[Viper2001]

Quote:

Isolation of a DAC from noise sources is far more important than jitter in my experience. The noise sources to watch
out for are the mains supply itself, and PCs. The average DAC has a connection to both. Galvanic isolation most
certainly can help a lot but the isolation is not complete at high frequencies, so its more important to know the
capacitance across the isolation barrier than know whether the isolation is of the 'galvanic' kind, or not.
 
If you look at the pics Viper2001 has posted of his D2, you'll see that it has a 'block' type mains filter (top left). In
my experience these aren't really adequate in taming mains-borne noise. It also shows a toroidal type mains
transformer (the blue thing with fins, bottom left). Designers keen to reduce mains noise don't use such trafos - a
good example is the extremely well regarded Berkeley Alpha DAC. Its transformers look positively pre-historic in
comparison with this one, but they've been chosen for good reason - isolation. Toroids have the two windings
(primary - mains, secondary - low voltage) wound over the top of each other. This means high interwinding
capacitance - an open window to mains noise.
 
If I was a D2 owner (and I wouldn't buy because I'm a multibit addict, not S-D) then I'd be looking into optimizing the
noise immunity before the clocking. That's not to say the clocking is unimportant, just getting the priorities right.

 
AC noise issue can be troublesome for some people, depending on where you live, your neighbors, etc., but some of the points you made are not quite valid.

For example, the blue fin transformer does not have high interwinding capacitance.  The interwinding capacitance starts to become a problem when you're into hundreds of VAs.  This is why it is better to use EI transformers in power amplifiers instead of toroids.  But for low consumption unit like the Anedio (10VA), it is not a problem.

The problem of leakage inductance on those cheap blue fin toroid is actually the bigger issue.  Cheap toroids are like that.  But James has done the RC damping network on the output of the transformer to kill leakage inductance induced noise spikes, so even if you replace it with a higher end unit, such as R-core, the gain will be very little.  The problem with R-core would be higher radiated stray magnetic field, so unless you can make it far away, it is going to affect the clock circuit.

The hum buckers used in Berkeley are mechanically noisy (hum).  So you gain some and you lose some.

Also, contrary to popular beliefs, the most damaging noise does not come from AC, it actually comes from inside your gear - the rectifiers (diodes).  Silicon rectifiers or fast recovery rectifiers are "noisy".  They emit tons of high freq noise at the moment of switch off.  The smart designers will put a snubber across each rectifier to damp out this noise.  James did the right thing on using the Cree SiC recitifiers, they have nearly zero switching noise.  And the other rectifiers he used for the display are of soft recovery type.

My assessment of D2 is that power supply related noise is not a problem on this unit.  The problem is actually the clocks on U2 and WM8805.

 

[Viper2001]

Quote:

So when're you gonna start your Anedio D2 modding business?   

 
Have you noticed the AC voltage setting on my D2?  I'm not in your part of the world.  I don't know how many US folks are willing to ship their units to me for modding...

 

[Viper2001]

Quote:

Just to be crystal clear, if I run a signal to my D2, (via Sonos) but without the pre/pro and amp turned on, it is still considered as " burn in " ?
 
I've been running my Sonos 24x7 for 12 days straight to the D2, with about 2 hours of actual listening via pre/pro + amp + speakers every day.
 
Doing the math, I am assuming that I have put 288 hours of "burn in" on my D2.
 
Is that correct ?
 
Or is burn in only when it's actualy playing through my speakers ?

 
Having the unit on is fine.  I'd consider your unit as "burn-in" at this point.

 

[Sapientiam]

Quote:

 
AC noise issue can be troublesome for some people, depending on where you live, your neighbors, etc., but some of the points you made are not quite valid.

For example, the blue fin transformer does not have high interwinding capacitance.  The interwinding capacitance starts to become a problem when you're into hundreds of VAs.  This is why it is better to use EI transformers in power amplifiers instead of toroids.  But for low consumption unit like the Anedio (10VA), it is not a problem.

