[new project] The BIY ("Buy It Yourself") High End DAC
Jan 5, 2008 at 8:43 AM Thread Starter Post #1 of 45

Atchoum

Head-Fier
Joined
Sep 19, 2006
Posts
91
Likes
10
[size=medium]PART 1 : the research[/size]

I'm building myself a (hopefully) high-end DAC and want to share back my findings because they mostly derivate from information gathered here.

My requisites after looking at several DIY DAC were :

- upsampling
- coax input with CS8416 (everybody seems to agree this is the best)
- DAC in dual mono
- price no object provided the value of component is "proven" (ie no voodoo black box inside)
- single ended
- I2S input for further tweaks
- "must" be better than my modded (with lm49710 for output stage) yulong mini-DAC (pcm1798)
- no need for an headphone amp (I own a Corda Opera
smily_headphones1.gif
)
- easy to build (no SMD soldering) because I'm very left-handed !

So I went through reading all the DIY DAC layouts threads, commercial products, DAC reviews...here were my findings :

- top commercial DAC are : Lavry DA10, Benchmark DAC1, Cambridge 840C
- top "DIY" DACs are : Opus Kit, Oritek & Zapfiltered Zhaolu, a lot of PCM1792/4 kits

So let's have a look at the insides of theses top DAC :

DA conversion :
- AD1955 for the DAC in the Lavry, the Cambridge (+ a special upsampling chip licensed from anagram tech). It looks like it is the most clean sounding DAC chip. That's what i'm looking for !
- Wolfson 8740/8741 for the Opus
- PCM 1792/94/98 in most other DAC


Output stage :
- Discrete analog stage seems to be a must and is present in almost all the top DAC (opus ballsie, oritek Zhaolu etc...)

PSU :
- Good psu (separated for the analog and digital part) seems a must and is found in most of the mentionned DAC

[size=medium]PART 2 : conception[/size]

The rave review (some are preferring it as a DAC over the Lavry !) the Cambridge 840 caught my attention so I went to look at it special upsampling chip and what I found is that its manufacturers do sell a evaluation board (http://anagramtech.com/getdoc.php?do...um=S2M-PL-100A) for this technology. The spec are dead on with my project with a CS8416 as a receiver, 384khz upsampler, quad AD1955 for the DAC, possible output to a custom analog stage. So I ordered one.

I choosed the zapfilter 2 for the output because it comes with its own psu and it gets rave reviews.

The only missing part is a PSU for the DAC board. I choosen a gold mini regulator (GOLD Reference Mini Regulator) for the +12V because of the great reviews this unit gets and because it is not too expensive.

Right now everything has be ordered and I can't wait to received everything and to try !

[size=medium]Part 3 : the building[/size]

Final Design is :

- Gold Mini Regulator +12V
- Anagram Tech EVAL Board with : CS8416 > 384kHz upsampler > dual AD1955 DA conversion
- Zapfilter 2 with its own PSU

in a few weeks !
 
Jan 5, 2008 at 2:12 PM Post #4 of 45
Quote:

Originally Posted by ferds /img/forum/go_quote.gif
How much is the Anagramtech Eval board?? tnx


495€. See : http://anagramtech.com/shop
 
Jan 5, 2008 at 2:51 PM Post #5 of 45
Quote:

Originally Posted by Atchoum /img/forum/go_quote.gif
[size=medium]PART 1 : the research[/size]


- DAC in dual mono
.
.
.
- single ended



Which is it? You want dual mono, but then convert to single ended? If you're going to the bother of dual mono (i.e. balanced), why make it single-ended on the output?
 
Jan 5, 2008 at 3:06 PM Post #6 of 45
Quote:

Originally Posted by ezkcdude /img/forum/go_quote.gif
why make it single-ended on the output?


...because my current amp is SE !
biggrin.gif

but I kept the dual mono layout because I "may" try full balanced once
 
Jan 5, 2008 at 3:09 PM Post #7 of 45
Atchoum

you'r planning to let the evaluation board do everything but the analog out?

I cant really comment on it as i havent heard the board nor the cambridge player but the website intrigued me.
I asume you have read trough their papers so,
this "Q5"module is used in their upsampler as well as in their masterclock where they claim to supress jitter below 1ps ,
Is the S2 board operating on the same masterclock? Are you planning to use clock injection to your source? And why not go I²S in the first place?

