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

[Zaphod Beeblebrox]

In addition to the Paging/VM space calculator - Zaphod's Guide (remastero.com) for paging/virtual memory, I have now added a storage calculator that lets you figure out the storage space you will need for storing gargle-blasted albums based on your output rate:

Storage space calculator- Zaphod's Guide (remastero.com)

[Mista Lova Lova]

10 hours ago, austinpop said:

But in DSD512, it became transcendent.

 

7 hours ago, jelt2359 said:

it's like watching OLED for the first time, and discovering that what I thought was 'black' was not actually black. The background on PGGB DSD is so jet black it is incredible- and the corresponding increase in resolving ability makes me feel like I upgraded a DAC all over again.

 

Just wanted to highlight the above observations as the words used resonate with how I'm feeling about PGGB DSD512. Some tracks (relatively simple ones) finally just feel both big (soundstage) and "right", with others I keep experiencing jaw-dropping moments because each sound within the soundstage is not only properly "grounded", but has its own space and holography, as if each sound was a separate musical event, happening independently of the others (in a good way, i.e. separation and precision, not lack of cohesion). This also applies to vocals which as a result feel much more life-like. Stunning.

[ajm]

As a DAVE user, I was thinking of trying the DSD512 suggestion until I recalled that DSD512 is available with DAVE only via the Windows ASIO driver. This is a bit of a problem for those of us who use non-windows software for our music players. I had recently finished converting my library to PCM 705/768 using PGGB. This included many DSD source files which took a very long time to process but I had been pleased with the results. Many of my sources sound much improved with the PGGB'd files. This has been most notable to me on files originally 44.1 PCM but having changed to routine use of SRC-DX, it became too inconvenient to change cable connections for higher res DSD files and I was pleased to have them all converted to PCM. Not being inclined to setup another device with windows at present, I may perhaps I will use the device that has converted the library over the last few weeks as a trial endpoint with an HQ player NAA instance. This machine, of course, was not intended as a music player and is currently sited away from the listening room with a standard power supply so is less than ideal for the purpose but I may give it a try. It is currently doing some DSD512 experimentation with PGGB and I can see that if I were to convert my library to DSD512 I would certainly require a further addition of disc drives to the NAS as the DSD512 files seem to be ~3x the size of the .wv PCM files. 

[Zaphod Beeblebrox]

10 minutes ago, ajm said:

As a DAVE user, I was thinking of trying the DSD512 suggestion until I recalled that DSD512 is available with DAVE only via the Windows ASIO driver. This is a bit of a problem for those of us who use non-windows software for our music players. I had recently finished converting my library to PCM 705/768 using PGGB. This included many DSD source files which took a very long time to process but I had been pleased with the results. Many of my sources sound much improved with the PGGB'd files. This has been most notable to me on files originally 44.1 PCM but having changed to routine use of SRC-DX, it became too inconvenient to change cable connections for higher res DSD files and I was pleased to have them all converted to PCM. Not being inclined to setup another device with windows at present, I may perhaps I will use the device that has converted the library over the last few weeks as a trial endpoint with an HQ player NAA instance. This machine, of course, was not intended as a music player and is currently sited away from the listening room with a standard power supply so is less than ideal for the purpose but I may give it a try. It is currently doing some DSD512 experimentation with PGGB and I can see that if I were to convert my library to DSD512 I would certainly require a further addition of disc drives to the NAS as the DSD512 files seem to be ~3x the size of the .wv PCM files. 

Are you sure about the 3x size? I would expect closer to 1.5x, DSD512 without compression is the same size as 16fS 32bits.

 

Would like to hear about your findings on using DSD512 and DSD256 with DAVE DSD+ and please note SRC-DX cannot go above DSD128, so you will need to use USB input.

 

Edit: Also note DAVE does not seem to play 48kHz DSD512 rates correctly either, so you will have to convert them to 44.1kHz rates.

[ajm]

Yes, but I'm using 24 bit PCM files. Limitation of SRC-DX meant that using DSD required changing cable connections for higher res files each time (which in practice meant I was unlikely to listen to DSD) so I was fairly happy converting all to PCM to avoid this. 

