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

[LowOrbit]

Roon advertise upto 768k support, so probably just being greedy there Chris!

[LowOrbit]

19 minutes ago, Fourlegs said:

 

 

ZB, using the Mscaler on pass through might of course be an interesting concept because it would enable feeding the hires PGGB files to the Dave via dual BNC which might have some advantages for the Dave instead of using USB. I guess that means it is time to experiment . . . . .

Done that - sounds great🙂

[LowOrbit]

1 hour ago, MarkusBarkus said:

@LowOrbitthat was at 24bit via FishScaler to Dave? Cheers...

PGGB Files @32 bit into MSCaler pass through over dual BNC into Dave (Wave Stratos cables).

 

Maybe encoding from PGGB at 24 bit would be more elegant but I was just too lazy to move the USB B-end to the Dave from the Scaler😁

[LowOrbit]

1 hour ago, kennyb123 said:

Agree with this.  M-Scaler on pass-through robbed the music of a good chunk of the PGGB magic in my system.  

And in mine I preferred having the HMS in the path. 

 

But I haven't done the exhaustive testing and comparisons others here clearly have. For now PGGB is a curio for me rather than a definite path as it does not provide an easy route for treating Qobuz streams, which is key for me and why I am focussing on the Dave/HMS and extracting as much performance as I can from those elements. To date I haven't heard enough "improvement" from PGGB to offest the considerable inconvenience adopting this as a core technology would entail for me.

 

 

[LowOrbit]

Streaming can of course be done via HQP with Roon, but I won't go down the Roon path.

 

Audirvana integrates well enough with both local library and Qobuz for my needs.

 

I have used HQP for many years as my primary player (that changed when I moved away from my DSD diy dac. Following other experiences here, I used HQP (all HQP processing turned off) to push my sample PGGB files to the HMS. Sounded excellent, but so does Audirvana/HMS/Dave, with the benefits of streaming functionality,  doesn't require industrial amounts of data storage, maintenance of a separate library (I need regular redbook files for portable use) and so on. 

 

We all have our priorities and PGGB - at this point in time - just doesn't make a compelling case for me. 

[LowOrbit]

40 minutes ago, kennyb123 said:

The improvement from PGGB was evident to me within 15 seconds of the first track I heard.   Pretty easy to hear with my TT2 - even when I just scaled up to 8fs.  My point simply being that it shouldn’t take “exhausting testing and comparisons” to hear the benefits of PGGB with a DAVE.  

Hi Kenny

 

I never said there were no benefits or differences were not apparent, nor that I thought it did not sound good - just that for me, right now, it isn't an approach that makes enough sense to compel me to pursue PGGB. This tool emerged from gestation just a week or so after I got my MScaler, which in turn was only a couple of weeks after my Dave finally arrived. Had PGGB launched a couple fo weeks earlier I probably would have been more willing to jump headlong into this approach. For the sake of the costs involved (in my case a bump in RAM and a couple of big SSDs) it would have been more attractive and I would have probably convinced myself that I could live with Audirvana upsampling streamed content for me.

 

Knowing what I am like, I am sure I will end up buying a PGGB license before the year is out. There are other aspects of my system that can be improved meanwhile, and those (Mains, server etc) will bring enhancements that will maximise future PGGB use.

 

If I had a Taiko Extreme sat on my rack I'd say PGGB is a natural step. 

[LowOrbit]

I have been playing my handful of PGGB 32/705 tracks again this weekend having just taken delivery of an SRC.DX to feed my Mscaler/DAVE (and of course just DAVE). I found that WTFPlay works with these files without issue (Fred only claims 352/384 as the upper rate, but I guess the capabilities of the USB interface hold sway).

 

Sound is - as always with WTF - exquisite. Enough to have me re-evaluating my initial thoughts re PGGB perhaps. 

 

ALso proved more reliable than Audirvana (3.5 or Studio trial) which is not happy with the SRC.DX driver - random failures to start playback and application crashes. Shame because the sound is fine.

[LowOrbit]

To briefly add my experience with PGGB< ZB and EQ following from ASRMichael above:

 

I bought a PGGB  license last week, stuck some additional RAM into the PC I built last year for running HQPlayer (and VR). I garbleblasted some files and, having added an SRC-DX (thanks to Fourlegs for making that an easy thing to do here in the UK), I finally heard what others were reporting.

 

So I whipped up an EQ file (I have been using HQP convolution to tweak my Raal SR1a 'phones for a year or so, homebrewed using rePhase to perform the necessary bass lift and smooth out the treble), pinged it off to ZB and quick as a flash I had a PGGB EQ correction file. 