If the board Does offer the masterclock qualities aclaimed by their clock module and a wc out it seems interesting
 
Jan 5, 2008 at 4:43 PM Post #8 of 45
Quote:

Originally Posted by threEchelon /img/forum/go_quote.gif
From what I've read, it seems like an R2R DAC like the PCM1704, a resistor I/V stage, and a tube or transformer output stage is the best.




I agree 100%.
 
Jan 5, 2008 at 9:00 PM Post #9 of 45
Quote:

Originally Posted by utilisateur /img/forum/go_quote.gif
you'r planning to let the evaluation board do everything but the analog out?
this "Q5"module is used in their upsampler as well as in their masterclock where they claim to supress jitter below 1ps]



The eval board does not embody their proprietary "timelock" clock which claims to provide extremly low jitter below 1ps. But the onboard Q5+Sonic modules do reclock (from a local "high quality clock") the I²S signal and this is claimed to help reduce jitter from source. See : http://www.anagramtech.com/wordpress...upsampling.gif

I'll therefore start with the whole board (except the analog out) including the "local" clock. Then i may bypass the digital input to use my own I²S input (using TwistedPear's I²S sources and selector to come).

I do not feel the need to inject clock to my source since their Q5 module is here to remove source's clock effects.
 
Jan 6, 2008 at 8:30 AM Post #10 of 45
There is much more to no compromise performance than you might think, it is not as easy as simply putting together a bunch of parts. While this will probably sound quite good, is not the way to go for a no compromise, cost no object solution.

Quote:

Originally Posted by Atchoum
- coax input with CS8416 (everybody seems to agree this is the best)


The best by todays standard is the WM8804. The jitter rejection of most other recievers look like a joke compare to this.

Quote:

Originally Posted by Atchoum
- easy to build (no SMD soldering) because I'm very left-handed !


The DAC, SPDIF and whatever upsampler you choose will be in SMD package. You will also need SMD capacitors for decoupling for the DAC, SPDIF and upsampler. This is not an option.

Quote:

Originally Posted by Atchoum
So I went through reading all the DIY DAC layouts threads, commercial products, DAC reviews...here were my findings :

- top commercial DAC are : Lavry DA10, Benchmark DAC1, Cambridge 840C
- top "DIY" DACs are : Opus Kit, Oritek & Zapfiltered Zhaolu, a lot of PCM1792/4 kits

So let's have a look at the insides of theses top DAC :

DA conversion :
- AD1955 for the DAC in the Lavry, the Cambridge (+ a special upsampling chip licensed from anagram tech). It looks like it is the most clean sounding DAC chip. That's what i'm looking for !
- Wolfson 8740/8741 for the Opus
- PCM 1792/94/98 in most other DAC



The big advantage with the WM8740/1 is that it is voltage output, this is much easier to handle for (I assume) a first timer like you. The WM8741 is very good, a definite step up from the WM8740.

Quote:

Originally Posted by Atchoum
Output stage :
- Discrete analog stage seems to be a must and is present in almost all the top DAC (opus ballsie, oritek Zhaolu etc...)



The Ballsie is not discrete, it uses an opamp.

Quote:

Originally Posted by Atchoum
PSU :
- Good psu (separated for the analog and digital part) seems a must and is found in most of the mentionned DAC



A good power supply is critical for no compromise performance. A LM317 or LT1085 is not enough here. You need a Jung Super Regulator, like this for example. You should use four of them at least, one +5V for analog section of the DAC, one +3.3V for the digital section, SPDIF reciever etc., and two +/- 15V for the output stage.

One thing you have also missed is a low jitter clock. This is of utmost importance for sigma delta DACs. A very good (and cheap) alternative is the TentLabs XO clock module. This also needs a good power supply, you might even want to use a dedicated +3.3V super regulator.
 
Jan 6, 2008 at 1:35 PM Post #11 of 45
Hi Cauhtemoc and thank you for all your advices !

About SMD soldering : I've choosen Anagram's eval board because all (receiver, upsampler, clock, DACs) is already soldered !
smily_headphones1.gif
Lazy me...The other thing that made me look at this board is that it is quite easy to bypass a part and to add another. I may quite easily bypass the onboard CS8416 receiver and feed the upsampler with one WM8804 as you suggested.

