Soekris' DAC implementations

[OlavHaal]

These are not so bad....

Hi Gentlemen....

From this company I have bought my transformator for my SITpreamp: Luminaria (the Pass forum). Very quick delivery and nice toroids! The smallest is 15watt(2x7.5V w 2x115volt Primay). So if this is an option - it should be a perfect solution.
I myself planned to buy and implement this R2R-dac, but faith has thrown me into a sickness and hospitality-cyclus. Maybe at a later time....

Best to all

Ol

 

[OlavHaal]

Forgot...

SORRY....

Here is the link: Toroidal transformers for 115V and 230V applications

Ol

 

[spzzzzkt]

I think Søren (soekris) mentioned that his FIFO scheme TRACKS the incoming clock to prevent FIFO overflow. http://www.diyaudio.com/forums/vend...-magnitude-24-bit-384-khz-57.html#post4135993 As per my humble understanding this is not the ultimate scheme in fighting jitter; Mark Levinson (name came because I used to have 360S) and many others have done that in 90'.

It seems Ian's (iancanada) approach is more comprehensive.

I'm not really sure where you think I wrote "ultimate" as a descriptor of Søren's FIFO?

I don't count myself as a jitter reduction guru by any stretch. I did build and code a PIC based PLL using a Linear Technology unbuffered DAC to control a Tent Labs VCXO around 2008/9 but never got the code fully sorted out. It used the 3/8 word buffer in the SM5847 filter chip to compensate for slight drift between the recovered clock and vcxo clock. It was fine once locked and sound far better than the recovered clock despite being a "rough as guts" perf board experiment.

Iancanada's method has drawbacks - the long buffer creates an input to output delay which makes it unsuitable for home theatre applications as Søren pointed out.

The reference to "TRACKS the incoming clock" belies what is actually occuring - while long term frequency of the data stream is followed, the FIFO acts as a low pass filter on the incoming clock to remove jitter. The corner frequency of the low pass filter is directly related to the frequency of the updates to the master clock.

I found in my PIC PLL setup there would be no adjustment made to the clock frequency for periods of 5-10 seconds even with a minuscule buffer and an unoptimised control loop configuration. This is the equivalent of a LPF with a corner frequency of 0.1 Hz.

:ADD: Most XMOS and Amanero based USB to SPDIF/I2S convertors now use XO to ensure that clock rate is stable, but even in the case of older transports the clock drift is not huge enough to require large adjustments over the short term. :END ADD:

With a well sorted FIFO and accurate adjustment of the master clock the update rate is likely to be in the order of once every 60-120 seconds or better, giving an equivalent of a loop filter with a corner frequency of 0.01Hz in the worst case. As a point of comparison the old CS8414 S/PDIF reciever has a loop filter corner frequency of 1000Hz. The low update rate means that the influence of jitter on the incoming clock is effectively eliminated. The infrequent update rate of the master clock means also that jitter induced by adjustments to the clock is very low.

So I contend that while a large buffer might look impressive on paper, in practice the advantage isn't nearly as significant as you might have been lead to believe.

And I apologise to Eldam for the digression into the technical.

cheers
Paul

 

[Julf]

Pretty good performance for a protocol that is supposedly far worse than coax. Perhaps the V-Link has a superior TOSLINK implementation to your Squeezebox, but regardless it shows that TOSLINK is not inherently flawed.

Anyway, I suspect you'll find that the dam1021 will have little sensitivity to the different cable formats.

Do you think mere facts will trump real testimonials?

 

[Eldam]

If you have a Squeezebox Touch you should definitely install the free EDO plugin and use the USB port as a digital output. Then get the excellent Amanero board and feed the Soekris dac with the I2S signal. I guarantee you won't look back.
The Amanero Combo board doesn't need drivers with Linux machines such as Squeezebox, and it's cheap too given it's marvellous performance.
It's a no brainer unless you use the USB port for music data. In that case I suggest you use the ethernet or wifi interface in stead.

Hi,

Unfornatuly I have the "bad" SB Duet.

 

[TNT]

It is the same mechanism used in Ian's reclocker. By tracking it is meant that the input sample rate is tracked and closely matched by the local oscillator allowing the use of a shorter FIFO.

No its not. Ians fifo doesn't track the incoming freq - it uses a large fifo to absorb any difference in clock speed. That the whole point with a large fifo - no tracking. If one only have a small fifo, then one have to track the incoming signal. Ians fifo has around half to a second delay.

Now, I hope that the R2-R dac docent adjust continuously, but rather does stepwise changes or has a smart also to gradually lock into the source. Floating around would be like a slow (? How slow?) PLL..

ok zzzzz has a grip on it!