The problem of leakage inductance on those cheap blue fin toroid is actually the bigger issue.  Cheap toroids are like that.  But James has done the RC damping network on the output of the transformer to kill leakage inductance induced noise spikes, so even if you replace it with a higher end unit, such as R-core, the gain will be very little.  The problem with R-core would be higher radiated stray magnetic field, so unless you can make it far away, it is going to affect the clock circuit.

The hum buckers used in Berkeley are mechanically noisy (hum).  So you gain some and you lose some.

Also, contrary to popular beliefs, the most damaging noise does not come from AC, it actually comes from inside your gear - the rectifiers (diodes).  Silicon rectifiers or fast recovery rectifiers are "noisy".  They emit tons of high freq noise at the moment of switch off.  The smart designers will put a snubber across each rectifier to damp out this noise.  James did the right thing on using the Cree SiC recitifiers, they have nearly zero switching noise.  And the other rectifiers he used for the display are of soft recovery type.

My assessment of D2 is that power supply related noise is not a problem on this unit.  The problem is actually the clocks on U2 and WM8805.

 
Let me take your points one at a time, you've claimed they are 'not quite valid' so let's have a look to see if your claim is quite valid.
 
Firstly the blue finned transformer does indeed have higher winding capacitance compared to other transformer types of the same VA rating. If you disagree, go measure some - I have made measurements myself of similar trafos. Higher capacitance to mains is always worse than lower capacitance from a noise point of view. Toroids are, in my experience, always the highest interwinding capacitance of the same VA rating and I explained why this is so.
 
Second point - you call that blue fin toroid 'cheap'. That is incorrect, those are amongst the most expensive for the VA. So the designer hasn't chosen them for the price, that's for sure. If you disagree go to Digikey here: http://www.digikey.com/product-detail/en/70041K/TE70041-ND/154886 and verify the price.
 
Third point - leakage inductance. Not a problem - in fact all toroids have the lowest leakage inductance of the VA rating. That's the upside of the high interwinding capacitance. Leakage inductance induced noise? Gimme a break! The noise I'm talking about here is common-mode and cannot be affected significantly by snubbers across a secondary. I agree on one point - R core would have higher leakage flux, so this is an issue. External trafo is probably the solution, but a steel screening can is a possibility too.
 
Fourth point about rectifiiers - once they're snubbered or low noise ones are used, they're not an issue. So a red herring.
 
Your final assessment is not based on full understanding of the issues.
 
<edit) Does your D2 have galvanic isolation between the USB ground and the audio output ground?

 

[Viper2001]

Quote:

 
Let me take your points one at a time, you've claimed they are 'not quite valid' so let's have a look to see if your claim is quite valid.
 
Firstly the blue finned transformer does indeed have higher winding capacitance compared to other transformer types of the same VA rating. If you disagree, go measure some - I have made measurements myself of similar trafos. Higher capacitance to mains is always worse than lower capacitance from a noise point of view. Toroids are, in my experience, always the highest interwinding capacitance of the same VA rating and I explained why this is so.
 
Second point - you call that blue fin toroid 'cheap'. That is incorrect, those are amongst the most expensive for the VA. So the designer hasn't chosen them for the price, that's for sure. If you disagree go to Digikey here: http://www.digikey.com/product-detail/en/70041K/TE70041-ND/154886 and verify the price.
 
Third point - leakage inductance. Not a problem - in fact all toroids have the lowest leakage inductance of the VA rating. That's the upside of the high interwinding capacitance. Leakage inductance induced noise? Gimme a break! The noise I'm talking about here is common-mode and cannot be affected significantly by snubbers across a secondary. I agree on one point - R core would have higher leakage flux, so this is an issue. External trafo is probably the solution, but a steel screening can is a possibility too.
 
Fourth point about rectifiiers - once they're snubbered or low noise ones are used, they're not an issue. So a red herring.
 
Your final assessment is not based on full understanding of the issues.
 
<edit) Does your D2 have galvanic isolation between the USB ground and the audio output ground?

 
Sure, higher winding capacitance is going to let in more high freq noise from AC.  The question is whether it is high enough to cause any problem.  I've done enough mods with custom R-cores to be able to tell you the difference is minimal if the rest of the circuit is done right.
 
