DSD Frustrations With Manufacturers

[The Computer Audiophile]

Hi Guys - Since I've been using HQPlayer more and more, I've run into an issue that is someone similar to several years ago with high resolution PCM. Back in the day some manufacturers didn't support 176.4 kHz, but that was soon fixed.

 

Today, the issue is related to DSD.

 

Most manufacturers view DSD as the following, just taking 44.1 and multiplying it by 64, 128, 256, or 512:

 

DSD / DSD64 = 2.8224 MHz

DSD128 = 5.6448 MHz

DSD256 = 11.2896 MHz

DSD512 = 22.5792 MHz

 

 

However, DSD isn't locked into multiples of 44.1. Proper DSD includes the following, multiplying 64, 128, 256, or 512 by 48.

 

DSD / DSD64 = 3.072 MHz

DSD128 = 6.144 MHz

DSD256 = 12.288 MHz

DSD512 = 24.576 MHz

 

 

 

My issue is that 99.999999% of manufacturers only say their DACs support DSD, DSD128, DSD256, or DSD512. They don't say anything about which sample rate family is supported or if both are supported. Not that big of deal you say? Except, when loud bursts of noise blast through one's speakers when attempting to figure out what is actually supported. 

 

Can manufacturers just specify in their user manuals which rates of DSD are supported? As it is now, it's equivalent of saying that high resolution PCM is supported at 2x, 4x, and 8x, without any mention of 96 kHz or 88.2 kHz etc...

 

OK, rant over for today. 

[Solstice380]

14 minutes ago, The Computer Audiophile said:

Can manufacturers just specify in their user manuals which rates of DSD are supported?

 

Or just make their #$%@%^ DACs play the 48K material!!!  Love my dCS Debussy but have to convert everything to 44.1K DSD128.  My Lampi does both. 

[Miska]

Funny thing is that they have clocks already for the 48k-family PCM playback. They just fail to switch to that clock with DSD input.

 

Even worse are DACs that have lost this 48k-family capability due to firmware update (iFi, when they introduced MQA support).

 

[Miska]

Good side is that most of the poly-sinc group of filters in HQPlayer can convert from 48k-base to 44.1k-base output. But it still limits available filter choices, because it cannot be done with all filters.

 

[jabbr]

1 hour ago, The Computer Audiophile said:

Most manufacturers view DSD as the following, just taking 44.1 and multiplying it by 64, 128, 256, or 512:

 

DSD / DSD64 = 2.8224 MHz

DSD128 = 5.6448 MHz

DSD256 = 11.2896 MHz

DSD512 = 22.5792 MHz

 

 

However, DSD isn't locked into multiples of 44.1. Proper DSD includes the following, multiplying 64, 128, 256, or 512 by 48.

 

DSD / DSD64 = 3.072 MHz

DSD128 = 6.144 MHz

DSD256 = 12.228 MHz

DSD512 = 24.576 MHz

 

 

True, though its not at all clear that there is any advantage to staying in the same rate family when converting to DSD so I routinely convert to 256 or 512 x 44.1

 

let's see: I have "Contrane" in three versions: 16/44.1, 24/96 and DSD64, all being played at: xtr-mp/7EC/DSD256

 

Anyone care to venture which is which, and there is an extra ;), that's that?

15895 hqplayer  10 -10 9653300 624144 185400 S 256.0  1.9   3:26.79 hqplayerd                                                                   
 1743 root      20   0 11.198g 3.351g  70868 S  37.4 10.8 368:03.60 RoonAppliance                                                               
 9346 root      20   0 2088920  42148  25292 S   0.7  0.1   1:57.37 containerd                                                                   
   11 root      20   0       0      0      0 I   0.3  0.0   0:22.41 rcu_sched      
Thu Apr 16 14:36:09 2020       
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 440.82       Driver Version: 440.82       CUDA Version: 10.2     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce RTX 208...  Off  | 00000000:65:00.0 Off |                  N/A |
| 40%   47C    P2    72W / 260W |    560MiB / 10985MiB |     11%      Default |
+-------------------------------+----------------------+----------------------+

now

1484 hqplayer  10 -10 9966.2m 915696 181800 D 282.1  2.8 215:25.55 hqplayerd                                                                    
 1743 root      20   0 11.244g 3.742g 447036 S  40.7 12.0 359:55.14 RoonAppliance            
 3326 root      20   0       0      0      0 S   9.6  0.0   3:03.31 cifsd