However, when last I tried using 32 bit files direct to DAVE USB, I did not find them preferable.

[Zaphod Beeblebrox]

8 minutes ago, ajm said:

Yes, but I'm using 24 bit PCM files. Limitation of SRC-DX meant that using DSD required changing cable connections for higher res files each time (which in practice meant I was unlikely to listen to DSD) so I was fairly happy converting all to PCM to avoid this. 

However, when last I tried using 32 bit files direct to DAVE USB, I did not find them preferable.

Wavpack also supports DSD compression, I plan to enable it on next update.

[taipan254]

33 minutes ago, ajm said:

As a DAVE user, I was thinking of trying the DSD512 suggestion until I recalled that DSD512 is available with DAVE only via the Windows ASIO driver. This is a bit of a problem for those of us who use non-windows software for our music players. I had recently finished converting my library to PCM 705/768 using PGGB. This included many DSD source files which took a very long time to process but I had been pleased with the results. Many of my sources sound much improved with the PGGB'd files. This has been most notable to me on files originally 44.1 PCM but having changed to routine use of SRC-DX, it became too inconvenient to change cable connections for higher res DSD files and I was pleased to have them all converted to PCM. Not being inclined to setup another device with windows at present, I may perhaps I will use the device that has converted the library over the last few weeks as a trial endpoint with an HQ player NAA instance. This machine, of course, was not intended as a music player and is currently sited away from the listening room with a standard power supply so is less than ideal for the purpose but I may give it a try. It is currently doing some DSD512 experimentation with PGGB and I can see that if I were to convert my library to DSD512 I would certainly require a further addition of disc drives to the NAS as the DSD512 files seem to be ~3x the size of the .wv PCM files. 

 

While I am a big fan of the SRC-DX, for the last few days I've been using a Topping HS 02 with my Mojo2, and it sounds pretty darn comparable to the SRC-DX. I've had the device for awhile, but haven't used it in a long time and haven't used it since I got the Mojo2 and SRC-DX. I was inspired to try this by @rayon who mentioned using an Intona with his Chord DAC (this mention was on Head-Fi). 

 

Adding a USB isolator and additional USB Cable could be worth your while. 

[jelt2359]

3 hours ago, ajm said:

As a DAVE user, I was thinking of trying the DSD512 suggestion until I recalled that DSD512 is available with DAVE only via the Windows ASIO driver. This is a bit of a problem for those of us who use non-windows software for our music players. I had recently finished converting my library to PCM 705/768 using PGGB. This included many DSD source files which took a very long time to process but I had been pleased with the results. Many of my sources sound much improved with the PGGB'd files. This has been most notable to me on files originally 44.1 PCM but having changed to routine use of SRC-DX, it became too inconvenient to change cable connections for higher res DSD files and I was pleased to have them all converted to PCM. Not being inclined to setup another device with windows at present, I may perhaps I will use the device that has converted the library over the last few weeks as a trial endpoint with an HQ player NAA instance. This machine, of course, was not intended as a music player and is currently sited away from the listening room with a standard power supply so is less than ideal for the purpose but I may give it a try. It is currently doing some DSD512 experimentation with PGGB and I can see that if I were to convert my library to DSD512 I would certainly require a further addition of disc drives to the NAS as the DSD512 files seem to be ~3x the size of the .wv PCM files. 

I understand. My taiko extreme’s usb card comes with its own driver that doesn’t work with usb out to the chord dave. In short, it’s suboptimal. But once I started playing pggb dsd, none of that mattered. The sq difference is so pronounced that I’ve never looked back since. 

[GryphonGuy]

I've just been trying the DSD512 in one pass option for a couple of test tracks on my Chord DAVE and I've got to say the DAVE's DSD presentation is sonically fantastic with great tonal presence BUT the soundstage is flat. That is, the sounds are strictly between my ears (pleasant and even very nice BUT no depth or 3D presentation like it usually does). So the AES input at max 192kHz 24bit is a superior presentation using Chord DAVE in my system. And this makes sense because Chord DAVE designer has not prioritised DSD on the DAVE since he is of the opinion it is inherently inferior in many ways to PCM.

