[MKHunt]

It can handle DSD128 but 256 is where it all falls apart.  The Gustard locks on to the 256 and 512 signal but listenable sound just isn't there.  I'm seriously thinking it's the I2S lines since the audiophonics provided jumper cables are about 8" long and the cards, for now, are sitting directly below a monitor.  Tomorrow I get my solder wick and new tips for my Aoyue 937+ which, at 45 watts IIRC, should be able to handle quick desoldering so I can remove the pin header from the DIYinHK board, break it into the right size, and directly solder the audiophonics card to it, eliminating the weak link.  From my reading, LVDS signals are fairly robust in comparison to the pure I2S from the card.  That should also allow me to break/make a header tor the DIYinHK screen so I can watch the VU needles go crazy and see what signal the card is outputting.
 
I also spoke with audiophonics about power draw (I don't have a .1-ohm resistor handy to measure across) and apparently the power draw is around 1A on the 5V line.  That board is seriously thermally efficient.  Unfortunately, that also eliminates the possibility of paralleling the power supply rails for thermal management.  Perhaps it's time to make a benchtop supply out of an old ATX psu.

 

[SimonBromley]

Hey MK

Reading through the data sheet for that LVDS chip, the performance charts indicate
that it uses less than 40mA at full speed. Add in the losses for the onboard 5>3.3v 
regulator etc, and I just don't see how the Audiophonics card could be drawing more
than 100mA max.

This would tie in with my own observations. I too have one of the DIYinHK 0.8 PSU's
My WaveIO card draws about 400mA. And when it was powered by the DIYinHK PSU
both the choke and the silver transistor heatsink used to get really hot.
The Audiophonics card was powered by the other half of the PSU, and the choke and 
heatsink on that side always stayed pretty much at room temperature, indicating a
low real world current draw.
 
At one point, when I was trying to hack into the X20's internal board to board
LVDS network, I had the WaveIO's I2S lines hooked up to the Audiophonics card by
8 inch long flying leads. At this length I could clearly hear static and interference
coming through. So yes, I2S lines need to be short and if possible well shielded.
As for LVDS, being a differential signal means it's less susceptible to interference
over a distance. Which is why it's used to carry data between equipment. However, adding
the necessary transmitter and receiver chip, along with all the extra power components,
cabling and various bits of socketry, inevitably means your adding extra jitter etc
to the data signal.
 
That's why I wired my Xmos card direct into the PCIe slot's I2S lines inside the X20. 
The step up from using the HDMI input was quite significant.
 
I've finished modding my X20 now and moved on. And, whilst I think it's a great DAC for 
the money (with a bit of DIY work). After working on it, it's been annoying to realize 
that it could have been a truly outstanding DAC with just a bit more thought and 
attention put into it's development. Personally, I think them moving the 100Hz master
clock away from the DAC chips was a big mistake. In fact, if you upsample everything to DSD
like me, the entire digital board is pretty much just a redundant pile of electronics that
simply gets in the way, and degrades the DSD I2S signals on their way to the DAC chips.
 
If they'd only put an I2S header near the DAC chips, and left the clock there too. This
would have been one great DAC. As it is, it's merely a pretty good DAC for the money.
 
I'm still very happy I bought it though. It's been a fun ride and interesting learning
curve. It's now sitting in my video editing system, where hopefully it'll give years
of useful and productive service.
 
Regards
 
Simon

 

[Quadman]

As I followed Simon in his quest and just recently soldered in the accusilicon clock close to the dac chips I must agree wholeheartedly with what he says on how this dac could have been great out of the box with no mods needed.  This digital DIY thing to me is new, I am more experienced in crossovers and things analog.  I had to ask a lot of stupid questions to get to where I am now and nobody has done two mods step by step like I have and reported on the "improvements" along the way.  At times it was very tedious and listening sessions turned into analyzing sessions comparing A vs B, and after a few years of just blissful listening it re-sharpened my listening skills to hear little differences between components.  It was/is a fun learning experience and at times very frustrating and I am glad I got to the point I am now at with this dac.
 