+-----------------------------------------------------------------------------+
| NVIDIA-SMI 440.82       Driver Version: 440.82       CUDA Version: 10.2     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce RTX 208...  Off  | 00000000:65:00.0 Off |                  N/A |
| 41%   51C    P2    83W / 260W |    565MiB / 10985MiB |     37%      Default |
+-------------------------------+----------------------+----------------------+

and

1484 hqplayer  10 -10 9945092 1.169g 182564 S 264.4  3.8 245:51.54 hqplayerd                                                                     
 1743 root      20   0 11.198g 3.707g 447992 S  38.0 11.9 365:04.68 RoonAppliance                                                               
 9346 root      20   0 2088920  42512  25292 S   0.7  0.1   1:55.55 containerd                                                                   
  823 systemd+  20   0   80172   3856   3520 S   0.3  0.0   0:06.76 systemd-network                                                             
 3326 root      20   0       0      0      0 S   0.3  0.0   3:08.42 cifsd 

Thu Apr 16 14:30:21 2020       
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 440.82       Driver Version: 440.82       CUDA Version: 10.2     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce RTX 208...  Off  | 00000000:65:00.0 Off |                  N/A |
| 41%   51C    P2    82W / 260W |    560MiB / 10985MiB |     35%      Default |
+-------------------------------+----------------------+----------------------+
                                                                               

and finally 

1484 hqplayer  10 -10 9945092 1.106g 182564 S 255.3  3.6 241:36.89 hqplayerd    
 1743 root      20   0 11.291g 3.800g 447924 S  35.8 12.2 364:24.03 RoonAppliance 
15536 jon       20   0   51476   4308   3456 R   0.7  0.0   0:00.08 top                                                                        
 3326 root      20   0       0      0      0 S   0.3  0.0   3:08.17 cifsd             
 4462 root      20   0       0      0      0 I   0.3  0.0   0:00.58 kworker
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 440.82       Driver Version: 440.82       CUDA Version: 10.2     |
|-------------------------------+----------------------+----------------------+
| GPU  Name        Persistence-M| Bus-Id        Disp.A | Volatile Uncorr. ECC |
| Fan  Temp  Perf  Pwr:Usage/Cap|         Memory-Usage | GPU-Util  Compute M. |
|===============================+======================+======================|
|   0  GeForce RTX 208...  Off  | 00000000:65:00.0 Off |                  N/A |
| 41%   49C    P2    70W / 260W |    568MiB / 10985MiB |      5%      Default |
+-------------------------------+----------------------+----------------------+

 

[One and a half]

Please pardon my ignorance,  which DSD source material is based on 48kHz multiples?

 

Don't seem to be offered 3.072MHz music from Blue Coast, Analog Productions, even the pirates... ☠

[Miska]

5 minutes ago, One and a half said:

Please pardon my ignorance,  which DSD source material is based on 48kHz multiples?

 

Don't seem to be offered 3.072MHz music from Blue Coast, Analog Productions, even the pirates... ☠

 

You can easily create such. I can also record such with my two RME ADI-2 Pro's. Up to 12.288 MHz.

 

[The Computer Audiophile]

9 minutes ago, One and a half said:

Please pardon my ignorance,  which DSD source material is based on 48kHz multiples?

 

Don't seem to be offered 3.072MHz music from Blue Coast, Analog Productions, even the pirates... ☠

This has much more to do with apps like HQPlayer that can resample the audio to higher rates and keep multiples of the original rate. 

[One and a half]

@Miska & @The Computer Audiophile, thanks for the replies, makes some sense now. Upsampling to DSD with HQPlayer didn't work for me, after giving it a run for many weeks. Switched to Native DSD and PCM rates one day, and that process has stayed, with occasional SRC by Roon, now also set to Native rates.

[jabbr]

Regarding above it seems that 24_96 source material when converted to DSD either 44.1x256 or 48x256, in both cases CUDA goes to 35% range, otherwise for 16/44 or DSD64 source, the usage <10%

[The Computer Audiophile]

19 hours ago, Solstice380 said:

 

Or just make their #$%@%^ DACs play the 48K material!!!  Love my dCS Debussy but have to convert everything to 44.1K DSD128.  My Lampi does both. 

Just tested the dCS Rossini and it accepts the following with DoP only.

 

DSD64 = 2.8224 MHz and 3.072 MHz

DSD128 = 5.6448 MHz and 6.144 MHz

[Solstice380]

1 hour ago, The Computer Audiophile said:

Just tested the dCS Rossini and it accepts the following with DoP only.

 

DSD64 = 2.8224 MHz and 3.072 MHz

DSD128 = 5.6448 MHz and 6.144 MHz

 

When I got the unit I had to update firmware that opened up 384 KHz PCM but not 48K x 128DSD.  ☹️  Nothing new since early 2018 for the Debussy, and I don't expect it to come.

