[middachten]

Quote:

Yeah I also don't understand why usb DAC doesn't just have a buffer in front of it, even just a tiny one, then no need to worry about jitter at all.

 
USB interface chips do have some form of buffer. Otherwise the interface would never be stable (its an asynchronous interface by design). Its the jitter on the I2S outputs that causes the problems. There are even implementations with no Xtal clock whatsoever. Re-clocking at the DF or DAC will improve this, but if the jitter is more than 1/4 of a clock cycle its impossible to re-structure the original clock structure. Its even possible to loose a whole sample this way.
This is where not only time/frequency domain problems kick in, but also the bit-accuracy is lost.

 

[Dynobot]

USB Buffer....I wonder if the AudioGD DI has one...as shown in the pic below as "USB external RAM"
 

 

[nick_charles]

Don't all DACs have  a 16 bit or 24 bit (word ) input buffer controlled by the clock so that the DAC processes the numerical value of a whole word at a time not just bit by bit ( for multi bit DACs) , in which case slight variations in the arrival order of bits at the receiver would be moderated by the clocking of the word to the dac ?? i.e as long as the input buffer is flled and the clock sends  words to the DAC at the correct intervals any jitter beforehand is massively attenuated ?

 

[middachten]

Yes, this is true. But still a lot of jitter related things can go wrong:

1. The clock that dictates the 'word' is derived from the original SPDIF signal or by the USB interface chip. Both are usually not the most stable clock sources > jitter
2. sometimes the edges of the (word) clock signal are so 'vague' that the actual timing of the clock gets blurred at the receiving end (the DAC) > jitter

 
If you look at the images of the Audio-gd DI you will notice the emphasis on high quality capacitors, line driver and output transformer. This is all needed to preserve a high quality SPDIF signal. This will improve the quality of any DAC at the receiving end which has a poor clock recovery / re-clocking scheme. Because the edges of the SPDIF signal are more stable.
 
If you have ever listened to a pro-audio interface using ADAT (an even more jitter-prone interface than SPDIF) with or without a master Wordclock connected you know all about jitter artifacts. The same is true if you have switched between 'internal clock source'  and SPDIF or ADAT clock source on a high quality pro-audio interface (like RME or Apogee). 

 

[regal]

Quote:

 
 
If you have ever listened to a pro-audio interface using ADAT (an even more jitter-prone interface than SPDIF) with or without a master Wordclock connected you know all about jitter artifacts. The same is true if you have switched between 'internal clock source'  and SPDIF or ADAT clock source on a high quality pro-audio interface (like RME or Apogee). 

This post points to the only real way to get rid of jitter,  put the clock in the DAC and slave the transport to it.   However popular DAC manufacturers have been touting their jitter voodo elimination technology for years,  some have even been proven to just be fake idle circuitry.   I just don't depend on anything in the DAC to reduce jitter,  it is like buying snake oil.  No company has really proven it has been done (hell it would take a MIT PhD thesis),  its just not empirical enough for my money.
 
Best thing you can do is listen and trust your ears,  something many people on the internet are opposed to.  If you don't have good ears for music (like say SlimA) then why worry about it ?
 

 

[middachten]

Quote:

 put the clock in the DAC and slave the transport to it.  

Even better: put the clock at the output of the DAC to re-clock the entire PWM or Delta/Sigma stream in the analog domain, after having fed it trough an opto-coupler for perfect galvanic isolation from the digital part of the circuit......
 
In other words: as Crystal did it with their CDB4303 evaluation board based on the CS4303.
I still have this board running once in a while. And when it comes to detail and soundstage its still serious contender for todays high-end systems.
Sadly I never came to actually build it into a nice case, and use it in my main system. I think I'll get it dusted off this weekend to see if it still works. Even if it only where to get sentimental about the days that I still worked in audio design......

 

[autonomous]

After reading the five pages so far on jitter in SPDIF and USB cables, I have concluded that it is the discussion of jitter that causes jitter! 

 

[Currawong]

This is very likely why the Sabre DACs have such a good reputation, as they bypass the need for all this circuitry by having a direct S/PDIF input.

 

[Dynobot]

So would Spdif be the way to go with Sabre Dacs as supposed to USB?
 

 

[Currawong]

I would guess it would depend on the USB implementation in the DAC.

 

[middachten]

Quote:

I would guess it would depend on the USB implementation in the DAC.

Exactly!
USB just got a bad reputation due to lazy engineers that used mediocre USB interface chips, and just relied on the glorious technical specs of the DAC chips to do the marketing for them.
So, I've said it

 

[Dynobot]

Quote:

Exactly!
USB just got a bad reputation due to lazy engineers that used mediocre USB interface chips, and just relied on the glorious technical specs of the DAC chips to do the marketing for them.
So, I've said it

 
So what are some good USB chips??
 
In an Asynch situation does the 'quality' of the USB chip matter as long as it can function in the Asynch capacity?
 
Also what are the inherent flaws in USB chips that bring less than stellar performance??
 

 

[middachten]

Both the TUSB3200A and TAS1020B from Ti can provide an excellent USB to I2S interface. But both chips need some form of control logic/configuration software to function properly. The TE7022L provides a good alternative with easier application.
If you're prepared to really dive into the hardcore coding side of things the XMOS devices and reference designs for USB interfaces provide an excellent starting point.
 
But as said before, it al depends on your implementation. A lousy clock topology at your DAC side will ruin even the most brilliant USB interface. And the other way around, a sloppy USB interface with a good re-clocking scheme (asynchronous buffer(!) and excellent power supplies could even provide excellent audio quality when connected to a TDA1541 of 25 years old.
 

 

[leeperry]

 

TAS1020B from Ti can provide an excellent USB to I2S interface. But both chips need some form of control logic/configuration software to function properly.

Most companies that make proprietary firmwares for TAS1020B will make you pay through the nose for them.

 

[Happy Camper]

If the I2S converter output were in the design of the DAC chip, you wouldn't have to worry about as much interference?