8xNOSDAC

Ulas · 11th August 2006, 05:07 PM

This diagram shows another way to get 8x linear interpolation with NOS DACs. All timing is derived from a synchronous counter clocked either by a local oscillator or S/PDIF. The oscillator also exports a clock to slave the CDP.

CLK0-3 are quadrature clocks. WS0-7 are 8-phase strobes. Everything else should be obvious. The circuit can be implemented with as few as 14 logic ICs (74xx type), 8 stereo DACs, a S/PDIF receiver, and transceivers to slave the CDP.

Use the local oscillator in conjunction with a slaved CDP. Use the S/PDIF clock source with non-slaved CDPs. Or, use the local oscillator with a non-slaved CDP if you prefer the sound of missing/duplicated samples, as do users of the DDDAC. If you like the sound of jitter, add a USB receiver but be sure to aschronously reclock BCLK for maximum the effect. The choice is yours.

poobah · 11th August 2006, 05:58 PM

linear interpolation? again?

Spartacus · 11th August 2006, 06:22 PM

Quote:

Originally posted by Ulas
Or, use the local oscillator with a non-slaved CDP if you prefer the sound of missing/duplicated samples, as do users of the DDDAC.

Can you tell me what a dropped/repeated sample every two seconds sounds like?

Ulas · 11th August 2006, 06:43 PM

Quote:

Originally posted by Spartacus
Can you tell me what a dropped/repeated sample every two seconds sounds like?

How about 7 per second? If you can't hear it, you have nothing to worry about.

dddac · 12th August 2006, 07:42 PM

Quote:

Originally posted by Ulas

How about 7 per second? If you can't hear it, you have nothing to worry about.

Ulas is not listening to music, he spends his time reading datasheets, so no need to ask him.

DDDAC users collectively think this dac sounds great (read my site, don't believe me, just read what others say ....) So if reproduced music sound great, even when you loos or duplicate a few samples, who cares? Ulas does, but he is not listening to music, so his opinion is only of technical value. He is mostly right, but missing the point what makes really a difference in musical experience.

If every one prefer blond, Ulas will tell you, for good reasons I guess, that brown is better for practical reasons, but who wants that ?

doede

maxlorenz · 12th August 2006, 09:15 PM

Quote:

If every one prefer blond, Ulas will tell you, for good reasons I guess, that brown is better for practical reasons, but who wants that ?

Hey! ...I too prefer brown!

Quote:

DDDAC users collectively think this dac sounds great

True indeed!

Cheers
M

Ulas · 12th August 2006, 10:38 PM

Quote:

Originally posted by dddac
DDDAC users collectively think this dac sounds great (read my site, don't believe me, just read what others say ....)

And that's my point. It is the unique characteristics of the DDDAC, the dropped or duplicated samples, 5us settling time, nonlinearity, and crosstalk that is appealing to some people.

poobah · 12th August 2006, 10:47 PM

Ulas,

When you say "non-slaved CDP", do you mean that the CDP is "free-running"? That reclocking has gone to point of dropping (or plugging) samples to accomodate 2 different clocks?

What am I missing here?

rfbrw · 12th August 2006, 11:11 PM

Common practice is to run the CS8412 in master mode. That way all clocks are synchronous with the clock derived from the SPDIF datastream.
The DDDAC runs the CS8412 in slave mode. In that mode one is required to generate FSYNC/WS/LRCLK and BCLK/SCLK. These are then used to clock data out of the CS8412. In the pro arena these clocks would be derived from a VCXO locked to a common reference as would the CDP so there would synchronicity across the board. Alternatively one can derive the clocks from an XO and in turn run the CDP off a feed from the same XO. Again there is synchronicity across the board.
Now DDDAC opts for neither of the above options, sticking solely with the XO. Needless to say any frequency other than that for which the XO was chosen is out. Even when the frequencies match one has to hope that the phase discrepancy between the incoming SPDIF datastream and the onboard XO are not enough to overwhelm any buffering the CS8412 may provide otherwise there will be dropped or repeated samples.

poobah · 12th August 2006, 11:24 PM

That is nuts... if I understand you correctly.

It would seem to me that dropping or repeating samples would be the most horrible form of jitter.

This is nuttier than the linear oversampling thread.

11th August 2006, 05:07 PM	#1
Ulas Banned Join Date: Mar 2003 Location: .	8xNOSDAC This diagram shows another way to get 8x linear interpolation with NOS DACs. All timing is derived from a synchronous counter clocked either by a local oscillator or S/PDIF. The oscillator also exports a clock to slave the CDP. CLK0-3 are quadrature clocks. WS0-7 are 8-phase strobes. Everything else should be obvious. The circuit can be implemented with as few as 14 logic ICs (74xx type), 8 stereo DACs, a S/PDIF receiver, and transceivers to slave the CDP. Use the local oscillator in conjunction with a slaved CDP. Use the S/PDIF clock source with non-slaved CDPs. Or, use the local oscillator with a non-slaved CDP if you prefer the sound of missing/duplicated samples, as do users of the DDDAC. If you like the sound of jitter, add a USB receiver but be sure to aschronously reclock BCLK for maximum the effect. The choice is yours.

Thread Tools	Search this Thread
Show Printable Version Email this Page	Search this Thread: Advanced Search

New To Site?	Need Help?
Register to Participate Search Privacy Statement Contact Us	Frequently Asked Questions Did you forget your password? Mark Forums Read


	Please consider donating to help us continue to serve you. Ads on/off / Custom Title / More PMs / More album space / Advanced printing & mass image saving

11th August 2006, 05:58 PM	#2
poobah diyAudio Member Join Date: Nov 2005	linear interpolation? again?

12th August 2006, 10:47 PM	#8
poobah diyAudio Member Join Date: Nov 2005	Ulas, When you say "non-slaved CDP", do you mean that the CDP is "free-running"? That reclocking has gone to point of dropping (or plugging) samples to accomodate 2 different clocks? What am I missing here?

12th August 2006, 11:11 PM	#9
rfbrw diyAudio Member Join Date: Oct 2001 Location: .	Common practice is to run the CS8412 in master mode. That way all clocks are synchronous with the clock derived from the SPDIF datastream. The DDDAC runs the CS8412 in slave mode. In that mode one is required to generate FSYNC/WS/LRCLK and BCLK/SCLK. These are then used to clock data out of the CS8412. In the pro arena these clocks would be derived from a VCXO locked to a common reference as would the CDP so there would synchronicity across the board. Alternatively one can derive the clocks from an XO and in turn run the CDP off a feed from the same XO. Again there is synchronicity across the board. Now DDDAC opts for neither of the above options, sticking solely with the XO. Needless to say any frequency other than that for which the XO was chosen is out. Even when the frequencies match one has to hope that the phase discrepancy between the incoming SPDIF datastream and the onboard XO are not enough to overwhelm any buffering the CS8412 may provide otherwise there will be dropped or repeated samples.

12th August 2006, 11:24 PM	#10
poobah diyAudio Member Join Date: Nov 2005	That is nuts... if I understand you correctly. It would seem to me that dropping or repeating samples would be the most horrible form of jitter. This is nuttier than the linear oversampling thread.