//

 

[TNT]

Hi,

Unfornatuly I have the "bad" SB Duet.

The Duet has opto out, doesnt it?

The Duet is bit correct on its opto output so it is OK.

And as a matter of fact, the Duet sound really good. Especially of you change out the output capacitors. I have heard a slightly modded Duet play wounderful over a pair of MBL so its not inherenty bad - it actually played music in a way many DACs would love to do. This is of cource using the built in DAC which you wont use in this case. It will be a mere bit transporter and its own intrinsic jitter will not matter because it will dissapera in R2D2

I would try it with a Toslink.

//

 

[Eldam]

Hi,

So for the lover of Folklores (people who love old spidf/AES because it's confortable despite the modern USB 380 Khz) :

There will be two AES/Spdif inputs:

1: Balanced into LVDS Receiver, can be connected directly to transformer and can be run single ended for SPDIF Coax, just a capacitor and two resistors needed when single ended 75R. To keep it isolated I recommend to always use a transformer for AES Balanced and SPDIF Single Ended inputs.

2: 3.3V CMOS level input, can be connected directly to SPDIF Optical Toslink receiver.

There will be example schematics in the manual.

Selection between I2S and AES/SPDIF sources can be automatic or manual with two pins that can be connected directly to a control switch. For more sources you can also just switch the inputs.

Can I link those transformer on the first page ? : SE spidf pulse transformer : PE-65612NL Pulse Electronics Corporation | 553-1580-5-ND | DigiKey
or a Newava S22083 Newava Technology Inc | 470-1003-ND | DigiKey

Do you think mere facts will trump real testimonials?

: Hey JULF, I'm not totaly bad when saying "sometimes/often" Tolinsk can sound worst ! Here you mix up yourself testimonial and prove... ! Here we have just the testimonial with this measurement which prove than tolinsk can be good (without problem), it doesn't prove than most of the time or sometimes (did I say the opposite ?, no !) tolinsk can be better or simply as good in situation when listening your hifi. I just said most of the times because of implementation and layout most of the time it's worst ! you talk about the concept I'm talking about expériences (testing). So without knowing why, I saw at home and some (most? long time ago?) listeners testimonie than spidf gave for us better results. Was is only jitter, I don't know. But if the expériences at the end don't validate the concept at the end ... we have a fact. Which is important is the experience, so this measurement prove it's good, not that it's always good in real life! . Yes the measure are real but which is real alo is that bad technicians sell us bad layout implémentations also !
You put few vodka in your Pepper, put just a little Pepper in the vodka ! Btw your photograph let me think to an old post war american song I have pleasure those days to listen: "Rhum & coca cola, lal lalla lalala !" Folklore... we are comparing bagpipes (which is not a bag where you put the rest of your pipes - private joke with TNT )

So ok, I hope people will share their respective AES & SPIDF shematic to help people like me here !

Thanks Olavhaall : I putt in in first page.

 

[TNT]

Eldam - most, if not all old testimonials about opto link sound is for DA which has no jitter buffers but more typically, rather poor implementations of PLLs. This DAs more or less played out all the jitter that was sent on the line which constitutes of the clock jitter of the source and the added jitter of the link.

So, it sounded bad and better with galvanic s/pdif due to the poor bandwidth of the opto links. This was 10-15 years ago.

Things change, Opto RX/TX units became better. There is glass cables. There is solutions indifferent to jitter.

Have you followed and understood the technical aspects and the development of technology?


I bet you watch analog tv and use a rotary dialing phone...

//

 

[Eldam]

The Duet has opto out, doesnt it?

The Duet is bit correct on its opto output so it is OK.

And as a matter of fact, the Duet sound really good. Especially of you change out the output capacitors. I have heard a slightly modded Duet play wounderful over a pair of MBL so its not inherenty bad - it actually played music in a way many DACs would love to do. This is of cource using the built in DAC which you wont use in this case. It will be a mere bit transporter and its own intrinsic jitter will not matter because it will dissapera in R2D2

I would try it with a Toslink.

//

Hi TNT, thanks.

Yes it has

I tested both on DACs and the experience gave me better result with spidf on the DACs I used to have.
I surmise an isolation traffo like Soren advise will cost me few if the result are the same as predicted ?
around 5 euros or less for one of the two references I linked (but I don't know if the ref I linked are good for that ?).
May you share your shematic about the aes receiver for the ones who want to try or need it ?

cheers

PS : have you a Duet ? I can advise you a particular cap on a particular place which transform it for a superb result on the Subbu V3. even if here we should not need it with the fifo input !

