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

[Kalpesh]

Hello everybody, I'm thinking of giving PGGB a try. What are the best use cases ? in which conditions do you find it to best players such as HQP, Audirvana etc? with redbook ? with hires source ? does it work better with recent digital or old digital ? music genre ?

[Kalpesh]

is there a way to losslessly compress 64 bits wav file (I did not find on Google) ? 32 bits PGGBed files lose the immediacy of the 64 bits files that might make PGGB worth the trouble, at least with DDD piano recordings of late 80's/early 90's where I found going 8FS through PGGB to be beneficial so far. 64 is twice 32 but since I can't compress it's 4 times heavier !! 20 Gb per CD is crazy ! 5 would be more acceptable !!

64 bits no dither because I explore a PGGB/HQP combo with HQP doing the last mile

[Kalpesh]

16 hours ago, peterlim8 said:

 

I don't have the answer for you. However, I am curious to know what dac supports 64 bit? Thank you! 

NOT a DAC !! the idea was to use a player, ie HQPlayer, to do the last upsampling mile, PGGB recommends 64 bits when further processing is intended (I  hear it's better than 32) . But since 8 FS 64 bits can not be compressed it's 4 times bigger than 8FS 32 bits compressed thus as big as 32 FS : not a good solution to save storage space.

 

I have a live recording of Richter and tried to combine putting the attached, adapted from your gift, it does not work

I have  tried to combine tracks of a live recording by putting the combine.json file provided by ZB, adapted of course, in the same folder as the tracks ; it does not work

 I even tried dropping all the tracks in the combine.json file, creating a list...

nothing works for me… what do I do wrong ?

@Zaphod Beeblebrox

is this a trial limitation ?
 

[Kalpesh]

2 hours ago, Zaphod Beeblebrox said:

I was taking a break for the past wo weeks. Though you should be able to combine files in trial mode, the problem in your case is you are trying to combine more than 5 tracks and so you hit the trial limit of 5 tracks. To experiment with combine, you will have to try and combine just the first 5 tracks.

yet you replied to a few emails during your break : thank you. I wanted to try combine with a live recording of Beethoven's Diabelli variations : 33... won't do then. never mind

[Kalpesh]

1 hour ago, Zaphod Beeblebrox said:

• Have you tried to compare results with 32bit dither option out of PGGB and further processing with HQP?
• What further processing are you doing on HQP? PGGB allows you to apply a EQ and also apply a fixed gain.

yes. 32bit sounds congested compared to 64f when further processed on HQP for eQ and last mile from 352 to 1.4M. 

 

 

[Kalpesh]

Actually, I think the Holo May should be SDM fed in the vast majority of cases

 

+ I have to admit I got rapidly bored to trial with a 8Gb machine : it takes forever to go 1.4 from 44 with eQ

 

However I have tested with quite a success the theory that PGGB could be a nice pre-stage from 44 to 352 64 bits f, no apodizing, before feeding HQP that does the eQ and the last mile up to 1.4M

 

But it seems to be only worth it with circa 1985 +-3 years CDs, thus presumably pre sigma delta.

 

I have found that PGGB pre stage to work ie with Marsialis's Black code of the underground, Shirley Horne's You won't forget me, Willie Deville's Miracle, and the bunch of digital recordings Arrau Richter and Serkin made in those times

 

Has anybody else found similar results, suggestions of CDs to try?

 

[Kalpesh]

47 minutes ago, austinpop said:

 

May I ask: what DAC are these findings on? My experience is nothing like yours, but then we may be on totally different DACs. I'm on a Chord DAVE.