 

I spent the weekend with some familiar tunes re-worked using the PGGB magic. It's eveything I hoped it would be - all the resolution and good stuff everyone else has reported - but the big win is just how damn easy it is to relax into the music, the sense of ease and realism. All the bits of music fit together a little bit more convincingly.

 

ZB is very responsive, the software works a treat, it yields insights and brings out the good things in music and the logistical challenges of file storage seems a small price to pay overall. 

 

I have not put the MScaler back into place. Chain is diy W10 server, HQP for replay from RAMDisk or SSD. USB out (Matrix Audio Element H USB card with outboard power via SLA/shunt regulator) to SRC-DX, Wave Fidelity Stratos cables into DAVE onto Jot R/Raal SR1a. Mains cables are screened but not fancy. Mscaler does not match the sense of ease or offer up the realism of space and instrument, but it's plug and play convenience still has a place in a streaming context and it's far from disappointing. Just now the bar has been raised.

 

Zb truly knows where his towel is.

 

 

[LowOrbit]

12 minutes ago, Zaphod Beeblebrox said:

PGGB counts physical cores not logical cores and the auto workers will match the physical cores.

That's interesting - On my i7 9700k which has eight physical cores and no hyperthreading, PGGB defaults to 4 Workers (Auto).

 

 

[LowOrbit]

I've processed five hundred albums so far, applying an eq curve to optimise for my Raal SR1a's. Very happy with the results in everything I've had a chance to listen to so far. 

 

 

 

[LowOrbit]

1 hour ago, peterlim8 said:

I've the Chord Qutest and HMS, and I still want to keep the HMS.

 

My question is:

 

Can I output the PGGB files to HMS via USB, and DUAL BNCs to Qutest?

 

Thank you!

 

P/S: The Qutest is 7m away from the HMS, and PC is just next tot he HMS.

 

 

Simple answer, yes you can do that. You will lose 3db going through HMS (even in non-upsampling mode), so will need to adjust volume level in your playback software to achieve your current output levels.

[LowOrbit]

6 hours ago, peterlim8 said:

 

Thanks for the prompt reply! I also realised that you had a Matrix Audio Element H USB card, I guess it's better than the onboard USB port.

Well spotted Peter, yes I do have and use the Matrix Audio USB card. It works very well - the key (and I suspect the same applies to other such cards) is good external power. I think that, more than any other factor, determines the results. 

[LowOrbit]

I've had the pleasure of being one of those to get beta access to the new Arbitrary Precision version of PGGB.

 

I've spanked it relentlessly over the last few weeks. I've doubled the ram in my computer to 128gb and upgraded the m.2 drive configured as virtual memory in Windows to 1tb. If you have a vast collection of pop and rock, you may not need to go to such lengths to enjoy the benefits of this new iteration of PGGB, but longer tracks do need lots of space for the magic to happen.

 

128bit precision brings new depth, more harmonic richness, more tonal body, gret acoustic realism to instruments and the space they were recorded in. If PGGB relaxes complex usic and lets you dig into it, 128bit spreads it all out, teases out more goodness and lets more enjoyment into your ears.

 

On my intel  machine, now a couple of generations behind the curve, at 128bit a typical album of 50 minutes is taking maybe eight hours to process. It varies because fewer, longer tracks take more time than more, shorter tracks. There's a lot of file management going on.

 

It is also important to sy that in my case processing times have been about 40% longer per track because I  use the EQ function. When I tried some test tracks without the EQ process engaged both track processing times and memory pressure were lower and for many people who don't EQ I think 64gb of ram at 128bit will suffice.

 

It's worth noting that the addition of disk caching was not in the last version of beta I have been testing. I think that will be a useful feature for processing longer tracks. 

 

At 192bit, you'ree going to be patient, you're probably going to want to be quite selective about the music you process because it is pushing the limits of typical desktop machinery. Even without EQ processing, it takes a lot of time. There can be worthwhile incremental gains in acoustic presentation and separation of intruments way down in the mix, but 128bit is very much the sweet spot.

 

In short, I am loving the new layers of reveal, the added tonal richness and harmonic life (pianos and guitars come alive) that this new update brings. 

 

So a big thanks to Zaphod, for letting me play early and for upping the game beyond what was already a great hifi product.

[LowOrbit]

6 hours ago, kelvinwsy said:

I hv the Buena Vista CD in Wav fornat and Courtesy of this forum here I downloaded the 128bit PGGB file of Pueblo Nuevo 

In straight passthru Mode .. It did sound gloriously rich with more warmth and air. No Filter no noise shaping and fixed 0 max/min volume...

