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

[atxkyle]

I’ve been able to trial a few PGGB converted files through a Hugo TT-2 and can confirm the increased realistic body of voices, instruments and ambient noted by previous posters. One step closer to having the live performance in my living room. Great stuff!

 

I mostly stream vs. purchase music, so am very interested in how the streaming PGGB options evolve. Would be very interesting to compare a PGGB streaming option vs. adding an M Scaler  vs. HQP upsampling. 

[atxkyle]

Downloaded PGGB today and emailed for a trial license (thanks @Zaphod Beeblebrox!)

 

I wasn’t sure if I’d be able to use it on my Windows PC, which is several years old and only supports 16GB of RAM, with an elder-generation quad-core i5. For 44.1khz files up sampled to 705/768 and 32 bits it worked like a charm on my hardware! 3-5 minutes to process a standard length track. I did assign the specified 128GB virtual memory on a fast SSD. 

 

I’m sure it runs much faster on a higher-power PC, but if you’re in the same boat as me and limited to 16GB RAM, don’t let that hold you back from trying it! With my system and my ears, it’s a fantastic SQ improvement.

 

[atxkyle]

Hey @Zaphod Beeblebrox - I just told my kids that I 'garble-blasted' a song for them... is 'garble-blast' the appropriate verb?? Feel like it rolls off the toungue a little easier than 'PGGB'd' 😃

[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.  

[atxkyle]

Thank you! I’m currently trying an album at 128 bit and can’t wait to listen!

 

Curious - has anyone compared the bits of the output files? What % of bits end up changing due to doing the upsampling math at higher precision?

[atxkyle]

I've enjoyed the further uptick in tranparency/realism in the first 128-bit precision upsampled album I tried out today. I started with a very well recorded and mastered album I thought sounded phenomenal using the prior PGGB (Diana Krall, When I Look in Your Eyes), and the AP upsampling squeezed out even more.

 

I'm gonna start the process of convering my albums that have 3-5 minute tracks, and am now tempted to more RAM (I'm currently at 32GB) so I can upsample longer tracks with 128-bit precision...

 

Thank you for the upgrade @Zaphod Beeblebrox, much appreciated!

[atxkyle]

+1 for the new 256 bit version. I’ve been transfixed by Coldplay’s Parachute’s this morning which sounds unequivocally the best it ever has on my system. The ultra-black background, the texture of the instruments, the euphony of the way certain chords ring… I agree with KennyB and AustinPop’s adjectives above. 

 

I also just moved from single sub to stereo subs setup this week (which also does very nice things to improve spatial cues for transparency, realism etc) so I confess I can’t 100% deconvolve the contribution of PGGB 256 vs the sub setup. But when comparing vs. the original 44.1 tracks, I think the delta is bigger than with previous PGGB versions. 

[atxkyle]

13 hours ago, WuBai said:

Hopefully I don't follow up with posting how putting Quartz crystals between the Shift and Z key 'really opens up the midrange' on these files. 

Wait ‘til you try isolation feet under your mouse. Mind blowing. :-)

[atxkyle]

Grateful for ongoing innovation and upgrades to PGGB! 

 

I did some listening tests comparing Reference-Optimal to Reference-Alternate to the prior (Infinity & Beyond) version and thought I'd pass along my findings. 

 

The resolution (separation, transparency, imagine precision, etc etc) is highest in Reference-Optimal, however, when using this Optimal setting in my system/room/ears I pick up some harshness during louder/loudest portions of tracks that reminds me a bit of the "digital" type noise I've worked hard to minimize in my system (by getting everything onto good LPS's, etc). 

 

Reference-Alternate has less of this, and Infinity and Beyond again even less harshness but with slightly less resolution. (Actually, I wasn't noticing harshness at all with the Infinity and Beyond edition). Curious what might be the cause of this, I plotted the same track upsampled the 3 different ways using the awesome new RASA tool (thanks ZB!), which is easy to use and quite educational.

 

See attached image. Based on this, I am suspecting that my system/room/ears may be sensitive to higher noise floor in the in the >85kHz range, especially during the loud portions of tracks. Since it sounds like beta testers didn't hear this, perhaps it is system dependent. 

 

Based on this exercise I'd be curious to try out a noise shaper option that uses the full 256 bit precision but with a flatter slope after 22kHz.  

[atxkyle]

7 hours ago, Zaphod Beeblebrox said:

Thanks for using the RASA and this was the exact intended use. I.e., try to look at objective metrics to better understand and compare your subjective experience, then make the right choice based on your preferences.

 

The improved resolution can be attributed to the reduced noise floor in the audible range. The harshness on some loud passages like you say could be system dependent as no one else reported it. Beyond 85kHz you can see the new Alternate Ns (yellow)is still quite low (-200dB) which is not too bad as it is closer to 32bit dither, and the previous NS from I&B edition (blue) was only slightly less at around -220dB.

 

Aso note that noise beyond 20kHz is not directly audible, and the only way they can become audible if it increases the noise floor of the analog front end and downstream equipment.

 

Unfortunately, there is no free lunch and there are trade off. The noise shaper simply 'pushes' the quantization noise around to a different region but it cannot just remove it. Else the simple answer would be why not just keep it flat from 20kHz and up to half the sample rate. If you look at the Optimal Ns (purple), you will see how the noise floor is kept low for an extended range (up to 38kz) but the noise that is pushed out of here has to go somewhere, and it ends up increasing the noise beyond 80kHz closer to -150dB which is closer to 24bit dither.

 

The analogy I like to give is that the tooth-paste tube. If the tooth-paste tube is, half full and you close the end, if you flatten out one end, the toothpaste, just bulges out the other side. So, you can 'squeeze' the paste to where it is flat at one end and the more area you try to make flat, the more it bulges on the other end. Alternately, you can decide not to shape it in anyway (this is dither) and it will be uniformly full throughout the tube and the noise stays uniformly high throughout the frequency spectrum.

 

This was the reason I had provided three choices of NS to choose from (pasting here from my original post below) in the hope that one of them will best suit your needs. The NS has to be designed within the constraints, so you will have to decide what you would rather have. Also, it is quite possible that you do not feel the same way with all tracks, and they could be processed differently.

 

 

 

Thank you for the thorough response. I enjoy learning more about this (as someone with no prior experience in this area) and your explanations are very helpful.

 

I guess one follow up question, then, is why was the "shape" of the 3 shaper options designed/chosen? Why push the toothpaste around in these 3 ways vs. other ways? You said the NS "has to be designed within the constraints" so does this mean that this shape is more a result of the math that allows for full use of the 256 bit resolution? (or was it more based on empiric SQ preferences, or for other reasons?)