Phison Audio PD2 Preamp/DAC

Looks interesting. Any idea why the AK4490 was used instead of an FPGA or AK4497?

I'm just guessing since the AK4497 was just announced at CES Feb 2016 and their site notes a mass production start this summer, so my guess the new chip wasn't available during the design phase of a lot of DAC companies like Phison and Bryston.
 
Looks interesting. Any idea why the AK4490 was used instead of an FPGA or AK4497?

Also, what other DACs have your customers compared the PD2 with?

The AK4490 was used because it has a DSD direct bypass mode. So when DSD is sent to it, it bypasses the internal SDM/SRC processing of the chip, and gets filtered by high quality discrete filtering. The AK4497 wasn't used because it won't be out till the summer. Using my server upsampling all PCM to DSD, there won't be much of a difference with the AK4497 anyways because the DSD bypasses the chip. It's basically just being used as a conduit to direct traffic.

Anyways due to the highly modular form factor, upgrading the DAC board as newer technology comes out will be very easy and inexpensive. Even the output stage of the DAC is in 5 pieces on separate boards.

I have only delivered one so far, and 2-3 are going out next week. My first client had a Lampi GG, and before that an MSB DAC but not sure which one.
 
Thanks, both.

Looks like upsampling to DSD followed by gentle filtering appears to be gaining in popularity. I can understand the preference for PCM/DSD conversion done externally, albeit slightly more inconvenient, especially when one already has something like the Melco/Aurender/Lumin.

Interesting times indeed for top end DACs. To be competitive with the GG (based on what i have read elsewhere) at the current (introductory?) pricing is an achievement. Looking forward to more field reports.

Would love to hear about a comparison with the Nagra HD Dac.
 
Thanks, both.

Looks like upsampling to DSD followed by gentle filtering appears to be gaining in popularity. I can understand the preference for PCM/DSD conversion done externally, albeit slightly more inconvenient, especially when one already has something like the Melco/Aurender/Lumin.

Interesting times indeed for top end DACs. To be competitive with the GG (based on what i have read elsewhere) at the current (introductory?) pricing is an achievement. Looking forward to more field reports.

Would love to hear about a comparison with the Nagra HD Dac.

No problem. It will work fine with any of those above boxes. But a box that does SOTA upsampling to quad DSD like the box I have in the works will get the best out of it. No different than the DCS Vivaldi upsampler concept, except it will have the server built in, run the Roon core, and upsample to quad DSD rather than DSD 64. Price will be a bit more reasonable as well at under $2500.
 
I just heard that Phison has new distributor/dealers pending in Australia, Korea, Hong Kong, China, Singapore and U.K.
 
I think its smart. NAS direct to i2s. Skipping all the usb ground pollution and packet mess. Not sure how audio data can be delivered to a dac any better?
 
I think its smart. NAS direct to i2s. Skipping all the usb ground pollution and packet mess. Not sure how audio data be delivered to a dac any better?

I think it's awesome as well. The only shortcoming is the UPnP dependency. It limit's you to using only media players/servers with UPnP media server ability. No Roon or anything like that. Also no external SRC/SDM engines can be used in conjunction with UPnP such as Hqplayer. A system that takes advantage of the best of both worlds is Ravenna. You get the plug and play ability with any media software like USB, and also the Ethernet connectivity with extremely low jitter, and minimal processing in the renderer.
 
Here's an example to the approach with DSP I'll be taking with my server system. The capabilities are far and beyond what's possible from FPGA's and DSP chips like dCS, MSB, EMM Labs, PS Audio, Mola Mola etc use for their SRC/SDM DSP. The PD2, is the ultimate hardware platform to handle the quad DSD in the purest way once processed through the SOTA DSP engine in the server.

http://www.intervalzero.com/revolutionizing-real-time-development-with-rtx/

To put things in perspective, here's the processing power of the FPGA's dCS uses in their latest Vivaldi upsampler, compared to an Intel I7 quad core Skylake processor used with the Interval Zero RTX64 system:

attachment.php
attachment.php
 

Attachments

  • Replacement-Engine-for-Vespa.jpg
    Replacement-Engine-for-Vespa.jpg
    337.9 KB · Views: 85
  • bugatti-veyron-engine-5.jpg
    bugatti-veyron-engine-5.jpg
    123.3 KB · Views: 87
I think it's awesome as well. The only shortcoming is the UPnP dependency. It limit's you to using only media players/servers with UPnP media server ability. No Roon or anything like that. Also no external SRC/SDM engines can be used in conjunction with UPnP such as Hqplayer. A system that takes advantage of the best of both worlds is Ravenna. You get the plug and play ability with any media software like USB, and also the Ethernet connectivity with extremely low jitter, and minimal processing in the renderer.

