diyAudio

diyAudio (http://www.diyaudio.com/forums/)
-   Digital Line Level (http://www.diyaudio.com/forums/digital-line-level/)
-   -   Multi-DAC: Lynx AES16 -> DIR9001/SRC4392 (-> SRC4192) -> PCM1794A (http://www.diyaudio.com/forums/digital-line-level/166450-multi-dac-lynx-aes16-dir9001-src4392-src4192-pcm1794a.html)

novec 6th May 2010 11:11 PM

Multi-DAC: Lynx AES16 -> DIR9001/SRC4392 (-> SRC4192) -> PCM1794A
 
After a lot of searching and not quite finding, I'll try a new thread...

I'll be running a Lynx AES16 sound card running digital XO and DRC in my 'puter, and outputting 96/24 balanced AES/EBU signal to a total of five stereo DACs. Three of these (fronts) will be as high end as I can get them, while the other two (center/surround) will be so-so. I've got a total of five DIR9001s, three SRC4192s, three SRC4392s and five PCM1794As to play with.

So, first question: Chip configuration. I'm planning on a simple DIR9001 -> PCM1794A with clock recovery for the so-so DACs, without any SRC. The "good" ones are the headache - do I go DIR9001 or SRC4392, and if I go for the DIR9001, do I really need an SRC4192 after it? And without any SRC, will the recovered clock from the DIR9001 be good enough? I doubt it, but it would make things easier... As all DACs will only be run at 96/24 from the Lynx there's only one format to worry about, and it's pretty clean. It can run from an external high-end clock, too.

Next: Balanced AES/EBU input circuit. The SRC4392 has a balanced input, making it simpler on that part, but it requires an extra 1.8V supply. The DIR9001 only has single ended input, and needs some kind of circuit for the interface. All Lynx outputs are already transformer coupled.

There's gonna pop up a lot more, but let's leave it at that for now ;-)

SoNic_real_one 7th May 2010 12:16 AM

Aren't you doing the upsampling in the PC software? I am not familiar with Lynx software and capabilities, but I guess you can have 96kHz from any source. The 96kHz samplingrate gives better TDH+N and channel separation on those DAC's. And DAC's are doing upsampling to the same final value.
Jitter on SRC output is 200ps and on DIR output is 50ps (I doubt that your card is worse). In this way, the SRC would be useless... and I will go for DIR->PCM on all of them.

novec 7th May 2010 08:45 AM

I'm using software 96/24 upsampling (or in rare cases downsampling) as the first filter on the source, outputting to room correction filters before it goes to FIR crossovers. So the AES16 just takes the processed 96/24 signal and sends it right through. It doesn't support 192/24 without using doube channels, but hey - how much difference does it really make?

I'll go for my original idea then, with DIR to PCM, so the only challenges now is to create the summing circuit for AES/EBU in and distribute a good master clock to three PCMs. Still don't like the idea of taking a recovered master clock from the DIR to drive the PCM on the high end channels.

novec 8th May 2010 05:37 PM

D'oh! I dove deeper into the datasheets and found out that the only way to connect the DAC directly to the DIR is to use the recovered clock from DIR as master clock for the DAC. The master clock on the DIR does essentially nothing (it will only output audio based on recovered clock anyways) and the DAC has to be in sync with the DIR do to any good. So the only way to be able to use a proper clock I need to put an SRC in the mix. And as far as I can see, a DIR9001 + SRC4192 would be better than a single SRC4392.

Any experiences to share?

novec 9th May 2010 01:23 AM

After a whole lot of ifs, ands and buts I've settled on DIR9001 -> SRC4192 -> PCM1794, but I'll make a couple of test boards with the option to exclude the SRC. Next up is power supplies and implementation, which another user has started a thread on here.

SoNic_real_one 9th May 2010 01:32 AM

I think it is better to use a good clock in the source, than trying to "fix it" in the SRC. I think that the lynx board has a decent clock.
SRC will not improve the jitter. As I shown, it has higher jitter than DIR.

novec 9th May 2010 12:43 PM

I know the Lynx clock is decent, but when has decent ever been good enough for us :-) The best thing is always to have a "perfect" clock at the source, then there would be no need for ASRC at all. Being a pro tool, the Lynx has external word clock input too. I'll check into how that works and see if I can run it from a better clock.

Both DIR and SRC attenuates jitter, but in different ways. The DIR recovers a master clock from the S/PDIF stream with a PLL (Phase Locked Loop) that will "spread out" the effect of incoming jitter, while the ASRC uses a separate master clock as a stable basis for some heavy interpolation and decimation filters.

I spent the better part of last night reading a thread about ASRC by "werewolf", co-founder of Silicon Labs with a masters from MIT and ten years of experience from Crystal, so I've finally gotten a pretty good idea of how the whole thing works. I recommend everyone with an interest in this (which is probably everyone in here, except tube guys :-) to skim through it.

That still leaves me with S/PDIF input circuit and power supply scheme to worry about...

tritosine 9th May 2010 12:58 PM

Quote:

Originally Posted by novec (Post 2181000)
I'll check into how that works and see if I can run it from a better clock.


Ive had an AES16. There are way better solutions now.
https://www.xmos.com/products/develo...kits/usbaudio2

Wordclock: no luck with this one, you need a DAC that has wordclock output, so it can act as master, and lynx becomes the slave, the lynx PLL is following the DAC clock.

Sounds good, but theres no such DAC. One is UA2192 , has ADC builtin, plus it has its own serious PLL/clock generation solution , and expensive as hell.

DIY solution would be SRC4392, disabling ASRC block, using 'clock out. Too much to mess around with. I2s is simpler.
USB+i2s might need LVDS rx / tx for usable cable lengths. Thats cool, jitter is kept very low for anything but 1bit.

fmak 9th May 2010 03:16 PM

Quote:

Originally Posted by novec (Post 2181000)
I know the Lynx clock is decent, but when has decent ever been good enough for us :-)

Have a look at the clock frequencies on your Lynx. Mine has one clock with over 450 ppm out!! This is why synchrolock does not work and why they recommend you don't use it.

There are better cards around and pre-upsampling using good software sounds better. HDDs are cheap and sata connection particularly easy and fast.

theAnonymous1 9th May 2010 05:08 PM

Quote:

Originally Posted by novec (Post 2180660)
After a whole lot of ifs, ands and buts I've settled on DIR9001 -> SRC4192 -> PCM1794, but I'll make a couple of test boards with the option to exclude the SRC. Next up is power supplies and implementation, which another user has started a thread on here.

I've done the DIR9001 -> SRC4192 -> PCM1794 thing. It's a good sounding combination IMO. Can't offer anything other than a very subjective opinion on it though.

http://i34.tinypic.com/np4eft.jpg

http://i41.tinypic.com/2vhu2wk.jpg


All times are GMT. The time now is 08:48 AM.


vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2014 DragonByte Technologies Ltd.
Copyright 1999-2014 diyAudio


Content Relevant URLs by vBSEO 3.3.2