[The Computer Audiophile]

Trying to figure out a better way to determine which sample rates are supported, other than pressing play and crossing my fingers. I just don't think there is a good way. 

 

Looking at the Bryston BDA 3.14 I see the following. DSD_U32_BE I believe means it supports native DSD, but there is no indication of which level of DSD (2x, 4x, 8x, etc...) or which sample rate families. Fortunately at least the user manual for the BDA 3.14 says quad DSD. It's ambiguous with respect to which sample rates can do DoP versus native DSD, although one could deduce this based on PCM rates supported etc... but Joe Sixpack isn't going to have any clue. 

 

Frustrating. Guess I just have to press play with crossed fingers. 

 

Bryston BDA3 Audio 2.0 at usb-ci_hdrc.1-1.1, high speed : USB Audio

Playback:
  Status: Stop
  Interface 1
    Altset 1
    Format: S32_LE
    Channels: 2
    Endpoint: 1 OUT (ASYNC)
    Rates: 44100, 48000, 88200, 96000, 176400, 192000, 352800, 384000
    Data packet interval: 125 us
    Bits: 24
  Interface 1
    Altset 2
    Format: S16_LE
    Channels: 2
    Endpoint: 1 OUT (ASYNC)
    Rates: 44100, 48000, 88200, 96000, 176400, 192000, 352800, 384000
    Data packet interval: 125 us
    Bits: 16
  Interface 1
    Altset 3
    Format: SPECIAL DSD_U32_BE
    Channels: 2
    Endpoint: 1 OUT (ASYNC)
    Rates: 44100, 48000, 88200, 96000, 176400, 192000, 352800, 384000
    Data packet interval: 125 us
    Bits: 32

[ericuco]

A few months ago, I wanted to get a DSD only DAC, no PCM at all. I use HQP to convert everything the DSD256 so no need for PCM.

 

I ended up with a Holo Cyan (DSD version) only to learn that it does not support 48k-based DSD once I got it installed into my system. A real bummer. HQP can still make the necessary adjustments but it adds some additional processing to my server.

 

Currently, I trying to resolve an issue of a "pop" in one channel (left) when music starts up and the DAC is in DSD256 mode. If the DAC is in PCM768K mode at start up, no "pop". However, once it switches to DSD256 mode, it remains in that mode so the "pop" is heard. Trying to resolve this with Tim at Kitsune for months now. It may be related to using Linux as the OS in opticalModule (same problem exists when directly connected to my Linux music server).

[jabbr]

3 hours ago, ericuco said:

A real bummer. HQP can still make the necessary adjustments but it adds some additional processing to my server.

 

Does it? This "top"  24/192 -> 44.1x256 (note the 261% CPU usage)

1959 hqplayer  10 -10 10.424g 1.829g 189832 S 261.6  5.9   1321:15 hqplayerd                                                                     
 9709 jon       20   0   51476   4172   3324 R   0.7  0.0   0:00.15 top                                                                         

whereas 24/192 -> 48x256 (note the 277% CPU usage)

1959 hqplayer  10 -10 10.291g 1.668g 189832 S 277.4  5.4   1323:46 hqplayerd                   
 1652 root      20   0 2163164  44716  25300 S   0.3  0.1   6:49.25 containerd                                                                                 
 

It looks like 48x256 uses slightly more CPU perhaps because its a slightly higher rate...

 

Is there any need/benefit to 48x256 (or any of the 48 rate family x DSD)?

[20hertz]

Chris, are you convinced that upsampling is the way to go now? Do you notice times when it is not better?

Also do you try to feed a DAC its highest native input or just upsample at will?

Thanx

Mike

[The Computer Audiophile]

2 minutes ago, 20hertz said:

Chris, are you convinced that upsampling is the way to go now? Do you notice times when it is not better?

Also do you try to feed a DAC its highest native input or just upsample at will?

Thanx

Mike

Much left to research and experiment with, no 100% conclusions. What I've heard has been incredible lately. 

[20hertz]

Please keep us updated with what and how you are hearing! Much appreciated.

Thanx

[The Computer Audiophile]

Just now, 20hertz said:

Please keep us updated with what and how you are hearing! Much appreciated.

Thanx

Absolutely. This is a really fun part of the hobby. 