I had considered the Opus kit (with the WM8740 and now WM8741) and while I'm sure it is easier to do something with the WM8740 / 41, i wanted to try an AD1955 based because the cambridge 840c and the Lavry that I do like very much use this chip.

For power supply units : I do understand it is important, a review on a french forum did an extensive comparison between JungsSuper Regulator and many others and found the Gold Mini Regulator to sound the best in a DAC thus my choice. I'll start with 2 psu (one for the digital + receiver) and one for the output stage. I may add more later following your advices.

I did not missed the clock : one "high quality, low jitter local clock" is availaible on the board and its easy to add an higher quality clock later (bnc connector + jumper already connected on the board).
 
Jan 6, 2008 at 8:47 PM Post #12 of 45
Quote:

Originally Posted by Atchoum
About SMD soldering : I've choosen Anagram's eval board because all (receiver, upsampler, clock, DACs) is already soldered !
smily_headphones1.gif
Lazy me...



You're right, that is lazy.
tongue.gif


Quote:

Originally Posted by Atchoum
I may quite easily bypass the onboard CS8416 receiver and feed the upsampler with one WM8804 as you suggested.


Does Anagram's eval board have their TimeLock technology built in? If so there is little to gain by changing the CS8416 for a WM8804. The advantage of the WM8804 compared to other SPDIF recievers is that it works very good straight out of the box, it's jitter rejection is far, far better than anything else commonly availible. TimeLock works somewhat better still, regardless of the actual SPDIF reciever used.

Quote:

Originally Posted by Atchoum
I had considered the Opus kit (with the WM8740 and now WM8741) and while I'm sure it is easier to do something with the WM8740 / 41, i wanted to try an AD1955 based because the cambridge 840c and the Lavry that I do like very much use this chip.


The thing with AD1955 is that for ultimate performance you need a proper discrete I/V converter. The way Zapfilter does it is okay, but it's not the best.

Quote:

Originally Posted by Atchoum
For power supply units : I do understand it is important, a review on a french forum did an extensive comparison between JungsSuper Regulator and many others and found the Gold Mini Regulator to sound the best in a DAC thus my choice. I'll start with 2 psu (one for the digital + receiver) and one for the output stage. I may add more later following your advices.


Do you have a website to this forum, and perhaps to where these regulators are sold?
 
Jan 6, 2008 at 9:16 PM Post #13 of 45
Quote:

Originally Posted by Cauhtemoc /img/forum/go_quote.gif
The advantage of the WM8804 compared to other SPDIF recievers is that it works very good straight out of the box, it's jitter rejection is far, far better than anything else commonly availible. TimeLock works somewhat better still, regardless of the actual SPDIF reciever used.


As far as I can read in the datasheets, the Texas Instruments DIR9001 is at least as good as the wolfson part. The graphs about jitter are also more detailled, with wolfson throwing in only one measurement. And the DIR9001 is cheap at digikey.


Edit: Wrt regulators for DAC, it's worth spending some time on pinkfishmedia 's diy forums.
 
Jan 6, 2008 at 9:17 PM Post #14 of 45
Quote:

Originally Posted by Cauhtemoc /img/forum/go_quote.gif
Does Anagram's eval board have their TimeLock technology built in?


Not the Timelock module itself which is nothing but a very good clock 'as far as I understand) but the "reclocking" solution is built in their Q5 upsampling technology using the local clock (which "should" be not as good as their Timelock but still on par with a XO clock I think)

Quote:

Originally Posted by Cauhtemoc /img/forum/go_quote.gif
The thing with AD1955 is that for ultimate performance you need a proper discrete I/V converter. The way Zapfilter does it is okay, but it's not the best.


This is interesting ! what do you think of exactly ?


Quote:

Originally Posted by Cauhtemoc /img/forum/go_quote.gif
Do you have a website to this forum, and perhaps to where these regulators are sold?


The thread is here : homecinema-fr.com : Voir le sujet - Alimentations ( DAC ) Faisons le point
The Mini Reg can be found here : GOLD Reference Mini Regulator
 
Jan 6, 2008 at 11:13 PM Post #15 of 45
This is an option for discrete output. And, one is availble in the Source FS section now.
 

Users who are viewing this thread

Back
Top