk through my music collection, really listening to albums again.

 

I do keep copies of all my PGGB albums on my NAS as well.  If I don't have an album I want on my music server, just copy it over, and hit play.  Really loving this back to the future experience, like when I had my "active" 50 albums staged by the turntable, with the wall of albums off to the left, staging more albums out as the music took me there.

[ray-dude]

7 hours ago, HeeBroG said:

To the early adopters,

 

With PGGB, is there still a need to purchase hi-res files?

 

In other words, would PGGB of 24/192 when available be superior to 16/44?

 

 

 

Geoff, I'll echo what others have said: get music as close to the native mastering format as possible, and process to as close as the native format of your DAC as possible (in the case of our Chord DAVEs, 16fs 32 bit PCM)

 

In digital processing, every time you change sample rate or bit depth, some of the original information is lost.  Alas, we don't have control of what happened during mastering and down conversion to redbook - it could have been done carefully and well, it could be done poorly.

 

As a thought experiment, take a favorite high res file.  Use your favorite conversion package and decimate it to 1fs 16 bit.  Take your favorite upsampling package and upsample it to the original high res sample rate and bit depth.  Repeat this 5 times.  Compare the resulting high res file to your original, and you'll have a sense of how "ideal" the down sampling and up sampling is.  If you can't hear a difference, it is effectively "ideal" for your system.  If you can hear a difference, you're hearing the result of the repeated non-ideal conversions.

 

For folks doing PGGB, PGGB does downsampling as well, using the same near ideal filters as it uses for upsampling.  You can do this experiment yourself by running PGGB on the same file over and over, going from (say) DXD format to redbook to DXD, etc.   Try the same with XLD or HQP or whatever conversion software you like.  It is very informative to hear the differences.

 

Note that differences are OK.  For example, if a digital process adds an echo effect, it can be quite pleasant (same with equalization).  Running through the same process 5 times will result in a mess, but running it through once may be awesome.  For this experiment, any difference is indicative of how non-ideal the processing is, not an indication of whether one prefers the SQ of a filter or not.

 

Addendum: if you want to isolate the effect of upsampling and downsampling as much as you can, do everything in 64 bits until the final step.  For example, if you start with a DXD file (typically 8fs 24 bit), use PGGB to convert it to 1fs 64 bit, then 8fs 64 bit, then 1fs 64 bit, etc.  On the file step, make the last conversion from 1fs 64 bit to 8fs 24bit and compare to your original file.  This will minimize the impact of bit rate changes, and let you focus on the idealized reconstruction (decimation) filter instead (the giga tap filter).

[ray-dude]

3 hours ago, hols said:

I must say it is not easy at all to find out the sweet spot of the DAC because it is very difficult to determine which one is more accurate. So this is some good experience that helps you understand your DAC more.

 

I would like to second this very important insight.  

 

A consequence of moving a lot of is typically intra-DAC processing (prior to the actual DAC stage in the pipeline) and externalizing it with PGGB is an opportunity to really get to understand how digital to analog conversion is working in your DAC.  The closer you get to matching PGGB output to what your DAC does internally, the better the result.  

 

Alas, how to define "better" is the challenge, esp. as new aspects of reference recordings start becoming audible...which is more accurate?  Which is more pleasing?  Which is more pleasing after extended listening?

 

Very interesting learning process.  Like a major component upgrade, PGGB does reveal things that may not have been apparent before

[ray-dude]

FWIW, before the Mac version was available, I was running PGGB in the free (for home use) version of VMware on my Mac Pro.  Worked like a charm.  That being said, I'm happy to be rid of Windows and running it natively on OS X now.

[ray-dude]

On USB input for Chord DAVE, 32 bit was the winner for me. With SRC-DX, 24 bit is the clear winner (and a big step up over USB direct…thanks Dan!).  That being said, all bit depths are pretty awesome, tweaking but depth is more about icing on the cake. 
 

So much about PGGB is about finding the right synergy with your DAC. The good news for me has been differences have been very easy to hear, so dialing in has been a pretty quick exercise. 

[ray-dude]

I have same issue on 2012 Mac Pro with 128GB (Intel) when processing some large DSD files.  The crash is the watchdog giving up the ghost.  I suspect the system is asking for swap but not getting it fast enough, so the watchdog barks.

 

If there was preallocated swap (like there is on Windows), we wouldn't be having this issue.

 

I do not have this issue when running Windows in VMware on the same Mac.  I preallocate the swapfile in windows, and Bob's your uncle

[ray-dude]

With my OOM issues, I was at 2B taps +.  When I manually capped off the number of cores, it ran cleanly.  Watching memory use tick up, it was pretty clear about when things tipped over.  All on Intel, so not related to a translation layer.

 

If you can dial down memory use (fewer taps, fewer cores) you can tune away from OOM zone.

 

 

[ray-dude]

2 hours ago, Zaphod Beeblebrox said:

The choice of 20 bits is based on linearity measurements that have been published. With R2R DACs, resistors need to be precisely matched and doing so to retain 24 bits of accuracy is close to impossible currently. Resistor variation would cause the DAC output to not be linear in the last few bits, this would in turn result in distortion. Zeroing out the last few bits would help but it will in turn result in increased quantization noise because fewer bits (20 instead of 24 bits) are used to represent the signal. This is where noise shaping comes in. Using a well designed noise shaper, even when only 20 bits are used,  quantization noise can be almost removed from the audible range by taking advantage of the 16FS or higher sampling rate. Zeroing out the last bits and noise shaping not only linearizes the DAC to 20 bits, but also effectively increases the dynamic range by reducing the noise floor further.