Holo May

[Kalpesh]

9 hours ago, mcewan71 said:

I did a trial of PGGB last Spring but didn’t follow through with a licence purchase. I can’t completely remember my reasons for this, but recently I rediscovered the remastered files and really enjoyed them. I recalled that some trial and error, mainly with the ‘onboarding process’, did determine whether or not it gave me what I was hoping for. So I basically just didn’t have enough time in the trial, and now am contemplating buying a license. I have a couple of thoughts, and wondered if anyone would help me out a bit with this. I am going to sound a bit daft here so please bear with me: 

 

1. So I get how this works in theory with the taps etc. I am just wondering how this actually happens - it’s kind of ‘breaking apart’ the PCM file then restructuring it from the pieces, right? I guess I have the experience of understanding that with digital audio files, you cannot put in what was not there in the first place (e.g. creating a lossless file from a lossy one would be a simple example). So - without reference to the information on the website (I know where to find it, and I think I understand the technical basics), could someone please humour me and translate this into layman’s terms? I guess I need to completely get my head round how I can feed in a PCM file of whatever bit depth and sample rate - and get a much better sounding file the other end. 
 

2. It’s probably fair to say I am definitely along the budget end of the scale when it comes to the DACs I have. Of course, this will change over time and circumstances, but my budget DACs max out at 24/96 just now. As much as I would love to buy a Dave or even a modest step up from what I have now, it’s not within my reach currently. For my listening rooms, I use iFi Zen DACs and for headphone listening I use Audioquest Dragonflies, both red and cobalt (told you they were budget!). Point is, I enjoyed what I heard with the PGGB files using these means. As I said, I will invest in higher spec equipment over time. I just read of what seems like insane sample rates folks’ DACs handle in here, and find myself wondering if it would be pointless getting a license given my current limitations? 
 

Thanks for any thoughts on these points! 

 

 

I'm now biased by @Miska 's claims that PGGB is messing with transients, and nobody has endorsed my idea of remastering old digital recordings to 352.8 via PGGB... We're supposed to prepare the files to the max of the DAC's capabilities and with my 8 Gb of RAM computer I have not processed many files and they had to be short. Yet, I'm under the impression that eQing in PGGB brings an extra bass push that might be very seductive and overcome any shortcomings of PGGB. But the paradox of PGGB to my ears is that while it seems to extract more details, the sense of space and of group of instruments in orchestras is lost

 

Yet, I have found a handful of early digital recordings that sound best first PGGBed 352.8 then sent to my Holo 1.4 M PCM thanks to HQP (Fagen's Nightfly, Cooder's Bop till you drop, Glenn Gould's 1983 Brahms (much better PGGBed than the 24/44 remaster), Harnoncourt's Mozart Idomeneo....). Maybe that's a trick of my brain and there's no rational... It might pertain to the way I get tuned to a track, this detail here, that extra bass energy there, sometimes even the impression of a voice "an octave" (figure of speech) lower buys me in. But the vast majority are very very early, as soon as 1979, digital recordiangs

 

Any suggestion of album to PGGB with great success is welcome. But to me my Holo works much better SDM fed in the vast majority of instances : so suggestions IMO should be rather old digital, pre SDM, justifying the PCM route in Holo + low transients in the sense of @Miska : limited bandwidth, no attack that trails in the very high frequencies...

 

I think anyone who can afford Chord Dave + Terabytes of storage of PGGBed files+ 128 GB RAM ultra powerful PGGB processing PCs and would not consider a DAC implementing, at least as a route, Sigma Delta Modulation as in the vast majority of ADCs since the 90's, should support @Zaphod Beeblebrox accomplishment and support by acquiring a license.

 

You.... ? I'd rather save for another DAC + HQP

[Kalpesh]

5 hours ago, kennyb123 said:

Not good to allow oneself to be biased by other’s claims - especially if the claims are theoretical in nature.

 

Why do you need an endorsement?  Can’t you just use your own ears to decide whether this is right for you?

 

This is available for folks in your situation:  https://pggb.io.  I have on several occasions shared processed files to folks who were interested in hearing PGGB for themselves.  But what each of them had in common was a demonstrated enthusiasm to give it a fair shot.  Your tone throughout your post is quite the opposite.  You seem to trying to find ways to find fault in PGGB.