When I said the blue fin toroid is cheap, I actually meant cheap in quality, not price.  Sorry about that.
 
For common mode noise, I use a balanced isolation transformer with Faraday shield upstream to my gear.  Then, there is also the differential mode noise you need to deal with in AC.
 
The leakage inductance can interact with the junction capacitance of the rectifier to cause a ringing spike.  James did the RC snubber on the secondary of the toroid.  He saw what I saw when the rectifier switches off.
 
Snubbers or soft recovery rectifiers are only band aid to the problem, they don't completely eliminate the problem.  It is better to start with something with the least noise possible.  And there are better sounding rectifiers whether they are snubbed or not.  The Cree SiC Schottky happens to be the best sound SS rectifier in existence, and I've listened to quite a few, including some ultra soft recovery ones from Vishay and true soft recovery ones from Diotec to be able to tell.  It is similar to tube rolling, some rectifying tubes are better sounding than others.  Same for SS rectifiers.
 
D2 has galvanic isolation b/w USB ground and output ground.  This has been much talked about in previous posts.
 
Ok, enough of this off topic stuff.  Lets go back to the discussion of the D2 DAC.

 

[Sapientiam]

R-core aren't the lowest interwinding capacitance - EI core are lower than R-core.
 
Using a differential/balanced isolation transformer with shield can help, it depends what you connect the shield to. If that's going to a noisy earth it can nullify the effect. Differential mode noise is easy to filter out in a power supply, because its entirely contained within the PSU. CM noise is dependent on what's connected externally. Anyway glad to see you've already taken steps to control CM noise.
 
Do you know how the galvanic isolation is done - with what capacitance across the isolation barrier?
 
None of this stuff is OT, its directly about the thread which is the D2.
 
<edit> Just as a general point, there's no 'complete elimination' of any problem whether CM noise or jitter - there's just reducing it to a level where its not significant. This is what design engineering is all about.

 

[Viper2001]

Differential mode noise depends on noise rejection of the voltage regulators, they are not good for rejecting high freq noise.  But this can be dealt with with an inductor (choke) after the rectifiers - but most designs nowadays do not do this.  It is also not good when you have large varing load with a large choke.  It is most suited with a class A circuit, where load is pretty much constant.
 
We talked about James taking a short cut to this galvanic isolation issue pages ago by using a pulse transformer on the input end of D2, where U2 is driving the input.  Remember, U2 is simply spitting out S/PDIF signal, and the ground is isolated right at the pulse transformer.  James used Scientific Conversion pulse transformer with Faraday shield, and the shield is connected to ground on the DAC side.  The SC PT has high leakage inductance and even higher when shield is connected to ground. The shield is suppose to suppress CM noise, but really, there isn't any CM noise to suppress at that signal path.  It just makes the leakage inductance even higher.  High leakage inductance means high TDR, not something you want.  I've replaced the SC PT with a better PT, one with lower leakage inductance.
 
Yes, there is no complete elimination of any problem.  We just try to make the effect as little as possible.  To me, D2 is already a superb DAC, but there are still some improvements to go, and that would really take it to a world class DAC

 

[Sapientiam]

I agree about chokes - they're a very useful element in power supply design, and yep - one that very few designers use. Would be good to see more use of chokes to cut down HF PSU noise.
 
So is James using the USB power to run his XMOS board (U2)? If not then he'll need a separately isolated supply. I haven't gone back over the complete thread so I appreciate you bringing me up to speed here 

 If using just one SC trafo then he must be going USB-SPDIF and feeding that into D2 right? If so then that rather does negate the advantage of async USB - which is low jitter. I agree that leakage inductance is undesirable, but then so is CM noise so one does indeed have to pick one's poison. I agree there's no point at all in using a screened trafo when there's no clean ground available to connect the screen to. Seeing as I'm currently a jitter skeptic (provisionally though, open to persuasion) I consider the CM noise a greater evil than the jitter so would live with the leakage inductance in order to get the low coupling capacitance.
 
Yep the D2 starts out being a great DAC at the price, where much attention has already been paid to important details, which is why its worth sweating all these small further details to 'gild the lily' so to speak.