 

But just to reiterate, the sonics and tonal presentation is really good at DSD512x1 on Automatic settings. Just don't expect fabulousness on a Chord DAVE.

 

GG.

 

PS I am a headphone user. hence the between-my-ears comment above.

[jelt2359]

9 hours ago, GryphonGuy said:

I've just been trying the DSD512 in one pass option for a couple of test tracks on my Chord DAVE and I've got to say the DAVE's DSD presentation is sonically fantastic with great tonal presence BUT the soundstage is flat. That is, the sounds are strictly between my ears (pleasant and even very nice BUT no depth or 3D presentation like it usually does). So the AES input at max 192kHz 24bit is a superior presentation using Chord DAVE in my system. And this makes sense because Chord DAVE designer has not prioritised DSD on the DAVE since he is of the opinion it is inherently inferior in many ways to PCM.

 

But just to reiterate, the sonics and tonal presentation is really good at DSD512x1 on Automatic settings. Just don't expect fabulousness on a Chord DAVE.

 

GG.

 

PS I am a headphone user. hence the between-my-ears comment above.

You did use dsd+ setting on Dave? And 9th order modulator on pggb dsd?

 

Thanks for the feedback. I’m also a headphone user. 
 

Here is the block diagram for Dave, btw. At DSD 512 you skip both WTA 1st stage <and> WTA 2nd stage, which in a normal Dave situation means you’re basically bypassing all of rob watts magic filtering. 
 

It hence makes absolute sense if you’re using Dave without PGGB DSD that you use PCM to maximise his filters. 

 

But PGGB DSD is a different beast…

[rayon]

Btw what does the second stage algorithm do technically? In know it further upsamples, but is doing interpolation or is it some kind of filter?

[Zaphod Beeblebrox]

5 minutes ago, rayon said:

Btw what does the second stage algorithm do technically? In know it further upsamples, but is doing interpolation or is it some kind of filter?

It is a much simpler/short filter.

[Zaphod Beeblebrox]

There are couple of reasons I implemented and provided these multiple options for both 2 stage upsampling and also modulators.

1. Obvious performance tradeoffs: Using lower multipliers for the first stage needs less memory and is faster. Lower order modulators are also faster.
2. Sound quality:
   1. 2 stage upsampling: Instead of me making the decision for you and just providing one or two options that I know would sound best, it is easy to hear quality difference as the first stage upsampling factor changes while keeping everything else the same. This is a nice way to demonstrate how reconstruction accuracy affects sound quality. It is also a way to demonstrate that better quality upsampling even beyond 16fS rates (which in some sense is equivalent to using longer and longer filters) still continues to make audible difference. 
   2. Modulators: It is often argued that reducing the quantization noise beyond a certain point in the audible range does not matter and that one needs to worry more about the out of band noise even more. Well, I keep the quantization noise in the audible range below -200dB even for the 5th order modulator, so you can easily compare whether you prefer 5th, 7th or 9th order modulator because when the order increases by 2, I drop the quantization noise in the audible range by 50dB.  It is easy for you to find out for yourself if the increasing order improves or makes it worse. Bottom line, no matter what you choose, the quantization noise in the audible range is below -200dB.

The key here is when comparing everything else is held constant. So, you can keep the modulator order constant and listen to what different first stage upsampling does. Or you can fix the upsampling algorithm and change just the modulator order and listen to what you hear. Of course, please feel free to share what you discover as that is part of the fun (i.e., how you arrive at what you like best.)

 

Edit: One of the reasons these comparisons are more relevant now is because with 1-bit DSD and a DAC with a pure DSD path, you know the signal is not touched in anyway, and that eliminates many variables.