 

EQing aims to address the shortcomings of your room.  It operates based on actual measurements one takes of their room.  I don’t know how you arrived at your impression but it is very uninformed.

 

It seems as if the only comments that that stuck in your mind about PGGB were the few outliers that weren’t positive.  If you were to balance these remarks against the many positive remarks and your own observations you would be better able to identify why a few didn’t favor it.  It is true that PGGB delights for a small subset of all the available DACs out there.  On that subset, it brings only positive benefits.  Common sense would dictate that one figures out if their DAC is in that subset and if it’s not just move on.  Painting with a broad brush doesn’t make sense given the realities of what makes PGGB special.

I don't consider being able to acknowledge and be conscious of one's biases as weakness or faulty, quite the contrary.

 

Then, inside the subset of Holo May owners, I'm under the impression that I belong to a minority that feeds them SDM ; and it's sooo much better that way in the vast majority of cases IMNSHO : YMMV can we say to each other, even with same DACs

 

Last but not least, my approach to PGGB is positive though I don't concur it delights for Holo and IMO, Holo owners should not privilege PGGB at all and , yes, I recommend  @mcewan71 should stay away from a chain of process delighting because of the idiosyncrasies of Chord DACs ; however, whatever PGGB's shortcomings might be if they are, such as transient truncating, I found it to shine sometimes and it's when the extra details extraction transforms porridge into something with sufficiently more cues and clues for me to be pleased, in the attached list, it works for me with the Madonna, the Fagen, the Hildegard, Arrau's Schubert, Wozzeck. The Farmer and the PIL are very early digital PCM Denon while Arrau's Schubert is 1990, so though it is true the millions number of sinc coefficients rather benefit to early digital, there are exceptions in both directions. Piano and vocals are good targets. But the main reason I attach the list is that gain is sometimes changed and should be checked when doing comparisons. In some instances, 1/2 a dB more for the non PGGBed file and details pop up and I don't see the point to store 20 GB for a CD... I found extremely difficult to compare when eQ is done inside PGGB; there's much more bass in the Craig Armstrong eQed in PGGB : it's better to match levels in dBA and appreciate, or not, if the extra bass is worth transients truncating

 

Bottom line : PGGB is an accomplishment beyond addressing some DACs oddities or idiosyncrasies and I totally respect it, however I can not recommend @mcewan71 to start by the SW, then upgrade his DAC to suit the SW choice

pggb_album_analysis_v5.csv

[Kalpesh]

15 hours ago, semente said:

 

Why waste time and electricity and storage space when you can have music playback software do it on the fly?

And you won't have to repeat the waste of time and electricity when a new and higher performance version of the software comes out.

If you try by yourself as you should :

 

I doubt your RME would benefit from full file conditioning as Chord DACs do (and though I soon unchecked apodizing systematically, I do not concur with Chord's Rob Watts that apodizing in PGGB is a liability)

 

There are good chances you would dig what Jussi means by transients truncation if you listen to music with real drums

 

BUT

 

Maybe if you have eQ you will be seduced by eQ imprinted in PGGB, I think deep bass are deeper and more impactlull (but then there's a trade off with drums transients and I ended up fiddling with HQP filters to obtain a besting compromise)

 

And if you have early digital recordings that sound congested with a lack of resolution that feels like harshness, ie  on vocals, you can use PGGB as I did : as a remastering tool upsampling 44.1 into 352.8, no apodizing, no dither, 64 bits f output maximum transparency. You can even further upsample those 352.8 PCM into DSD 256. I found circa 1979 recorded operas to benefit, providing a more relaxed listening experience and enhanced sense of venue acoustics : Wagner/Boulez Ring, Wozzeck with Anja Silja, Mozart/Harnoncourt/Idomeneo. I also much prefer my PGGG remaster of Gould's Brahms over the official 24/44.1 remaster. Several early digital piano recordings benefitted too from the tiny bit of extra resolution thanks to reconstruction with millions of sinc coefficients

 

 

 

[Kalpesh]

49 minutes ago, kennyb123 said:

Have you actually heard this yourself?  If so what were you listening to and specifically what did you find truncated with it?  