Ya but I am about SQ first and foremost. Mimserver is good enough for me as far as interface goes. I would not marginalize SQ for interface. Also I just plain like what MSB does top to bottom in the dac from upsampling to filtering to conversion. Maybe someday Jussi will show me a reason to go a different direction but not today.
 
Ya but I am about SQ first and foremost. Mimserver is good enough for me as far as interface goes. I would not marginalize SQ for interface. Also I just plain like what MSB does top to bottom in the dac from upsampling to filtering to conversion. Maybe someday Jussi will show me a reason to go a different direction but not today.

I agree 100%. Next to Ravenna, I think it's the best Ethernet renderer I have seen yet (except maybe what's in the Bel Canto black system as it's the same thing). It uses an AD Blackfin DSP chip, and slaves to the master clock of the DAC. Extremely hard to beat that jitter wise with USB. I was ready to throw in the towel with USB as well until Sonny built this masterpiece of a USB implementation for the PD2. Now I just need to make my Ravenna interface outperform it :)
 
Here's a little information on why the USB implementation is so good on the PD2. First of all it's powered 100% by very clean power provided by very low noise, high PSSR regulators on the DAC board. It's isolated from USB noise by the best galvanic isolators available. It uses 2 very low jitter Crystek 957 clocks for the async operation. Because the galvanic isolators add a bit of jitter themselves, the clock is reclocked with a flip flop to eliminate 100% of the jitter caused by the galvanic isolation. So really the signal from the USB is galvanically isolated and reclocked twice. This process not only eliminates jitter, it also provides unparalleled noise isolation from the USB bus. The clocks are also powered with ultra clean power from their own dedicated ultra low noise, high PSSR regulators. Without powering the clocks with ultra clean power, no matter how good the clocks are, you will never get low jitter performance from them. Now on to one of the most important features. The DAC chip is synchronously slaved to the reclocked, clock #2 of the USB interface. This provides the lowest possible jitter performance between the DAC chip and the USB interface, just like the MSB Ethernet renderer does. Most DAC's don't do this. They use mediocre clocks for the USB interface, and send the jitter riddled signal to the DAC chip which is clocked by a different master. Other important features are 4 layer PCB with dual ground planes and extensive via stitching for the ultimate in ground noise isolation. And finally, the DAC chip is located in very close proximity to the master clock and USB interface. All of these attributes are a huge contributing factors to the astonishing performance of this DAC.

attachment.php
 

Attachments

  • Phison Audio PD2 DAC board.jpg
    Phison Audio PD2 DAC board.jpg
    284.6 KB · Views: 83
Here's the discrete modular DAC output stage/DSD filter board that goes on top of the DAC board. It uses 4 of the same discrete Jfet folded cascode gain stage/servo boards that the preamp uses. These are the best discrete gain stages I ever heard and I've tested many of them. It's designed not only so the gain stages themselves can be replaced, or upgraded in the future, but the backplane that has the SOTA discrete DSD filter components on it can be removed from the DAC and upgraded separately as well. This way when a new DAC comes out, the output board will still be compatible. Only the inexpensive main DAC board will need to be replaced to upgrade.

attachment.php
 
So I have a bit of feedback on the PD2 from my 1st client. While waiting for delivery of this Phison A2.120 amp, he borrowed 2 amps to try in the meantime. 1 was the Nord 500 with Sparko discrete opamp input stage, and the other was one of the higher end Bryston's (not sure the exact model). He also has new Magico S1MKII speakers he's breaking in. He will have a better evaluation of the unit once he has the matching amp.

He has given me permission to share this from an email he sent me:

"The PD2 really allowed me to differentiate between the two amplifiers but with the PD2 and Magico S1MkII it made both very enjoyable to listen to for long listening sessions. Even with the Bryston, at moderate volumes the music sounds so pure with that analog like sound. There seems to be just the slightest hint of warmth giving it a very natural type sound which I really like and it does this without sacrificing details which it seems to excel at (micro details). Although with different speakers at the time, this natural warmth with details reminds me of the MSB Dac but with that DSD liquidity of the Golden Gate Dac using it's DSD mode. To a lesser degree than either one (MSB Dac IV had more micro detail with PCM but was slightly drier, relatively speaking and Lampizator GG definitely sounded more smooth or I would describe as "wet"). The balance the PD2 strikes fits my preferences almost perfectly."
 
I forgot to give source gear info used in the above setup/impressions. What was used is a general purpose tower PC running Windows. For software Roon/HQplayer was used. This computer is located in an office, and connected via standard copper Ethernet cable, and a standard router to the $145 Intel NUC based streamer in this thread:

http://www.whatsbestforum.com/showthread.php?19221-Streamer-to-kill-the-big-boys-for-under-250

He is using the cheap and noisy $2 power supply that comes with the NUC, and none of the fibre optic noise isolation upgrades recommended in that thread.

