How immune to jitter are modern dacs?

[The Computer Audiophile]

Hi Johan - Here is a link to a discussion about buffers etc... http://www.computeraudiophile.com/content/buffer-universal-panacea

 

 

 

[pj]

@Gordon

 

There was some a discussion over at diyaudio in 2004 on ASRC instigated by a former VP of Engineering at Crystal (manufacturers of ASRC chips). The poster mentioned that ASRC output a clean clock but embedded the jitter in the data, and that in his opinion there was little difference in the output between jittered data & clean clock from ASRC vs clean data & jittered clock from a PLL if the incoming jitter level was the same.

 

http://www.diyaudio.com/forums/digital-source/28814-asynchronous-sample-rate-conversion.html

 

cheers

Paul

 

 

[Wavelength]

Thomas,

 

It doesn't "in theory," and it doesn't "in practice."

 

I would have to agree with you that there is really no way to get to zero jitter. While the asynchronous process no matter what the interface is, does significantly reduce the overall jitter. There is always some intrinsic jitter in the Master Clock Oscillator and the DAC chip which will be present in the output.

 

Paul,

 

I have some stories about the Crystal/Cirrus ARSC... On the CS8420 when you played a Linda Ronstandt first cd the part would litterally go into a notch filter. They fixed the part by putting a hard reset anytime an internal bit was set true that meant the state machine was no longer working as it should be. Yea that's a good fix!!!!!

 

But what I am saying is the problem occurs at the entry to the ARSC or between what ever the receiver chips is (SPDIF, USB, Firewire whatever) and the ARSC. This guy is only talking about what happens in and after which is not totally correct either. Oh sure the math is there... but the performance is not as he suggests.

 

Thanks

Gordon

 

[Thomas J]

Gordon, I appreciate that async does reduce jitter significantly, and also appreciate your descriptive honesty, which is a breath of fresh air in audio discussions. It seems to me an attempt to understand what happens in a piece of audio gear must be guided by descriptive accuracy. Phraseology like "zero jitter" doesn't meet that standard.

 

[pj]

Gordon,

 

Yes, as you said above the master clock, psu, board layout, logic gates all have an impact - or are you thinking of something different? And I definitely agree that any manufacturer claiming zero jitter or a derivation of that claim really needs to justify that claim as something more than spin.

 

As an footnote I'm currently running a DAC ( Spencers D1V3 which is a loose clone of the Pass D1 with jfet i/v and output stage) and have built a PIC controlled PLL/VCXO setup using code based on Jos van Eijndhoven's description (see http://jos.vaneijndhoven.net/dac2/index.html) and the discussion you instigated on the topic at diyhifi a few years ago.

 

cheers

Paul

 

[Wavelength]

Thomas,

 

High End Marking at it's best is coming up with terms that apply to something that cannot be true. As I write this I am looking at an ad on Chris's site that makes me want to laugh.

 

Paul,

 

That is actually a very good way to pull of low jitter SPDIF. I only wish the Crystal/Cirrus parts had much longer FIFO's on them as it is really too short to really make something like that work really well. I have not looked at the Wolfson, AKM or TI's offering but a longer FIFO would mean that you could hold the VCXO at a single voltage for a much longer period of time. Do you monitor the OSLIP bit to see if you are slipping samples?

 

Really you maybe better off running the receiver in master mode going to a dual FIFO (DATA and WCLK) and then putting the VCXO on the output of the FIFO. This way with the PIC you can setup the FIFO so when it is half full that it will start outputing to the dac and thereby having allot better window in which to change the VCXO.

 

If all the rumbling on DIY is correct the new NAIM stuff looks interesting. They have a set crystal with a couple of relays to inc/dec or keep stable the Master Clock going into one of the AD DSP's which has of course tons of memory and they control their jitter that way and retain flow control by adjusting up/down or steady the Master Clock.

 

Thanks

Gordon

 

[pj]

Gordon,

 

I'm firmly in jurrasic tech mode - cs8412 and sm5842 with only concession to modernity being the pic/tent vcxo. I'm using FSYNC output of the 8412 rather than MCK, and the SM5842 is running in jitter free which gives a buffer of +/- 3/8 of LCRI which allows the PLL to drift slightly against the incoming clock. A FIFO would definitely be a better approach but the SM5842 is considerably better than relying on the buffering in the RX.

 

The NAIM does sound very interesting. Thanks for the heads up.

 

cheers

Paul

 

[Wavelength]

Paul,

 

The 8412 is my fav... I have tons of them here. But if you have it in slave mode isn't the FSYNC/WCLK actually an input?

 

The way I did it was that I used a Valprey Fisher VCXO at 2.8224 this is a killer unit with really low jitter and had the CS8412 in slave mode with the MCK/4 (74HC74) into PIC T0CKI and the VCXO into the CLKIN of the 12F509. I did not use a DAC as much as a 1024 variable resistor and then this into the base of a 2N5089A via a 4K99/100uF cap to lower the noise even more. This worked really well for 44.1K... but then was the last time I worked on SPDIF.

 

Even that with all the work that went into it had jitter just over 100ps.

 

I was never a real big fan of any of these digital filters.