 

A great track to try is "Dances on one Foot" from Charles Lloyd's "Sangam" album.  This features Zakir Hussain on tabla and Eric Harland on drums.  What I hear from my system when playing the 16FS version is the most realistic reproduction of the rise and fall times of drums, tabla, and percussion that I ever heard from any non-PGGB processed track (even if it's scaled with HQPlayer).  I get that the math says that some truncation is happening - but if that's what it takes to bring drums and percussion to life, I could care less.  The goal of my system isn't to excel at math.  It's to deliver sound that suspends disbelief.  PGGB helps me get closer to achieving that goal.

since you're a drummer and your system is no slough that sure is food for thought...

 

to answer your question : I first dug what I understood of Jussi's claim with the first drum hit of W Marsialis' Black codes of the underground : shyer PGGBed and played the PCM route of the Holo, could almost go unnoticed while it's a beautiful statement that catches my attention up to the reverb with native via DSD 256.

 

Seems that what you laud is the sense of rythm, that is different (or is it not ?) from how high in the bandwidth the harmonics of a hit, a strike, go, that is how fast and transients they are : you might both be right with different focus which makes your wording "what were you listening to and specifically what did you find truncated with it" perfectly sound

 

I don't own** Sangam (beautiful recording, the tabla at the beginning is a treat then Lloyd, on that one, bores* me a little) so can not PGGB it but what I can say after streaming it from Qobuz via HQP is that it sounds much better via Holo's SDM route than via the PCM : thus, from my trials there's no chance PGGB would change the verdict with my DAC. Better refers to realism and physicality of the placement in the soundstage

 

*edit : his first solo ; the coltranesque one after the drum solo is beautiful !

** edit : that's also quite something to have in mind ; remember @mcewan71 is on a budget and so would I consider myself when it comes to purchase albums PGGB them with powerful machines and store TB vs streaming Qobuz via HQP

[Kalpesh]

3 minutes ago, kennyb123 said:

I appreciate your response.  
 

When thinking of transients, the acronym ADSR can be really helpful.  Attack, Decay, Sustain, Release. What I was looking for was looking for was a example that pointed specifically to harm in one or more of those areas.  
 

It’s late here now so I want to keep this short.  Tomorrow I’ll provide an example of specifically where I’ve heard harm done to these areas by a component or cable.  

 

I have no doubt that HQPlayer is working wonders with your DAC.  It’s really nice to have options as each of us have different listening preferences and different sensitivities.

 

I looked back at some of your posts to try get a better sense of where you were coming from.  This reply to you explains a lot.  Do your ears tell you that PGGB “just destroys time domain performance without benefits”?  

you're quoting @Miska, not me ; but there again it might be a matter of semantic rather of maths : rythm/time domain ; i can see how you possibly relate them semantically

 

I think I have already described how I heard time domain destroyed by PGGB : when I lose the soundstage, space, perspective, all things phase/time domain related afaik : it's clear with the Sengam exemple with my DAC but with a few recordings it's the other way round : PCM shines and within that subset some are even better, including space cues, thanks to better micro information extraction, I guess, overcoming shortcomings, PGGBed

[Kalpesh]

23 hours ago, semente said:

 

What do you mean by "remaster"? Are you EQ'ing as well as upscaling? Also I have the impression that upscaling requires the use of dither...

And in what way is it different from on-the-fly upscaling?

I take the term "remaster" from @austinpop featured article where he writes : "Finally, I used PGGB, a remastering tool, to upsample my music files offline, so the DAVE only ever sees 24-bit, 16FS (705.6/768kHz) music streams. "

 

I used PGGB to upsample to 352.8/64 bits float with benefit a few 44.1 music files but my HOLO MAY sees 20 bits 32FS music streams, the final upsampling being made by HQP where dithering occurs.