 

Even with DAVE, one of the main reasons why  going 24 bits via SRC-DX  is  still effective even though only 24bits are used instead of 32, is because noise shaping is used. Even Mscaler noise shapes to 24 bits and it is not in pass through if the input is 32 bits.

 

Fantastic post and summary.  The magic of PGGB is all these factors coming together, and control being externalized from your DAC so you can optimize digital processing in software to get the absolute most out of the actual digital to analog conversion stage of your DAC.  

 

Each DAC will have its own optimal recipe, and some will be easier to bypass their internal digital processing than others.  Hopefully over time we'll see some world class DACs that are designed from the get go to externalize all their digital processing (for example, I'd love to hear PGGB content directly feeding a Chord-like pulse element array digital to analog conversion stage to a cost no object discrete analog output stage)

 

 

 

 

[ray-dude]

These apples to apples comparisons are very difficult.  Different USB chipsets matter. Any internal sample rate conversion or bit depth conversion matters.  Isolation and RF and ground paths matter.

 

You will hear differences, but untangling what is causing the differences can be more involved than obvious at first glance.

 

That being said, if one sounds better than the other, there is much happiness to be had to just enjoy the better sounding music ;)

[ray-dude]

We did a lot of this math at the very beginning of PGGB (before it was called PGGB).  Very hard to make it work.

 

Only idea we came up with is find a buddy with symmetric fiber to the home and unlimited data and no throttling, and put a shared server in their closet (says the guy with symmetric fiber to his house and a server in his closet)

[ray-dude]

Since things are heating up again, a reminder to keep things civil, and an ask (as OP) to keep discussions focused on listening impressions, and technical context that may be helpful in understanding listening impressions.

 

There is a lot of value in debating different digital processing strategies for audio, but an impressions thread for PGGB is not the best place for that. 

[ray-dude]

(putting on my OP hat)

 

Folks, let's stay on topic with listening impressions for PGGB please (thank you)

[ray-dude]

EDIT (oops, ZB replied before I hit POST)

********

Yes, file size is the same regardless of the precision of processing.  File size is set by the target high res frequency (amount of over sampling) and bit depth.  For each DAC, there is an optimum setting for these.

 

The precision used for the math impacts the accuracy of the upsampling and degree of noise shaping that is achievable.  Most definitely you should compare 64 vs 128 vs 256 bit precision processing and see what you hear in your chain.

 

Generally speaking, if the trial mode of PGGB doesn't let you do the tests you're keen on doing, most PGGB users are happy to process some test tracks for folks.  Find the PGGB settings that are working best for you using the trial version, and ask here or in PM and a bunch of folks would be happy to process your test tracks for you at different precision using your preferred settings.

That being said, the trial mode of PGGB is pretty permissive

[ray-dude]

Crushing it.  Astonishing performance for reconstruction accuracy and noise shaping.  As close to true reference as you can get.  Congratulations on an amazing engineering achievement ZB

[ray-dude]

These are great questions, and ones that there has been a lot of head scratching over.  

 

The short answer is that there seems to be no harm in using the whole track (quite the contrary, during group listening tests there was strong consensus that more is better).  There is a question about whether combine helps at all.  The only way to really get a sense for individual recordings is to try and see (or hear) if there is a difference.  

 

As ZB said, there has been consistent benefit to using more and more of the tracks.  I will echo the need to go USB direct to DAVE if you're comparing to mScaler (and adjust DAVE volume 3 clicks down).

 

At the end of the day, what matters is what sounds better for you.  There is no right or wrong when it comes to preferences 

[ray-dude]

Thank you to @Miska and @Zaphod Beeblebrox for the amazing deep dive into digital signal processing...incredibly informative to see different design philosophies compared and explained by experts.  These posts are a solid reference that I'm sure me and others will be coming back to on a regular basis.  Thank you for all the time and energy to translate the complex maths you guys live and breath into language and explanations that the rest of us can learn from.

 

In the end, the proof of the upsampling algorithm is in the hearing.  If one were to hear a difference between HQP and PGGB, the last couple pages of debate and explanation are invaluable for understanding where those differences may be coming from (and similarly, if you don't, that too is interesting to understand why)

 

 

[ray-dude]

The challenge is usually related to how they buffer or stream the next file while the current one is playing.  If the files get too big (whether 16fs or DSD256), things can get glitchy.  As soon as you layer in what the player has to do to support music seeking (jumping forward or back in a track that is playing), and things get complicated again with very large files.

 

If the player is reading and playing the file from disk (playing music as it is reading the file vs, say, loading the whole file into memory then playing), then gapless is much more robust.  I have yet to hear Roon or HQP stumble on this stuff (as Rajiv said, solid)

 

 

[ray-dude]

Folks, as OP, I think we're into non-productive territory here.  There is a PGGB vs HQP thread if you want to compare and contrast the different design approaches and intent, but these have been covered multiple times here.  

 

There are differences in approach and design philosophies, and that's OK.  Both PGGB and HQP very generously offer free trials so folks can come to their own conclusions about what works best for their ears and preferences.