2 others who have also tried that streamer with standard Ethernet gear, as well as the included supply are members of this forum. 1 is Dimfer who posted earlier in this thread, and the other is Gopher. Both reported lousy performance from the unit most likely due to noise from the Ethernet network, and the cheap switch mode power supply included with the NUC.
 
Hi guys,

I would just like to clarify a few things as my last post has seemed to trigger a frenzy of behind the scene activity. I'm not in any way suggesting Blizzards Superstream is capable of competing with dedicated streamers on the market costing $2000-20000. I have acknowledged from the beginning that to get top performance from the unit that it must be used with a very high quality LPS, and fibre optic isolation. However the fact that the level of sound that was heard by my first client was heard from the cheap base configuration was so good can only mean 2 things:

1: The USB implementation is so good in the PD2, that all of the shortcomings of the Superstream when powered by a the cheap SMPS and no fibre isolation is overcome

2: The PD2 will be even better if the source is a highly competent and proven commercial streamer such as an Aurender.

And to add to this, don't be afraid to comment openly in public. I have nothing to hide, and can answer any questions anyone has in a clear, concise, open fashion. I don't bite :)
 
Here's a little information on why the USB implementation is so good on the PD2. First of all it's powered 100% by very clean power provided by very low noise, high PSSR regulators on the DAC board. It's isolated from USB noise by the best galvanic isolators available. It uses 2 very low jitter Crystek 957 clocks for the async operation. Because the galvanic isolators add a bit of jitter themselves, the clock is reclocked with a flip flop to eliminate 100% of the jitter caused by the galvanic isolation. So really the signal from the USB is galvanically isolated and reclocked twice. This process not only eliminates jitter, it also provides unparalleled noise isolation from the USB bus. The clocks are also powered with ultra clean power from their own dedicated ultra low noise, high PSSR regulators. Without powering the clocks with ultra clean power, no matter how good the clocks are, you will never get low jitter performance from them. Now on to one of the most important features. The DAC chip is synchronously slaved to the reclocked, clock #2 of the USB interface. This provides the lowest possible jitter performance between the DAC chip and the USB interface, just like the MSB Ethernet renderer does. Most DAC's don't do this. They use mediocre clocks for the USB interface, and send the jitter riddled signal to the DAC chip which is clocked by a different master. Other important features are 4 layer PCB with dual ground planes and extensive via stitching for the ultimate in ground noise isolation. And finally, the DAC chip is located in very close proximity to the master clock and USB interface. All of these attributes are a huge contributing factors to the astonishing performance of this DAC.

attachment.php

Blizzy-poo,

Is that triceratops or stegosaurus? I am no expert in this stuff but I still think its inferior to ethernet direct to i2s and the "real" reason designers are stuck using it (and hence try to argue its pretty snazy) is cuz they haven't developed an eloquent way to get dsd through a direct i2s implementation like msb. My Femto 33 controls all clocking in the i2s input. Its a fairly good clock and it gots a fairly good lps. No matter how much you clean up a nasty ground and icky packet noise its hard to argue its better than a connection that is squeeky clean from the start. (Of course I plan to use ethernet glass with a linear ps.) Save the usb stuff for a museum.

Now don't bite, you promised...(I am just haven fun and trying to catch up on this stuff with the deep thinkers.)
 
Blizzy-poo,

Is that triceratops or stegosaurus? I am no expert in this stuff but I still think its inferior to ethernet direct to i2s and the "real" reason designers are stuck using it (and hence try to argue its pretty snazy) is cuz they haven't developed an eloquent way to get dsd through a direct i2s implementation like msb. My Femto 33 controls all clocking in the i2s input. Its a fairly good clock and it gots a fairly good lps. No matter how much you clean up a nasty ground and icky packet noise its hard to argue its better than a connection that is squeeky clean from the start. (Of course I plan to use ethernet glass with a linear ps.) Save the usb stuff for a museum.

Hi Paul,



Now don't bite, you promised...(I am just haven fun and trying to catch up on this stuff with the deep thinkers.)


I agree with you 100%. It was through my extensive research with USB interfaces starting back in 2012 with the Amanero that I discovered first hand the limitations of USB. After that I experimented with several more advanced interfaces until I finally hit a brick wall for performance. I couldn't find anything even close to as good as a well implemented SD card transport connected I2S direct to the DAC. I even got banned from Computer Audiophile for my strong objection against USB interfaces, and favoritism's towards Ethernet implementations such as MSB uses. Here's a couple great threads on it here:

http://www.computeraudiophile.com/f...tween-computers-and-electronic-devices-25467/

http://www.computeraudiophile.com/f...s-and-electronic-devices-vs-i2s-direct-25653/

So when I say I approve of this USB implementation, it's for a reason. But I agree with you that Ethernet has an edge. This is why I have been working on a Ravenna based Ethernet interface for quite some time.
 
Back
Top