 

I'm in a delicate situation to answer your last question : I'm neither a PGGB champion nor a PGGB basher, and you should ask champions and consult PGGB's FAQ. Right or wrong I just tried to find a use for the thing : IMO there's no point in PGGBing all files with my DAC and I'll stick to on-the-fly upscaling of native files with HQP ; however, I found PGGB, thanks to hundred of millions taps or sinc coefficients that require extensive off line calculations, to better reconstruct a few early digital recordings that I find harsher, probably due to a lack of resolution with mere on the fly upsampling : a welcome tiny bit of extra focus/resolution that is beneficial in borderline situations such as forte vocals in 1979 digitally recorded operas in exemple, and that overcomes whatever shortcomings (or, more often, absence of clear benefit justifying the trouble) might exist

 

 

 

 

 

 

 

 

[Kalpesh]

5 minutes ago, Miska said:

 

This is your mistake. It is not "mere on the fly". There are no trade-offs made in any HQPlayer algorithms for the reason that they run on the fly. My approach has always been that "you need to buy fast enough computer". If you cannot run some algorithm, you need to go shopping for a faster computer. Everything is 100% quality driven.

 

Of course, if you want offline processing, there's HQPlayer 4 Pro for that purpose.

 

and which algorithm would you recommend testing with  borderline situations such as forte vocals in 1979 digitally recorded operas ? PCM route ? SDM route ? 

[Kalpesh]

7 minutes ago, Miska said:

 

Digital recording from -79? Or digitized analog recording?

 

poly-sinc-gauss-long is pretty good all-rounder for classical music. DSD256 with ASDM7ECv2 since you are on a Holo DAC.

 

With Holo DACs, remember to set PCM gain compensation to -6 dB when comparing PCM vs DSD. Since the DAC has 6 dB level difference between the two.

 

thank you. I'll answer on the HQP thread

[Kalpesh]

raw data : of the 8 non classical music albums I have PGGBed with success, that is offering the most pleasant experience PGGBed 352.8/64 b f/Maximum transparency/no apodizing, 4 are listed in https://www.digido.com/honor-roll/ : I still believe that with ancient ADC/Digital recorders it brings upsampling/remastering benefits that overwhelm whatever limitations Rob Watts or Jussi Laakso have pointed.

If someone asks I will specify said albums... In classical music Philips DDD 80's recordings by Arrau and Richter account for almost half of the successfully PGGBed albums. 1979 1982 brackets most of the successfully PGGBed albums. Interestingly enough, PGGB is a liability with a 1984 DDD Arrau recorded in NY while the rest is recorded in Europe

 

My trial period ends this week-end : if anyone has a suggestion, now is the time !

 

I want to thank the author and everybody involved for allowing me to better enjoy a handful of albums and understand the enthusiasm when the idiosyncrasies of the DAC allows PGGB to make the positive difference with all albums

[Kalpesh]

I'm trialing PGGB at 256 bit precision

got my interest

though my RAM limits the length of tracks I can process and thus I'm aware I can't get PGGB at its best

so far, vs HQP 256 7ECV2, the most catchy feature of PGGB was the electric bass line in a jazz record : more there, richer, more articulated; as of piano, drums, placement cues, would not bet which is more accurate

Same with the one classical track I tested : would not bet which is more accurate regarding bells and choir ; however I can write that the most catchy feature with PGGB was at 16 FS : the harp was more real at the center of the church, than at 8, 32 FS (all 20 bits) or HQP 256 7ECV2. That might relate to low frequencies better reconstruction and thus resonances?