 

[Viper2001]

I once believe that we need isolated liner power supplies for everything, but I'm no longer a believer.  You don't need isolated PS for XMOS to sound amazingly good.  That's the case with U2+my mods.  You just need to use LC filters where appropriate, and there are certainly a lot of LC pairs on U2 and D2 (James used ferrite beads, for the L in LC).  But you do need ultra low noise voltage regulators.  This is also where U2 and D2 can improve upon.
 
The advantage of async USB is that it no longer depends on the crummy clock in the PC, but relies on the clock in the receiver circuit.  James did the Pierce oscillator in the XMOS schematic to reclock the S/PDIF signal.  He could use Crystek to achieve better result, but he was counting on WM8805 and ESS9018 to further suppress jitter.  Yeah, they work, but won't bring you world class performance.  For world class performance, an ultra low jitter clock is still needed.  So even if James were to use I2S between U2 and D2, an ultra low jitter clock is still needed to clock the I2S signals.  I think it is far easier to reclock the S/PDIF signal, than to deal with I2S.  With I2S, you faced with the issue of galvanic isolation.  And doing that won't be easy (forget about opto-isolators, they don't sound good).  Also, current USB opt-isolators are limited to USB Class 1.0 Audio only.
 
I think we have dissected D2 deep in the hardware level more than any other DAC on the market.  I don't think it is fair, as anything negative here could imply lost sales for Anedio.  And the negatives here aren't really all that bad consider the price point, and Anedio never claims their DAC to be a world beater.  FWIW, my friend just listened to the M2Tech Vaughan, he said it is in no comparison to the D2.  And Vaughan is EUR 6K!

 

[Sapientiam]

I tend to agree that total isolation isn't always a requirement - given that at HF, even a fairly small capacitance looks like a low impedance. So yes, judicious use of ferrites can be a help. Use them for their resistance (loss) though, not their inductance (which gives out above a certain frequency anyway). Hard to get enough isolation though just relying on ferrites - common-mode chokes can assist as they do go to higher impedances than ferrites alone.
 
I can't see anything unfair in dicussing the detailed decisions made in terms of the D2's design - it shows that a fair bit more attention than the average has been paid, so all credit to James for this design. Especially as he's not charging a greedy price - its worth bearing in mind the price point here is a very fair one. As for the Vaughan I'm already on record elsewhere for trashing that design so no need to say more here.....

 

[cky8]

Quote:

 
Have you noticed the AC voltage setting on my D2?  I'm not in your part of the world.  I don't know how many US folks are willing to ship their units to me for modding...

 

Well I am at 230V so that is not a problem for me if you are willing to offer your modding services. But seeing that you have stacked another PCB on top I am not sure if that would still fit in the U2 housing? I have a external U2.

 

[Viper2001]

Actually, James added a 2 ohm resistor before many LC pairs, like you said, the small resistance is needed for the filter to be effective.  But then again, the inductance of ferrite bead is too small to be useful.  I'd rather use non-magnetic inductors.  It has built-in resistance of 2-3 ohms, and that would be better than R+ferrite bead+C.
 
I'm just going thru' some onboard voltage regulators James used, and started to replace them with lower noise version.  So far I did one on the U2's output circuit, and I got a warmer and more velvet vocal.  So I'm going to work on the regulators on the D2 mainboard.
 
Perhaps all these should go into a separate thread on modding the Anedio D2

 

[Viper2001]

Quote:

Well I am at 230V so that is not a problem for me if you are willing to offer your modding services. But seeing that you have stacked another PCB on top I am not sure if that would still fit in the U2 housing? I have a external U2.

 
Well, U2 does not run on 230V, so any part of the world would be fine with me

 
So far one person has contacted me about modding his D2.  But I want to finalize everything before I start offering this service.
 
I don't know if I can make my clocks fit inside U2.  Anyone has the casing pictures of U2?  I need to see how much internal space there is b/w the circuit board and the top cover.

 

[cky8]

I can make some pictures for you 

 
EDIT:
 

 
 

 
 

 

[customNuts]

@Project86
 
did you get your Auraliti pk90 +psu?? If so can you fill us in how it turned out, specifically against the SBT +EDO??