ALSO

Using polysinc gauss long and LNS15 to upsample the 44.1 khz version was also very nice ( 705.6 khz PCM output mode)

 

Then I switched bk to my GO-TO ASDM7EC v2 DSD512 mode which has a very sweet hifreq Tone! This is critical as I listen using the Abys AB1266 phi/cc - great slam but a real detail monster..

ON 1 of my favorite musical flavours (Kpop which has tremendous amount of Hifreq music content) , the AB1266 can be mercillously revealing!!

Pcm705.6 fr 44.1 does not cut it!! The hifreqs were a bit too shrill .. on Kpop 

 

ASDM7ECV2 DSD512 was the sweet spot!

 

Initial tests with the PGGB128 Pueblo Nuevo track gives a v nice rich midrange tone

 

I think the new PBBG AP at 128 bit will be v useful to fill out the Abyss 1266 slightly dry Midrange tone
 

i would love to put thus new version of PGGB AP to test really tough music genres!

 

Your experience mirrors mine (though with the equally revealing Raal SR1a's rather than the 1266's). HQPlayer sounds terrific, PGGB-AP just goes deeper, comes up richer.

 

Your CPU is probably up to the task but if you can run to another 32gb of RAM PGGB-AP will benefit in terms of speed (still not speedy at 128 bit though) and you'll run less risk of hitting out of memory on longer tracks. If you're only planning on "pop" length tracks and no EQ you should be fine based on my experiences over the last few weeks with PGGB-AP (and having gargle-blasted all my tracks previously at 64bit). 

 

Incidentally, I have found the GA version of PGG-AP seems to cope better with demanding files (long stuff), even without using the cache function. 

 

 

[LowOrbit]

I've been listening to a Gustard X26Pro/GustardU18 DDC combo with 16fs PGGB files (processed at 182 or 192 bit precision) for a couple of weeks now, lined up alongside my Chord Dave/SRC.DX with the same files.

 

Gustard have an impressive combo there. If I didn't have the Dave I'd buy the chinese pair and be very happy.

[LowOrbit]

20 hours ago, Progisus said:

So has anyone reconverted their whole library to higher than 64b precision? Although I wish mine was converted to 128b it is only new adds for me.

I've done maybe a couple hundred albums at 128 and a few dozen favourites at 192. The extra processing time for convolving eq that I use adds a 40% time overhead which  is exacerbated with longer tracks. I've converted some albums at 192 which have taken in excessive of forty hours to process. (These albums used to fail to process, but the latest PGGB versions are much improved in that regard - thanks ZB)

 

So then thoughts turn to what level of desktop computer would significantly improve the processing times as I would like to have most of my library at 128 bit a least and much more of it at 192. Maybe worth a discussion here to understand what might yield worthwhile processing time improvements... I have a fairly powerful machine at the moment, but a day per album means I'd probably be too old and deaf by the time I'd got through everything I have at 192bits. (i7 9700k, watercooled, 128gb ram, about a terabyte of swap space.)

 

 

[LowOrbit]

4 hours ago, Zaphod Beeblebrox said:

I agree, the addition of EQ is painful when it comes to processing times. Currently I use P192 for my headphones with no EQ and P128 for my speakers that need EQ as with P192 with EQ, you can't do many albums. I use an AMD EPYC 7443P liquid cooled PC with 512GB of RAM, overclocked to run at 4GHz and has 24 cores, I average about 4 albums a day without EQ at P192. That will give you some sense of how it can scale.

 

When configuring a new PC, having more RAM will be the most important factor, once all processing fits in RAM, then having more cores help as PGGB is able to parallel process and scale with cores, then faster processor helps. Once you exceed 256GB of RAM, you are likely in server territory, so you are more likely to find CPUs with more cores than higher speeds. When paging is needed, having a fast NVME drive for virtual memory is also helpful, with 1TB of spare space.

 

I am currently working on PGGB-AP that will halve the memory footprint for P192 and will run close to the current P128 speeds, so this would make P192 more practical at 128GB of RAM and one may not have to venture beyond 256GB of RAM into server territory. P128 processing will also be significantly faster but will have the same memory footprint as now.  I do not have an ETA but I hope before end of year. There will be additional cost involved as this involves collaboration with researchers (hand optimized assembly for quad and octuple precision) and licensing. My hope is to provide the faster processing as an add-on for those who wish to us it (instead of having to invest a lot more in hardware).

 

 

 

Great stuff - I shall plod on with my current hardware and await further developments. EPYC money for a single use system is probably more spendy than I want to get right now so improvements in the software would be fantastic.

 

 

[LowOrbit]

Not much help but I had no issues with the Gustard driver on Windows (build 1903) or Linux, 768/705k worked great and sounded sweeter.