[rayon]

Yes, it's really nice that you gave that option. Thank you! Actually my motivation behind the question was related to the fact that DSD1024 was slow to process. Did quick comparisons on one 7:25 long redbook track (psytrance) with 13900k (cooled with AIO) and 96gb 5600mhz RAM. Managed to kill the PGGB process before it finished and didn't dare to start it again :)

 

First pass (then to DSD1024)	Min	Sec	Total sec	Percent	SQ
16	15	45	945	100 %	100 %
64	18	34	1114	118 %	110 %
128	21	16	1276	135 %	115 %
256	33	51	2031	215 %	120 %
1024	>120	#N/A	#N/A	Everything	#N/A

 

• 16fs x 64
  • Baseline
  • I think I already prefer this over PGGB→HQP, but it's not a fair comparison as I'm listening to this with my virgin ears instead of doing proper A/B comparison.
• 64fs x 16
  • Bass has more definition
  • Sound becomes more dimensional
  • Cleaner
    • When coming from 16fs x 64, this has less bite
    • Compare other way around and with 64fs x 16 and go down and 16fs sounds is harsher
    • Two sides of the same coin, also may be taste dependent
• 128fs x 8
  • Dimensionality from 64fs jumps even more than 16fs x 64 -> 64fs x 16
  • Bass is further better articulated
    • It also changes it’s nature into smoother direction
    • However not a night and day (or is shadowed by changes in dimensionality)
• 256fs x 4
  • Dimensionality is further refined, but not as big jump as earlier
    • In this section returns seem to diminish first
  • Some kind of realism is increased (this is first step when this happens)
  • Bass jumped more this time and became more visceral
    • This annoys me as processing times increase so rapidly and this is very addictive
  • Any kind of harshness is only a memory

 

Playback chain:

 

NUC -> May -> Bliss -> Abyss 1266 TC

[rayon]

Sounds great!

 

Btw one more thing, I heard some popping sound in the beginning of some songs when listening to one album, which is gapless. Haven't heard that before. I was using HQP to play those. Is it related to processing and is there something I could do in processing?

[Zaphod Beeblebrox]

4 minutes ago, rayon said:

Sounds great!

 

Btw one more thing, I heard some popping sound in the beginning of some songs when listening to one album, which is gapless. Haven't heard that before. I was using HQP to play those. Is it related to processing and is there something I could do in processing?

First test is if it is repeatable. Also are you using EQ? If it is repeatable and you are not using EQ, please DM me with the two tracks (the transition where you heard the pop). If you are using EQ, I am aware of the issue and I am looking into it.

[rayon]

Sounds good. I'll try to reproduce tomorrow.

[jelt2359]

50 minutes ago, rayon said:

Sounds good. I'll try to reproduce tomorrow.

If you're worried about processing time and already have a 139k (I have a 139k and 149k with custom cooling, so I know this setup well)- at least compare 512x1 to 1024x1 one time before you decide. The choice between 512 and 1024 seems very dac dependent. But 512x1 is where things start to get pretty magical imo..

[GryphonGuy]

11 hours ago, jelt2359 said:

You did use dsd+ setting on Dave? And 9th order modulator on pggb dsd?

 

Thanks for the feedback. I’m also a headphone user. 
 

Yes DSD+ is mandatory on DSD512 so I used that and the automatic setting shows 9th order for DSD512.

 

I'll be sticking with my heavily clocked and synchronised AES/EBU path. I don't want to go down the rabbit warren of reclocking USB with appropriate cables. I think the future of digital audio is via networking protocols and optical at that since it makes such a positive effect on sound quality and staging depth. At least in my system.

 

Regards

GG

[rayon]

Or if Windows can't use two drives simultaneously, forget 660p completely and just use Samsung (as 660p top speed is slower than Samsung sustained speed).

[rayon]

15 hours ago, Zaphod Beeblebrox said:

First test is if it is repeatable. Also are you using EQ? If it is repeatable and you are not using EQ, please DM me with the two tracks (the transition where you heard the pop). If you are using EQ, I am aware of the issue and I am looking into it.

And wanted to confirm that this isn't related to PGGB. If I play files in playlist mode in HQP, I don't get pops. Just when playing albums.

[Miska]

1 hour ago, rayon said:

And wanted to confirm that this isn't related to PGGB. If I play files in playlist mode in HQP, I don't get pops. Just when playing albums.

 

In playlist mode, HQPlayer is performing de-popping algorithm since the tracks are unrelated. This also means it is not completely gapless. In album mode it is assumed that that all the content belong together and are in correct order, and thus this is not needed and the playback can be completely gapless without the de-popping algorithm.