 

Chance or could it be that the sweet spot with Holo May is 16 FS not at its 32 FS max capability ? But I also wonder if my interest is triggered by a bigger difference vs HQP 256 7ECV2 characteristics, ie bells are snappier and I perceive longer decay at 32 FS or with HQP ; yet overall I prefer 16 FS

 

I'm upsampling all automatic. What would be your suggestions of xFS and other settings with the May? I mean for better objective results. I admit that in the exemple above I could reverse my hierarchy based on knowledge of what is supposed to be better

 

[Kalpesh]

8 minutes ago, Zaphod Beeblebrox said:

 

The May DAC accepts up to 32fS, objectively 32fS processing at 20 bits will result in about 40dB lesser noise in the audible range (-600dB) vs about -560dB of 16fS processing. But you have to weigh this against the DAC's settling time which may increase nonlinearity at 32fS and negate the benefits. Something to try at 32fS is try lesser bit depths at 32fS, you can go as low as 16bits and even at 16bits, 32fS processing is objectively better (-580dB noise floor) than 16fS 20bits (-560dB)

 

You are not the first to show a preference for 16fS with May DAC, so if 16fS works best for you then it is good to stay with it as a bonus it takes less time to process, needs less memory and occupies less space. But it is worth trying 32fS at lower bit-depths too (i.e up to 16bits at 32fS).  

 

Here is -6dB, 44.1kHz 16bit band limited white noise upsampled to 16fS 20 bits, here you see the noise floor (the lowest point in the graph) reaches about -560dB

 

Here is -6dB, 44.1kHz 16bit band limited white noise upsampled to 32fS 20 bits, here you see the noise floor (the lowest point int he graph) reaches 40dB lower (-600dB)

 

And finally, here is -6dB, 44.1kHz 16bit band limited white noise upsampled to 32fS 16 bits, here you see the noise floor (the lowest point int he graph) reaches 20dB lower (-580dB) than 16fS even at 16bits. Which is why I suggest trying as low as 16bits at 32fS.

thank you very much for this comprehensive answer. I trialed earlier version and I would recommend anyone to trial the reference version, much improved I think, so far. Worth it vs HQP?  a 256 Licence + the need to acquire a giga giga RAM computer might be too much for me but kudos : it sounds really good 

 

 

 

[Kalpesh]

sounds really good... yet an upgrade in Mac world would have to be a Studio with the $500 extra RAM... Or maybe not : the swap capabilities seems in line with what you write about Windows : slow with heavy paging and allowing the same extra length. Do other users confirm ?

 

For the time being I have settled on 16 FS and 256 precision

 

My player is HQP and despite the red warning, actually it's better and no big deal to 0 the volume

 

remains the eQ. I do eQ/DRC and I prefer to keep it separate for the time being. Should I, objectively, noise shape in HQP ? NS9 would be my choice.  Sonically, subjectively, I can't make up my mind, it's snappier with NS9 engaged and it nulls the difference I perceive between 64 and 20 bits output with no NS in HQP. On the one track I obsessionally tested, there are very catchy tones and resonances in bells and harp with 64 output + eQ in HQP but no NS. But maybe at the expense of a less defined choir compared to the NS9 engaged option in HQP.

 

Should I, objectively, prefer to engage dither/Noise shaping in HQP because I eQ ? If yes, am I right to assume that 64 output (OMG the space !) is not necessary ?

[Kalpesh]

3 hours ago, Zaphod Beeblebrox said:

  

Did not realize you are running a Mac, upgrading Macs are hard and can be expensive. Unfortunately, unlike Windows or Linux I do not think you can change swap size on Mac, if you know how to then I would like to know too.

 

Because PGGB takes care of inter-sample-overs, yes, it is safe to leave the volume fixed (0 min and max) unless you are doing further processing.

 

The choice of Noise shaper and/or dither depends on the output sample rate. If you are playing and processing 16fS signal, I would suggest LSN15 it is objectively better, and I think it is recommended for 8fS or higher rates. How are you playing 64bit signals? It will still need to be converted to the bit-depth accepted by your DAC and if you have not changed your settings, it would explain the 'null difference'

 

Objectively, yes, it is better to engage dither/noise shaping anytime you are processing. The reason is that when any processing is done at a higher bit-depth, typically 64 bits, the result after processing needs to be reduced in bit-depth to fit your DAC. This re-quantization adds noise and introduces distortion if no dither or noise shaping is done.