 

Really excellent dac - and not just "for the money". As I've said elsewhere, if I didn't have the Dave, I'd be happy with the X26pro. (I had the U18 at the same time and felt that using that and IIS to the dac was gave a little more sq too, but only had the pair for a few weeks, so didnt swap stuff about too much.

 

A Sablon EVO USB cable that I received a couple of days before I boxed the Gustard kit up for Amazon return made brough a good jump too.

[LowOrbit]

It would be nice to get the new flac bersion in WTFPlay - still my player of choice for PGGB files.

[LowOrbit]

Golly - I've been waiting for this, checking in frequently to see if ZB had dropped the announcement here.

 

And I wake up in a HIlton on a freezing Hew Hampshire morning, jetlagged and 3000 miles from my desktop machine reserved for processing my PGGB files - and here it is.

 

It'll make the redeye flight back to the UK Friday much more bearable.

 

I don't see how I will not be upgrading, so thanks in advance ZB and the testing team.

[LowOrbit]

OK - so I got back from the US 6am Saturday, so other than dowloading the new version of PGGB, reading the updated documentation and trying out a couple of tracks I wasn't really in fit state to make an evaluation. (Though apparently I snored loudly on the couch for a good long while😁).

 

Since then ears recovered and what passes for normal brain function restored, I found time to run a few known tracks through the new PGGB on Sunday and definitely liked what I heard. Purchased new license pretty soon after that.

 

Jeez, this is so much faster, even with EQ (which was always one of the big time factors with earlier PGGB versions, and I can't live without some EQ for my Raal's). I am so pleased with this as it saves me spending a lot more cash on computing hardware.

 

Soundwise, yes, I can hear immediately that PGGB 256 is quite different to the previous AP versions. Leading notes in complex piano runs are much more clearly delineated, instrument separation is remarkable and musical flow is more compelling. Vocals are definitely clearer making some lyrics easier to discern but the depth , body and nuance to voices is one of the most compelling improvements.

 

This is an excellent upgrade and brings a fresh perspective I think on recordings I though I knew very well.

 

Thanks ZB and the testing team. 

 

 

 

[LowOrbit]

18 hours ago, Atriya said:

I'm in the process of buying a Weiss DAC501-4ch, and wonder if there might be any merit to using PGGB with it.

 

While the DAC accepts 384Khz PCM and DSD128, it is perhaps unusual in that all incoming signals - both DSD and PCM - are internally converted to 195.312Khz PCM (yes, that strange number!) before being fed to ESS chips. Apparently Daniel Weiss determined that this is some kind of "sweet spot" for the chips.

 

The filter used internally cannot be changed, but I was able to find that it is a "conventional linear-phase filter with a symmetrical ringing before and after the single full-scale sample."

That is a very specific number (possibly a result of clock divisions in play).

 

Quite a lot of pro-audio DACs are designed to operate at a rate not divisible by the standard sample rates we store and replay from. Benchmark, Merging & CraneSong all spring to mind as doing this. Noise management and filter performance are usually cited (where you can get any information at all) as the reason for these choices.

[LowOrbit]

5 hours ago, tupuhirez said:

Using Windows, no matter how much you strip it down, isn't the same as using an OS with a kernel that will give I/O priority to the calling process over everything (without crashing the system of course) vs the usual scheduling of things.  Devices in our playback chain need to operate more like appliances in the traditional sense.  Based on what I and others are hearing in my system, most recordings deserve better.  You'll be amazed at what you can bring out of regular ol' Redbook tracks.  

 

Okay, this is long enough.  Enjoy the music!

 

This aligns with much of my own experience. Using a very stripped-out W10 build I get a satisfying sound (HQplayer with no upsampling features turned on so it just plays PGGB files back gaplessly). Using WTFPlay on the same hardware the sound is fuller, richer, less tense. Just remains a shame that gapless playback (essential with multi-part PGGB files) is not implemented.

[LowOrbit]

If you like a revealing, musical presentation, the T+A Dac200 is a very fine option.

 

(Of course, feed it PGGB upsampled files and it jumps to a higher level altogether, but very good without that.)

[LowOrbit]

2 hours ago, Kalpesh said:

Yes I have different eQ profiles

Nevertheless I tried PGGB eQ with my main profile, took over 4 hours for a 9 minutes file with a 32 GB Mac Studio

Nevertheless I wanted to try linear eQ and installed 45 after adding a 43 to remastero's folder name

Now it's broken, reverting to 43, or unchecking linear eQ does not help

I've seen the same errors when trying to use the latest version. These lead to aborting the specific file(s) so I have reverted to the previous build.