[ericuco]

1 hour ago, jabbr said:

Does it? This "top"  24/192 -> 44.1x256 (note the 261% CPU usage)

1959 hqplayer  10 -10 10.424g 1.829g 189832 S 261.6  5.9   1321:15 hqplayerd                                                                     
 9709 jon       20   0   51476   4172   3324 R   0.7  0.0   0:00.15 top                                                                         

whereas 24/192 -> 48x256 (note the 277% CPU usage)

1959 hqplayer  10 -10 10.291g 1.668g 189832 S 277.4  5.4   1323:46 hqplayerd                   
 1652 root      20   0 2163164  44716  25300 S   0.3  0.1   6:49.25 containerd                                                                                 
 

It looks like 48x256 uses slightly more CPU perhaps because its a slightly higher rate...

 

Is there any need/benefit to 48x256 (or any of the 48 rate family x DSD)?

You may be correct. Just assumed that it had to do more work but perhaps not. Be interested in getting @Miska thoughts on the pros & cons. 

[Miska]

2 hours ago, jabbr said:

Is there any need/benefit to 48x256 (or any of the 48 rate family x DSD)?

 

Yes, if you want any filters that can do only integer or power-of-two multiples...

 

With single stage poly-sinc filters in many cases it is is also lighter CPU load to work within rate family. For example try poly-sinc-xtr-lp from 48k to 48k x512 or alternatively from 48k to 44.1k x512. Latter works at least on RTX2080Ti, but not on i9-9900KS while the former I think does.

 

[The Computer Audiophile]

Just testing out the Ayre QB-9 Twenty and found the DAC supports DSD256 at 12.288 MHz (multiples of 48 kHz). However, the front panel display goes blank with this sample rate. Playing DSD256 at 11.2896 MHz, the AB-9 Twenty front panel correctly says 256.

 

@Ryan Berry, can you weigh-in on this one? Is everything playing correctly at both 44.1 and 48 multiples, but just the display doesn't recognize DSD256 at 12.288 MHz?

[Ryan Berry]

H

21 hours ago, The Computer Audiophile said:

Just testing out the Ayre QB-9 Twenty and found the DAC supports DSD256 at 12.288 MHz (multiples of 48 kHz). However, the front panel display goes blank with this sample rate. Playing DSD256 at 11.2896 MHz, the AB-9 Twenty front panel correctly says 256.

 

@Ryan Berry, can you weigh-in on this one? Is everything playing correctly at both 44.1 and 48 multiples, but just the display doesn't recognize DSD256 at 12.288 MHz?

 

Hey Chris,

 

The QB-9 has handlers written for DSD256 at 11.2896 MHz but not 12.288MHz right now.  We looked into this before releasing the QB-9 and could not find any source material that was recorded in a DSD format with a multiple of 48kHz, so it came off as more of a marketing gimmick to us than actually a useful feature and only adds confusion to what DSD256 is.  Adding a handler is REALLY trivial, as I imagine it would be for most DAC manufacturers.  The only hesitation is if there's any actual gain doing so.  If there's some source material out there we're missing, let me know and I'll talk with Ariel some more about it. 

[The Computer Audiophile]

21 minutes ago, Ryan Berry said:

H

 

Hey Chris,

 

The QB-9 has handlers written for DSD256 at 11.2896 MHz but not 12.288MHz right now.  We looked into this before releasing the QB-9 and could not find any source material that was recorded in a DSD format with a multiple of 48kHz, so it came off as more of a marketing gimmick to us than actually a useful feature and only adds confusion to what DSD256 is.  Adding a handler is REALLY trivial, as I imagine it would be for most DAC manufacturers.  The only hesitation is if there's any actual gain doing so.  If there's some source material out there we're missing, let me know and I'll talk with Ariel some more about it. 

Thanks Ryan.

 

This is more about using apps like HQPlayer to upsample content, using different filters and modulators, to DSD256 and keeping the sample rate as a multiple of the original. I assume the QB-9 upsamples internally to a multiple of the base rate. This is very similar just done outside the DAC and enables the user to select what he prefers sonically.

[jabbr]

7 minutes ago, The Computer Audiophile said:

Thanks Ryan.

 

This is more about using apps like HQPlayer to upsample content, using different filters and modulators, to DSD256 and keeping the sample rate as a multiple of the original. I assume the QB-9 upsamples internally to a multiple of the base rate. This is very similar just done outside the DAC and enables the user to select what he prefers sonically.

 

I gather than when converting from say 24/96 PCM to say 44.1x256 DSD that certain filters need less processing when staying within rate family, yet I have not heard that the final result depends on staying within rate family. For chip families such as ESS which operate internally at ?100Mhz we aren't staying with a multiple of either 44.1 nor 48Khz regardless.