Dec 4, 2014 at 5:30 PM Post #1,471 of 1,613
Much thanks for sharing your conversation with Amanero, as they did the actual software and firmware work on behalf Yulong, I hope it becomes a matter of waiting for Yulong to release this or offer some type of exchange for the USB interface cards.



Back to your original problem.......

I had a very similar problem to what you described so I am hoping this may provide some useful information on your issue

Yulong is one of the first companies to offer the Sabre32 DAC with 384kHz PCM sampling and higher than DSD128 operation.
The DA8 is at the bleeding edge as the Sabre32 was originally designed and qualified for 192kHz use, I was half expecting issues from the day I bought it.

I use the DA8 with Audirvana on the Mac upsampled to 352kHz or 384kHz and in stock form I will get periodic glitches and cutouts every 20-40mins.
This is only at the highest sampling rates, 192kHz was never an issue, similarly DSD128.

The behaviour was not consistent as Foobar with DSD256 on the PC via ASIO had a far lower incidence of glitching, so maybe this is why Yulong is not seeing the issue.



I fixed my DA8 by replacing the master clock oscillator and rebuilding parts of the powersupply regulator going to the Sabre32.

The clock fix is the most significant at reducing the glitching @352/384kHz. I have very limited DSD256 material so the observations is not complete but I did see any substantial glitching.
It went from 2-3times/hour to maybe once per listening session.

Fixing the power regulator brought glitching down to maybe once a week.

Clock fix
=======
It cost about USD$30 in parts from Mouser and about 20mins of my time.
If you have the soldering skills or can find someone to help you, the replacement clock is a Crystek CCHD950-25-100M.
Both Mouser (25ppm version) and Digikey (50ppm version) stock this part and they may deliver to your part of the world.

I used an adapter board so I can swap and try various XOs like the 82fs Jitter CCHD575 (aka "femto-clock")



Power regulator fix
===============
Yulong uses a 7805 5V pre-regulator for the Sabre's Analog Vcc (AVCC) supply.
The noise output is about 40uV but it suffers from poor transient response when handling the fast changing power demand of highspeed digital circuits.
The Sabre32 running at it maximum design limit is such a case.

The fix is to replace this with TI's TPS7A4700, this is readily available regulator with low noise (4uV 10x better than the 7805) and fast transient response.
It is a linear regulator designed for digital circuits.

There is a vendor in Hongkong that supplies these pre-mounted on a PCB with the 7805 form factor . They also sell on Ebay.

Cost is about $30USD but extraction of the existing 7805 is complicated, it took me about an hour.
Most of the time is spent slowly heating up the board and heatsink with hot-air to avoid damaging the PCB copper traces.
 
Dec 5, 2014 at 8:32 AM Post #1,472 of 1,613
Thanks a lot.
My da8 is under warranty (2 years, in europe), so, even with my sufficient skills, I'm quite afraid of modding it.
But I may consider the clock change, if it gives other "features" (because I had deterministic glitches, no random ones even in 3-4 hours of listening)
 
Dec 5, 2014 at 9:22 PM Post #1,473 of 1,613
Changing the master clock can alter the Yulong's sound.
One of the shortcomings of the Yulong is it can sound quite thin when there are complex passages as in classical symphonies or multiple voices in the case of a chorus or choir.

With rock music, it can bring out the sibilance in those recordings and this was promptly picked up by my HD800, not a happy combination in stock form.

I also wanted deep bass extension, some of my albums go down to between 20-30Hz, on regular speakers this a nice rumble thru the floor and up thru the listening chair.
On a headphone this is transmitted to the cheekbone, the HD800 is quite capable of doing that.

While delivering the bass, I wanted the bass transient to be clear, kick drums must display the transient of the foot hitting the membrane and not just a dull thud.
Similarly the high frequency plucking sound of the string on a bass guitar should be heard clearly.

The stock form's bass was no better than my M2Tech HiFace DAC which is a USB bus powered dongle.

The top end dynamics and extension on complex material was primarily influenced by the choice of master clock.
The master clock was important to bring out the top end clarity without overexciting the HD800 upper midrange tendencies @6kHz

Low end extension is a combination of the right master clock and rebuilt analog and digital power regulation on the DAC

The end result I was looking for was a nice balanced sound.
In order to make sure I also have other headphones
HD600: I wanted the top end veil lifted without losing the midrange sweetness or bass extension.
HE60/SR009: I wanted bass extension without losing the imaging accuracy ie no muddy flabby bass.

I believe this has been largely achieved as I can listen to the HD800 for hours at a time without wanting to take it off.
I did not believe that you could feel the bass coming out of the SR009 but this has changed my mind.

Below are the crystals I experimented with, all of which can be bought from Digikey or Mouser



Crystek CCHD950:
===============
This is my XO of choice on the Yulong and it gives the balance I was looking for.
I am using the CCHD-950X variant right now, it is a better specced version of the 950 series with a wider temperature range, in practice this means a better quality cut of the quartz crystal.
Overall result is a cleaner sound wit the top end extension coming close to that of the CCHD575

Crystek CCHD575:
===============
This is similar to the XOs offered as "femto" clock upgrades by several DAC manufacturers, this has a rms jitter of 82fs.

The imaging is more precise and the topend is more extended that the CCHD950 but the lower midrange and bass sounds thin in comparison.
Overall less balanced.

Some of this is due to how Yulong is supplying 3.3V to the XO, it comes off the regulator shared with the rest of the Sabre32.
On better implementations, the XO has a dedicated low noise regulator such as the ADP150.

At this point I consider that the DA8 is not able to make full use of the femto clock, I am thinking of making a small PCB with an board regulator to find out for certain.