The clock, last night was my first listening session after soldering it in and letting it run for 24 hours with a signal.  It is in my opinion a very big upgrade (re-enforcing Simons comments about clock location) especially after going backwards to the stock coax cable.  The custom silver coax cable is a definite upgrade over the stock coax, the clock proximity to the dac chip is an even bigger upgrade.  Using  typical audiophile terms, the glass is now cleaner, less distorted, background noise has even further diminished leaving performers more solid and 3D with greater space between them, leading edges where better defined and had more energy and impact, bass I thought was also improved.  Piano which I have mentioned sounds wonderful at 512 had even more clarity, the felt hammer hitting the string and the soundboard sustain were the best I have heard in digital.  A reference recording I have listened to over 70 times (roon count) in the last 2 months is a voice and acoustic guitar in a studio, as the mods advanced his voice bouncing off the far walls would be more obvious, last night the sustain/decay of it was by far the best I have heard.  Each reference track I played had a bit more than before.  Suffice it to say I am very pleased.  Like Simon said, move the clock close to the dac chips and have a I2S header nearby and wow.  Our I2S lines are still going into the PCIE holes, a long ways from the new clock but still excellent sound.  6 hours of listening last night at DSD512 and no noise, Whoo Hoo.
 
I am still not fully done with the mods, I have another clock to try the Abracon AOCJY1 a more power hungry clock that probably requires a separate external PSU and a regulator capable of 1A. @b0bb on the LKS forum says this is above the others he has tried to date.  I also want to externally power the accusilicon clock just for giggles.  

 

[MKHunt]

Okay 256 is up and running, but only in HQPlayer for some reason.  I think DSD in JRiver might be bugged.  Somehow 256 in JRiver only uses 8% CPU (i7 6770k, 16gb DDR4 @ 3200MHz).  512 is still elusive. 
 
I do NEED PCM for skype meetings, youtube, amazon video, etc.  Maybe the way to go is to go full internal for DSD and then use an alternate input method (optical?) for PCM.  I have an EVGA Classified 4-WAY Z170 board so the onboard sound should be........ decent?  I dearly hope so.
 
Quadman where did you find an Accusilicon clock for the Gustard?  I tried searching online and wasn't able to dig up any sources at work anyway. 
 
Also did you solder the clock directly to the board or did you use a DIP14 socket so you can swap clocks?  I found a Crystek CCO-083-100.000 but stability is only 50Ppm.
 
I guess for the clock, what made you choose accusilicon?  There's Crystek, NDK, etc out there.  I don't know anywhere near as much as you, so I bow to your advice.
 
 
As for Simon, I greatly appreciate all the work you did.  I still have so much to learn but I'm slowly getting a handle on it.  DAC guts are completely new to me.
 
So just to recap, in case I decide to go full internal:
 
DATA - Gustard B5
LRCK - B7
BCLK - B9
MCLKSEL - B13 Is this even needed?
DSDDETECT - B15
MCLKOUT - not used?
3.3v and 5v GND - A3
+5V  - A4
 
Is that right?  Can I use the center GND pin on the XMOS for the 3.3 and 5v ground?

 

[b0bb]

   
I am still not fully done with the mods, I have another clock to try the Abracon AOCJY1 a more power hungry clock that probably requires a separate external PSU and a regulator capable of 1A. @b0bb on the LKS forum says this is above the others he has tried to date.  I also want to externally power the accusilicon clock just for giggles.  

A master clock board with powersupply might be useful. The one below is made by ACKO labs, the regulator is capable of up to 800mA.
Check with ACKO to see if it is sold without the XO and if will work with the AOCJY1.
 
This keeps wiring to a minimum for good RF performance.
 
https://sites.google.com/site/ackodac/home
 

 
The XO shown here is the Pulsar OCXO. I got the AOCJY1 as  part of the exercise to see if an OCXO is feasible for the ES9018.
 
The AOCJY1 performance was an unexpected surprise, however the one I got from Mouser took a very long time to settle in, about 4 weeks.
The sound was transparent but very bright in the first 10days or so, it mellows out over the course of the next 2-3 weeks.
 
I got the Pulsar about 2.5 weeks ago and it is just about done settling in.

 

[Quadman]

@MKHunt That is a good idea, to use another input for PCM.  Simon has tried a lot of ways to improve the DSD signal so listen to him.  As far as your wiring thoughts, Simon has a older dac with Dec 15 or early jan 16 production he had to reverse B7 and B9 which is as you list.  I, who has probably the next batch made late Jan 16 or Feb 16 found I need to go exactly per the schematic.  So B5 is data, B7 is BCLK and B9 is LRCK.  I did not use either B13 or B17.  For the grounds at A3, I used the center pin of the DIY 3 pin header by the USB input (which is ground, 5V) and then took the ground from my external 3.3 Vdc PSU and also connected to A3.  +3.3Vdc to B15, and +5Vdc (USB handshake) to A4. I actually soldered some wires to the bottom of the board at A3, A4 and B15.  Make sure you have return grounds for the DSD lines I used the A side for the grounds.  I personally would not use the 3.3V header pin on the DSD512 card for the 3.3V B15 connection, even tho it may be okay.  I prefer the external PSU.  When I had the digital board out I removed the AES input so the USB cable can fit inside the dac and my up tone regen can sit outside.  I have made my own custom 6" usb cable for this which I must say I am very pleased with. And keep you i2s lines short and use silver wire 24-28 ga.
 