 

[Julf]

I'm not totaly bad when saying "sometimes/often" Tolinsk can sound worst!

No, but still... I know it is a bit petty to make remarks about spelling, but I think we would probably all prefer it if you spelled it "toslink" (as in "TOShiba LINK") instead of a place name in Siberia...

Here we have just the testimonial with this measurement which prove than tolinsk can be good

So without knowing why, I saw at home and some (most? long time ago?) listeners testimonie than spidf gave for us better results. Was is only jitter, I don't know. But if the expériences at the end don't validate the concept at the end ... we have a fact.

The problem is that testimonials are anecdotes, not proof or fact.

 

[TNT]

Chapt 5 Application:

http://media.digikey.com/pdf/Data Sheets/Toshiba PDFs/TORX147(F,T).pdf

//

 

[Eldam]

.

Have you followed and understood the technical aspects and the development of technology?


I bet you watch analog tv and use a rotary dialing phone...

//

Whazat.... are you crazy no phone nore TV ! just computers and hifi, and some stuffs for cooking, that's all !

As ancient IT manager I just used fiber glass for transport datas on a long distance between two places while in local coper and some protocol go as fast than fiber in short distance for disks bays and virtualisation e.g. !).

All this thread will go into Lounge aera....

 

[Eldam]

No, but still... I know it is a bit petty to make remarks about spelling, but I think we would probably all prefer it if you spelled it "toslink" (as in "TOShiba LINK") instead of a place name in Siberia...

The problem is that testimonials are anecdotes, not proof or fact.

You are a little bit sophist You don't like Siberia and dyslexia... comon JULF, not you please ... and don't speak for others ! And "Not" is not "Still"... a bit petty but really more interresting than spelling .

Have you prove than most of the time in real situation (commercial products) aes is better with all the parts it needs than spidf ? No ! You extend one good result to all the devices : this is not scientific attitude. And myself I accepted this measurement as a fact aes can be as good ! So what ?

You speak only technic while I'm only speaking about expériences. Expériences (not speaking about scientific validation, please don't put some words in my mouth I never said ) are always anecdotes. And I've already said your remark was good above; What do you want more ? Saying than aes is always as good as spidf. Yes with such devices like Soekris 's one. But let take 10 DAC on shelf. Maybe the experience will show than most of the time the spidf output on a classic DAC will give better results! People testimonied that. As TNT remarks, maybe it's different today ? Maybe yesy, maybe not for curent products. So without proves or many testimonials of the opposite....
Steril conversation, yes folklore, there are better protocols todays no limited at 96 hz ! for ending : prove and fact are two different words even used in scientific situation. While being certainly a good technician (I'm not technician) you certainly could re open epystemology books (does it take two y in English... sorry no English native, nore cartoons TV in English in our country to allow celtic/latins to have a fluent English when youngs ! )

So anyone to say if the two traffos linked are good enough to follow Soren advice ? Or do I stop here ?

 

[Julf]

You don't like Siberia and dyslexia...

I definitely don't like Siberia. I have much more understanding for dyslexia, now that I know that is the issue - it was just that your misspelling was so systematic and consistent.

Have you prove than most of the time in real situation (commercial products) aes is better with all the parts it needs than spidf ? No !

I think it has been proven quite a number of times that both are perfectly fine for their intended purpose.

Maybe the experience will show than most of the time the spidf output on a classic DAC will give better results!

Maybe, yes, if you do enough controlled tests, so that the results are statistically relevant.

there are better protocols todays no limited at 96 hz !

Indeed, but 96 kHz is not really such a bad limitation considering our ears, even under the best of conditions, only go to 21 kHz or so. Anything beyond 96K is overkill, done just because bigger numbers must be better.

 

[TNT]

"Maybe the experience will show than most of the time the spidf output on a classic DAC will give better results!
Maybe, yes, if you do enough controlled tests, so that the results are statistically relevant."



One can't compare total system behavior and think it correlates to one aspect of the system.

//

 

[Eldam]

He can compare if the whole system is the same and if he just swap a toslink link for a spidf one...

It's true in the other sense to !

 

[Julf]

One can't compare total system behavior and think it correlates to one aspect of the system.

Unless that is the only variable you change in a series of tests.

 

[Eldam]

No testimonials or measurement proof on the two traffos I linked above to follow Soren's advice ? Does it slide to an old men conversation about who have the biggest... last word!

Those traffo links help no one ?

 

[Julf]

No testimonials or measurement proof on the two traffos I linked above to follow Soren's advice ?

So what do you want to know about those transformers?