 

PGGB 256 uses noise shaping even for 64-bit output when the precision is set to 128 or 256 bits. Objectively, converting 256 bits to 64 bits keeps more of the information in the audible range than converting to 20 bits and 64 bits, which may be beneficial for further processing. However, any subjective and objective difference after the EQ is still contingent on the quality of the noise shaper used after EQ. So, you will have to judge for yourself if the size difference is worth any difference you hear. At the very least, I would still suggest 32 bits instead of 20 bits from PGGB, and you can request HQP to noise shape to 20 bits.

 

Here is -6dB, 44.1kHz 16bit band limited white noise upsampled to 16fS 64bits, here you see the noise floor (the lowest point in the graph) reaches about -900dB

 

 

Here is -6dB, 44.1kHz 16bit band limited white noise upsampled to 16fS 20bits, here you see the noise floor (the lowest point in the graph) reaches about -560dB, about (340dB more!)

 

Thank you very much.

I don't have any trick regarding swap and Mac ; I can just report having seen a 16 GB swap file for a mere 8 GB RAM. 

So far it allowed me to get up to 6 minutes plus files at 16FS/20 bits and my Mac is struggling for hours processing a 8'57 track. I hope I will get a bunch of tracks for fair evaluation but with my M1 I'd be dead before I finish processing all my music.

 

Should I try the new release or am I at risk it limits the length I can process ?

 

I took a deep breath and went back fiddling with REW to export a stereo wav and now wonder : where will the negative gain to compensate for my low end boost take place ?

[Kalpesh]

18 minutes ago, Zaphod Beeblebrox said:

If anything the new version may help you as it is slightly faster and needs less memory during processing.

 

PGGB takes care of all that and it tries to maximize the dynamic range when EQ is used. It may make A/B comparison a bit hard but it will make sure you do not clip so no further gain adjustment is needed.

 

GREAT

THANK you

great. I have also issued the filters with a known soundproof negative gain applied in REW which should help A/B. I'll try both options once the 8'57 track will be processed at last (processing block 233 of 256 stage 2 : seems 8.57 can be processed 16FS/20 bits by a 8GB RAM M1 albeit it's excruciatingly slow ). The positive side is that I could probably go for 32 GB machine when I replace and withstand the wait once in a while for long tracks

[Kalpesh]

46 minutes ago, Kalpesh said:

GREAT

THANK you

great. I have also issued the filters with a known soundproof negative gain applied in REW which should help A/B. I'll try both options once the 8'57 track will be processed at last (processing block 233 of 256 stage 2 : seems 8.57 can be processed 16FS/20 bits by a 8GB RAM M1 albeit it's excruciatingly slow ). The positive side is that I could probably go for 32 GB machine when I replace and withstand the wait once in a while for long tracks

took 4 hours 29 minutes and 32 seconds but the 8GB RAM M1 did it : processed a 9 minutes track in 16 FS /20 BITS  at 256 precision   Of course it's no recommendation but it's 3 times theoritical max length and if it can be derived that once in a while a 32 GB RAM Mac can cope with a 36 minutes long track, that's positive since a 32 GB machine could suffice 

 

[Kalpesh]

13 hours ago, pavi said:

 

i use the two steps simply because i prefer dsd to pcm on both the may kte & the dac 200. i also prefer — by far — the hqplayer dsd modulators to pcm.

 

i don't use pggb alone because 1. pcm not dsd; 2. i listen to a lot of streaming music from qobuz.

 

am only doing 8fs because that's the input limit for my other dac, the wavedream sig/bal.

 

any advice you might have will be gratefully received. truly superb work on pggb, zb.

 

btw — why did you suggest 64bits earlier, and not 256?