As to the clock I actually contacted accusilicon and was able to get 2 samples of them, I had to pay freight from HK so not sure what they end up costing me.  I have been told by accusilicon they will offer the clocks for direct sale off their web site soon.  I picked this clock because newer dacs have this clock installed and my friends dac which I modded and reported on had it and I was familiar with the sound.  Check out this link for some info on clocks Post #274 http://www.head-fi.org/t/745032/lks-audio-mh-da003/270

 

[Quadman]

b0bb, I followed the link but I don't see that master clock board on the site, am I missing something?
 
Quote:

  A master clock board with powersupply might be useful. The one below is made by ACKO labs, the regulator is capable of up to 800mA.
Check with ACKO to see if it is sold without the XO and if will work with the AOCJY1.
 
This keeps wiring to a minimum for good RF performance.
 
https://sites.google.com/site/ackodac/home
 

 
The XO shown here is the Pulsar OCXO. I got the AOCJY1 as  part of the exercise to see if an OCXO is feasible for the ES9018.
 
The AOCJY1 performance was an unexpected surprise, however the one I got from Mouser took a very long time to settle in, about 4 weeks.
The sound was transparent but very bright in the first 10days or so, it mellows out over the course of the next 2-3 weeks.
 
I got the Pulsar about 2.5 weeks ago and it is just about done settling in.

 

[Quadman]

b0bb, I finally found the master clock board.  $750 USD, I know that includes the clock too but wow thats expensive

 

[MKHunt]

Well, toslink cable ordered, so I guess I'm truly following in Simon's footsteps.  In other news, I found an interesting Abracon clock.
 
http://www.mouser.com/ProductDetail/ABRACON/AOCJY1-100000MHz/?qs=sGAEpiMZZMsC2cQJVRBSBUVTnCXPHX6IDcln7PotGtA%3d
 
Pinouts match up, voltage matches up, and the frequency stability is 50 parts per billion rather than parts per million (.05ppm stability).    The noise numbers look pretty good too. -90dBc/Hz at 10Hz and -150dBc/Hz at 10kHz.  It seems to edge out the Crystek offerings, though at $62/piece compared to $35/piece it probably should.
 
Hahah whoops, that's the exact clock b0bb was talking about.

 

[Quadman]

  Well, toslink cable ordered, so I guess I'm truly following in Simon's footsteps.  In other news, I found an interesting Abracon clock.
 
http://www.mouser.com/ProductDetail/ABRACON/AOCJY1-100000MHz/?qs=sGAEpiMZZMsC2cQJVRBSBUVTnCXPHX6IDcln7PotGtA%3d
 
Pinouts match up, voltage matches up, and the frequency stability is 50 parts per billion rather than parts per million (.05ppm stability).    The noise numbers look pretty good too. -90dBc/Hz at 10Hz and -150dBc/Hz at 10kHz.  It seems to edge out the Crystek offerings, though at $62/piece compared to $35/piece it probably should.
 
Hahah whoops, that's the exact clock b0bb was talking about.

I have that clock but need a bigger regulator before installing, I will power that externally since it is so power hungry, .6A at startup and .3A after (this crystek and accusilicon draw 25mA).  The regulator on the digital board is only 1A and I am not sure what else is on the load, tho it does run cool when dac is on.  The regulator on the output board is rated at 3A and also feels pretty cool, but since that is the only regulator I can see, it is powering other stuff and the .6A may be a bit much.  So I will eventually use one side of the DIY psu to power that, I already replaced the stock choke with a .8A choke, then I need a big regulator near the clock, sparkos maybe.  If we really want to kick things up a notch we need OCXO's with IHR (integrated heating resonators) like the pulsar clocks, but they run $300+.

 

[b0bb]

  b0bb, I finally found the master clock board.  $750 USD, I know that includes the clock too but wow thats expensive

That price is for the completed unit, ACKO is supportive of DIY, they may sell you just the board if you ask nicely.

 

[kaylu]

Hey HeadFi, just thought I'd give my review of this DAC. I bought the Gustard X20 because of this thread's positive comments. I am not a headphone user but I appreciate this community nonetheless so I want to contribute when I can.
 
I'm writing this review because the previous comments about how awesome this DAC is are correct in my view... but they didn't paint the whole picture for me and my tastes.
 