Abracon ABLNO:
==============
The best specced stability of the bunch at 12ppm over temperature, 28ppm over the life of the device.
RMS jitter performance is also good at 100fs, this makes a "femto" clock class device
(This is over 20MHz as opposed to 100MHz measurement interval used by Crystek, so they cheated a bit.)

Sound quality is very close to the CCHD575 but a lot more punch on the lower midrange, bottom end performance does not come close to the CCHD950 that is I cannot feel the bass on the cheekbones.

Underlying cause is likely the same issue with the CCHD575, it needs a better and quieter power supply from the motherboard.

Epson XB42:
===========
This is a Surface Acoustic Wave (SAW) based XO, unlike the other three which are quartz crystal based, this one is a high stability ceramic resonator than is stable to 50ppm with temperature.
Phase noise/Jitter performance is not so great.

This provides a very lush sound that is addictive and very easy to live it, bass extension is better than the femto class clocks and is not very fussy about power supply.

Downside is it is not that accurate.
A Violin(E-string) may sometimes sound like a Viola (C-String),
I did not notice initially until I checked the album's liner notes and found it was a violin section making all the noise rather than a bunch of violas.
 
Dec 6, 2014 at 6:23 AM Post #1,475 of 1,613
TI'll try with the CCHD-950X. Is it "plug 'n play" with the original clock?

Yes, just orient the dot for pin1 on the motherboard and the dot on the XO

Here is where I bought mine
http://www.mouser.com/ProductDetail/Crystek-Corporation/CCHD-950X-25-100000/?qs=L4klnTtkofObrfUJUtC8ig%3D%3D

The 950X is SMT, an adapter to the 14pin DIP outline will also be needed.
Please let us know about the results of making the pcb with voltage regulation.

Will do.
 
Dec 6, 2014 at 11:56 AM Post #1,476 of 1,613
Yes, just orient the dot for pin1 on the motherboard and the dot on the XO

Here is where I bought mine
http://www.mouser.com/ProductDetail/Crystek-Corporation/CCHD-950X-25-100000/?qs=L4klnTtkofObrfUJUtC8ig%3D%3D

The 950X is SMT, an adapter to the 14pin DIP outline will also be needed.

 
Is this adapter sold by mouser, also?
 
Dec 6, 2014 at 4:34 PM Post #1,479 of 1,613
It is a 4 pin adapter in a 14 pin dip frame with only pins in the corner.

The Accutek part number for the CCHD-950X is AK14D300-XTAL-04SOJ 4 PIN SOJ CRYSTAL TO 4 PIN (14 PIN FULL SIZE) DIP
 
Dec 9, 2014 at 3:51 AM Post #1,481 of 1,613
I find it difficult to put a specific figure, the main improvement, balancing of the presentation without losing the clarity details and sound staging.
The midrange in the stock DAC is adequate but is overwhelmed by the top end especially on the HD800
In stock form it can fall apart on complex music both in classical and rock.

The XO change addresses mainly these short comings on regular 44.1KHz 16bit CD Redbook music

My main interest is in high resolution tracks such as 352kHz DXD and DSD128, I also have a collection of 24/96kHz and 24/192kHz DVDA type tracks
I upsample most of my redbook tracks to 352kHz

In stock form, the high resolution stuff has better definition, instrument separation but comes across as quite lean and somewhat analytical, the sweetness and lush soundstage is sometimes lost even with tube gear. Bass impact was also considerably reduced when the higher sampling rates were used, particularly on upsampled material.

This is an exercise to make the DA8 work to my satisfaction on 352/384KHz material. I am looking for the attributes of the unsampled redbook audio (deep bass, sweetness and lushness) without losing the benefits of the high sample rates (top end clarity, soundstaging and overall improvement in pacing )

The XO change is one of the aspects of this effort, I also rebuilt the analog and digital powersupplies and voltage regulation on the DA8, the latter to improve bass and lower mindrage preformance. The Sable32 demands for low noise and current supply capability increases exponentially with increasing sample rates.

The DA8 is only part of the audio chain, I also rebuilt the headphone amp, I use the A28 . I wrote this up in the A28 thread.

The modified DA8+A28+HD800 combo is a serious alternative to my other setup DA8+SRM-007t2+SR-009
 
Dec 10, 2014 at 5:11 AM Post #1,482 of 1,613
In order to solve a noise problem for DSD256, there is a new  DA8 firmware:
 
http://www.yulongaudio.com/en/down.asp ( DA8 DSD256 firmware_update )
 
The installation Requires a PC and the opening of the DAC...
 
Dec 10, 2014 at 5:19 AM Post #1,483 of 1,613
ps yes, I was frustrated, because I'm not a rich audiophile, and paying 1149 euros for the unit, without "necessary" support (I was an IT engineer, and, simply, I can't conceive of the lack of support on something involving drivers and firmware) for me is too much.

According to the european laws, it is the seller who is responsible for the customer care and support.
 
Dec 10, 2014 at 5:21 AM Post #1,484 of 1,613
  Yes, I wasn't surprise, also. But don't trust blindly on "local" companies, maybe I'm too much categorical, but I'm quite sure that each dac producer has support (in terms of fast bug fixing or new formats implementation) difficulties. Sometimes justifiable, but in most cases, specious.

I had to send two times my Mytek from Italy to Poland for "assistance"...
 
Dec 10, 2014 at 5:28 AM Post #1,485 of 1,613
   not to include a button sequence to start a firmware reset. It is really simple, and a lot of other dac producers supply it. 

Again: Mytek no button and firmware update only with the firewire 
wink_face.gif

It's a difficult world 
bigsmile_face.gif

 

Users who are viewing this thread

Back
Top