 

So if you get pops in album mode, then you know those tracks have pop issues.

 

[Zaphod Beeblebrox]

On 4/21/2024 at 4:05 PM, ajm said:

As a DAVE user, I was thinking of trying the DSD512 suggestion until I recalled that DSD512

in DAVE and DSD+ mode, with both DSD256 and DSD512 both WTA1 and WTA2 are bypassed, that is good as it takes one more stage out (compared to 16fS PCM). But ...  DAVE does apply a filter to remove HF noise from DSD.

• For DSD256 this is non decimating, so what comes out is 256fS PCM.
• In the case of DSD512, what comes out is 512fS PCM, however, it is decimated to 256fS PCM to stay consistent with the next stage. Which takes 256fS to 2048fS. 

Given both DSD256 and DSD512 benifit from bypassing WTA1 and WTA2, it is worth giving both a try. On paper, that looks good; however, it is hard to predict if the results will be better/worse than 16fS PCM. Do try and report as the feedback so far has been very mixed.

[rayon]

4 hours ago, Miska said:

 

In playlist mode, HQPlayer is performing de-popping algorithm since the tracks are unrelated. This also means it is not completely gapless. In album mode it is assumed that that all the content belong together and are in correct order, and thus this is not needed and the playback can be completely gapless without the de-popping algorithm.

 

So if you get pops in album mode, then you know those tracks have pop issues.

 

Actually after further debugging, there is some truth in that these tracks have pop issues. It happens on some tracks, but not on others. I can reproduce the same results. It also doesn't have anything to do with rate family. I have content both in x44.1 and x48 families both with pops and without pops. To me it seems like it has something to do with silence in the beginning of the track. Those that don't pop, don't start immediately. I have processed these tracks with a setting that trims track and adds silence in the beginning of the track, but it I seems that it only applies to PCM. Also not sure if that would fix the issue anyway, this is just my educated guess. Is there some technical reason for why that silence can't be trimmed/added for DSD @Zaphod Beeblebrox?

 

And on the other hand I miscommunicated earlier that I wouldn't get these pops in playlist mode. I at some point thought that these pops wouldn't occurr in playlist mode, but it seems to happen everywhere. Also: this problem seems to occurr even if I pause the track, move in the beginning and continue. The beginning of the track is somehow corrupted or something. I get pop, one second of silent hiss (or silence, not sure), another pop and then music starts. Otherwise I can pause, start and move within track as much as I like. It's just the beginning.

 

I also found one track that is just 1:45 min long that I can use for debugging. I will try upsampling with different settings and see if I can find the root cause.

[Zaphod Beeblebrox]

8 minutes ago, rayon said:

Actually after further debugging, there is some truth in that these tracks have pop issues. It happens on some tracks, but not on others. I can reproduce the same results. It also doesn't have anything to do with rate family. I have content both in x44.1 and x48 families both with pops and without pops. To me it seems like it has something to do with silence in the beginning of the track. Those that don't pop, don't start immediately. I have processed these tracks with a setting that trims track and adds silence in the beginning of the track, but it I seems that it only applies to PCM. Also not sure if that would fix the issue anyway, this is just my educated guess. Is there some technical reason for why that silence can't be trimmed/added for DSD @Zaphod Beeblebrox?

 

And on the other hand I miscommunicated earlier that I wouldn't get these pops in playlist mode. I at some point thought that these pops wouldn't occurr in playlist mode, but it seems to happen everywhere. Also: this problem seems to occurr even if I pause the track, move in the beginning and continue. The beginning of the track is somehow corrupted or something. I get pop, one second of silent hiss (or silence, not sure), another pop and then music starts. Otherwise I can pause, start and move within track as much as I like. It's just the beginning.

I have not run into any pop issues, truly gapless albums (like Opera) transition smoothly without gaps or pops. 

 

There are too many variables here, so please DM me couple of problematic tracks (originals not converted) so I can look into it. What I would also like to know is how the same tracks behave at DSD512.

 

Regarding silence, you are correct, PGGB does not currently support adding silence to DSD. However, I am planning to add that option.