My previous system: JRiver Media Center PC > Schiit Wyrd (USB cleaner) > Gustard U12 (USB>SPDIF converter) > Schiit Bifrost Multibit (~600 USD) > Coincident Dynamo EL34 tube amp > Vienna Acoustics Bach floorstanding speakers.
 
Basically the Gustard X20 is a superb DAC. The sound is huge; palpable and weighty with a deep, powerful bass. It immediately jumps out as a luxurious, high value presentation. The sound was so big I was smiling involuntarily. DSD sounds especially awesome, so clear and tangible. I listened using the Gustard U12's HDMI I2S output into the X20.
 
I can't imagine a better DAC under $1000. It's a monster. Seriously, the sound is massive.
 
That said, this DAC was not for me. I auditioned it against a Metrum Musette ladder DAC (~1300 USD) and greatly preferred the R2R sound. The Musette's sonic character is almost identical to the Bifrost Multibit, just clearer with better image focus, air, and depth. Instruments sound authentic, like listening to microphone feeds in a recording studio. The Musette sound isn't huge or mindblowing like the X20... but to my ears it's right.
 
I mainly listen to acoustic jazz and classical (mainly symphonies, concertos, and solo piano). 3D reverb and realistic instrument timbre are my highest priorities.
 
My problem with the Gustard X20 is that all instruments had a comparably poor timbre, like the difference between an electric piano and a real one. The difference was not subtle at all. Along the same vein, reverb and the studio recording ambience were sorely lacking for my taste. Collectively the X20 made everything sound like synthesized instruments, not like microphone feeds coming from a real room.
 
I chalk all this up to the difference between Sabre and R2R. This conclusion may be an oversimplification but even so I don't see myself trying another Sabre-based product.
 
If I had never heard ladder DACs... or mainly listened to rock/EDM/pop/rap and didn't care about hearing an acoustic soundstage that sounds like all the musicians are playing in one room together... I'd be utterly thrilled with the X20. It just doesn't suit my priorities at all. I consider it a fantastic product just the same. The sound is HUGE and DSD sounds noticeably smoother, blacker, and fuller than lower resolutions.
 
 
One note of caution. I bought the Gustard DAC directly from Shenzhen Audio's website. If I didn't like it I expected to return it within 30 days. At the time of this post their website policy clearly states I should expect a full refund, minus shipping. But when I contacted them to find out exactly how much they would deduct for shipping they said I'd get 80% of the sales price back ("because it's opened we have to send it back to the factory") and they wouldn't state what the shipping deduction would be.
 
The customer support rep I traded emails with was clearly not fluent in English. When I pointed him to their own website policy he simply reiterated that I would not be receiving anything near a full refund. I'm in the USA and didn't have any real avenue to fight them on this. And I wasn't about to spend $150 (yes) shipping it back to them for some undetermined refund amount. So Instead I sold the DAC on eBay to a lucky buyer. Future customers: be advised.

 

[Triplefun]

My problem with the Gustard X20 is that all instruments had a comparably poor timbre, like the difference between an electric piano and a real one. The difference was not subtle at all. Along the same vein, reverb and the studio recording ambience were sorely lacking for my taste. Collectively the X20 made everything sound like synthesized instruments, not like microphone feeds coming from a real room.
 
May be a bit late now but did you try the Gustard with DSD upsampling in realtime AND remember to drop the volume in keeping with the specs of your pre amp (300mV full output). Given the 6V output of the Gustard you would need to drop the Gustard output  volume by about 15db !

 

[MKHunt]

Kaylu if you sold your X20 on Jul 21 then I now have it.

 

[kaylu]

  May be a bit late now but did you try the Gustard with DSD upsampling in realtime AND remember to drop the volume in keeping with the specs of your pre amp (300mV full output). Given the 6V output of the Gustard you would need to drop the Gustard output  volume by about 15db !

 
Perhaps you may be confusing things? My Metrum Musette puts out 2v RMS and Gustard X20 supposedly puts out 2.3v RMS single ended. I didn't hear anything to indicate these figures are incorrect; their volume levels were pretty close.
 
I used the DAC (and all DACs) at its full output volume; the digital volume control in both my software and the DAC were effectively disabled. I used the analog attenuator (typical Alps resistor knob) in my 8w tube amp to reduce the signal as needed for my listening comfort.
 
I didn't upsample any PCM to DSD. I don't believe I should have to trick a DAC - or tax my computer - just to make it sound acceptable. And anyway I don't expect that switching to high octane gas will turn a Cadillac truck into a BMW coupe...
 

  Kaylu if you sold your X20 on Jul 21 then I now have it.

 
Well if you saved a few bucks on it... could be. Looking forward to your Head-Fi review.