 

Thanks

Gordon

 

[pj]

Gordon,

 

I'm using the CS8412 in master mode. The 8412 FSYNC goes straight to T0CKI on a 16F818. VCXO drives the DF and CKI/T1CKI on the PIC via a buffer. The PIC controls the VCXO via an 16bit LTC2641 serial dac. The DAC output is unbuffered and I've got a small film cap on the output forming a LPF. I'm not quite sure what jitter levels are like, but the sound is much more focused with the pll/vcxo than without.

 

I did have the whole shebang connected to a stopped clock reclocker based on Guido Tent's DIY Audio DAC which allowed me to clock each of the PCM63's from the VCXO buffer. This was all piggybacked onto Spencers boards and was a bit of a mess so I've removed till I can do something neater. There was a definite improvement in SQ clocking the DAC chips as directly as possible from the VCXO.

 

I'd gather you'd abandoned SPDIF to concentrate on Async USB. It does appear to be a far more elegant solution to the transmission problem.

 

cheers

Paul

 

 

[Wavelength]

Paul,

 

Feel free to email me about this, not everyone wants to hear about SPDIF jitter reduction.

 

A couple of things... What I was saying above is very critical to your design. If you don't reclock the data and word clock into the filter then it will result in jitter related data errors as the VCXO and the data/wclk will no longer be aligned. Actually a better way if you don't use slave is with a fifo. TI makes some dual 64K fifo's that would work great for something like this or even some of the small 16 bit units would work great. Allow them to 1/2 fill then start the output.

 

Yes async USB trumped anything I did with SPDIF so I figured why go backwards.

 

Thanks

Gordon

 

[pj]

Hi Gordon,

 

I won't labor the point but the 5842 has a split clock domain - data is clocked into the DF using BCKI/MCK so there is no misalignment between data and clock. The clean clock drives clocking data out of the internal input buffer, processing and clocking data out to the DAC chips.

 

I agree this is all getting a bit oblique to the topic.

 

cheers

Paul

 

[jostp]

On Christmas eve, we spend a nice time with friends at the table, with music in the background, adapted to peoples wishes from my ipod remote .

Later, late in the night, we had some dedicated listening with a couple of music lovers.

What a disappointment the rendition via the mac mini, compared to the cec transport!

 

What should I do?

- choose another DAC then the Bel Canto

- throw the apple mini away

- upgrade the mac mini (sound card, replace opto-couplers, ...????)

- wait a couple of years till computer playback via dac is is on par with top notch transports? (but the transports only plays '44,1/16')

 

 

[Audio_ELF]

1 or 3

 

You should be able to get comprable sound quality (if not better) out of a Mac Mini into a DAC as via a transport into the same DAC.

 

Depending on your budgets you could consider FireWire or Async USB DACs such as Weiss DAC II and Ayre QB9. Or consider looking at inproving the SPDIF interface using a FireWire or USB device. A true Mini-TOSLink to TOSLink cable rather than using an adpator with your exisiting cable may also offer an improvement.

 

You should also check that the settings in iTunes and OS X Audio Midi control pannel are correct -- set the correct sample rate before starting iTunes; make sure all sound enhancements in iTunes are disabled and set volume to 100%. If you still feel the SQ is lacking - consider Amarra (or wait for Pure Vinyl 3 or Audiophile Engineering Twilight) or possibly Play as an alternative to iTunes.

 

Eloise

 

[The Computer Audiophile]

"- wait a couple of years till computer playback via dac is is on par with top notch transports? (but the transports only plays '44,1/16')"

 

Comparing one transports to one computer and making the above question/statement is doing yourself a disservice. In terms of accurate sound I'm willing to put computer based audio up against any transport at any price. If you are talking about what you like soundwise then I can't argue with hour conclusion.

 

[pj]

Johan,

 

The Mini's toslink output isn't particularly good in terms of jitter performance, and as Gordon and others have observed the "jitter elimination" attributed to many DACs should be read as marketing speak for what is effectively "Jitter Reduction". The outcome is the lower levels of jitter you can feed into the DAC better job it will do of output low jitter levels (limited by the DAC's intrinsic jitter levels).

 

If you are happy with the overall sound of the Bel Canto feed by the CEC I'd suggest rather than choosing another DAC, or throwing the Mini away you really need to look at how to get the cleanest output digital from the Mini. Ultimately computers are not designed as music transports and the interface to the dac is the area that needs the most attention regardless of the computer you use.

 

I've had a bit of a look at the Mini output board and it looks tricky to work on. The digital -> optical conversion is done in the mini-jack unit so it would be feasible to replace the mini-toslink with a coax output. I haven't seen any "previous art" so this would require some research and work. Ultimately this mod will be limited by the necessity of using the relatively dirty on-board clock.

 

This leaves the practical approach of a firewire or usb to spdif converter of some description. My experience is that a reasonably basic interface like a Echo Audiofire matched the performance of a Teac VRDS T1 when feeding the same dac. I've recently replaced the Audiofire with a Konnekt 24D and this is clear improvement over the T1.

 

It might be worth seeking out a Hiface to try as that seems to be very promising.

 

 

cheers

Paul

 

 

 

 

 

 

 

 

 

 

 

[jostp]

Paul,

This is the kind of information I was after.

I 'll wait some time till the problem with the 'tics' gets resolved, and then I may try out the Hiface.

(Although I just don't understand the 'trick', it has a good clock, but then, the bel canto also has a good clock? Apparently the implementation is better then the bel canto).

cheers,

Johan