Seems we are both trialing

I have been a SDM May route proponent when PCM Sinc L LNS15 was the trendy flavor. Then vice versa. But now I use HQP with my May per the consensus : SDM 7ECV2 @256 + PG Short for pop, xla for classical/Acoustic jazz etc at 1FS, hires lp above for Any.

 

My advice would be you do as romans do in Roma, I mean I now get the most from PGGB by :

 

Preparing 16 FS/20 bits files for my May, with the eQ/DRC/Convolution processed in PGGB

setting all to none in HQP with Volume 0 and going the PCM route : pure PGGB

 

if space is an issue open the hidden menu and click to get .wv

 

I'm just getting my first tracks so processed and I'm impressed. A/B is really difficult but maybe not relevant : I'm 100% confident I never heard familiar tracks the way they sound processed and played as above. Actually, I might have to reconsider my convolution to lower the bass level... Will it be an overwhelming persistent as positive feeling ? More processing and more listening sessions ahead...

 

 

[Kalpesh]

On 3/29/2023 at 5:38 PM, hanshopf said:

I just now made a listening test myself, m-scaler vs 128bit vs 256bit. Unfortunately I was not able to come to a reliable result. More often than not m-scaler sounded best to me. But that may be due to confirmation bias or just because I am used to it. But then I felt that 128bit sounded rounder and fuller than 256bit. But this should be impossible and therefore I can only conclude that I must further test repeatedly over several weeks.

 

Anyway: a Rob-Watts-like time window of 1.5 to 25s makes more sense to me because then at least the majority of the windows will not contain an edit. The base assumption of long unedited music passages does unfortunately not correspond with reality. You will basically have a hard time finding unedited music tracks at all.

If you then argue that listening test have shown otherwise, then I hope the test listeners are better than me, which very well maybe the case (even though I am a trained listener working professionally in the field with still good hearing). I would have hoped that there would be a less vague explanation for using a time window in length of a whole music track. My guess is that a compromise might offer a better solution. A time window of let's say 1 minute length would still give you 32M taps, but would as well allow to capture most probably the majority of unedited passaged of music inside a music track. 

I would really like to trust the ears of your test listeners, but that's not easy when I feel that I cannot even trust my own ears. I wish it would be easier to decide mathematically what is more correct. As it seems to me right now, Rob Watts is still in the game, even though I hoped a software solution could be far ahead.

 

In my understanding of PGGB, and the conceptual beauty of it, at any given moment in time, it looks both past and future and adjusts t time reconstruction taking into account past and future. Both metaphysically mind blowing and trivial, the note emitted by the piano being the result of the hammer strike and resonating and mingling with the next, etc. So, if that past and future do not exist for as long as the track, the time window looked at by PGGB, the reconstruction is not as enriched/accurate but there is no strong reason for PGGB to alter the reconstruction at t time by adjusting it with uncorrelated information from a past and future of the track that would not present continuity.

@Zaphod Beeblebrox am I correct ? 

[Kalpesh]

@Zaphod Beeblebrox I found out that 0 volume in HQP by crancking right is not the same as 0 min 0 max as you recommend. Main reason seems to be that in the former it leaves the volume control I forgot I set ages ago in Preferences and that lowers PCM 6 dB in order to match SDM output.

Issue is that I lost the magic and the fascination I was enjoying with PGGB. Sounds much thinner with correct 0 min 0 max. Might be the stress resulting of the assault of 6 unexpected dB, further fueled by a zeroing of my preamp (losing its settings) while I adjusted gain, might be because I use my preamp in a less favorable zone, with a less optimised overall gain control...

 

Or could it be that I got hooked by quantization noise introduced back by HQP because of the 6 dB fixed volume control in Preferences?

 

I'm pretty sure you'll confirm I absolutely must get rid of that interference but it sounded so good... Too bad if I developed a taste for distorsion or quantisation noise creating some kind of euphony. The sound was so full and rich and sweet with the volume crancked 0 to the right thus leaving the -6 in Preferences !