Asynchronous I2S FIFO project, an ultimate weapon to fight the jitter

[qusp]

Originally Posted by dvb-projekt

The TPS7A4700 low noise LDO board.

connect it as you would the battery board... its that simple. spinning a new gen board is a significant cost vs respinning the same board and its easy enough to do it as it is. most people that are going to go to the trouble to build another reg for the board will go with something a bit more exotic than another slightly lower noise LDO anyway IMO.

[andrea_mori]

The Si570 shows -112dBc@100Hz at 120MHz, so theoretically around -130dBc@100Hz at 10MHz; so, I presume -100dBc at 10Hz or less, since the manufacturer does not shows phase noise performance at 10Hz and 1Hz, that are much more important in audio than higher testing frequencies. A good oscillator should be reach -120dBc at 10Hz, not at 100Hz.

A good crystal should be in cold welded package, 10 ohm ESR or so, a few fF of motional capacitance, translated in 100-150K unloaded Q.

But the cheap Citizens like this
CSA309 11.2896MABJ-UB - CITIZEN AMERICA - CRYSTAL11,2896MHZ | Farnell Italia
should work fine in a simple cheap picogate oscillator. As far as I know these crystals does not perform all identical, so the best way is to buy a lot of them (50 pcs for about 15 USD) and listen to the result. I think you could find at least 5 good crystal from the bag of 50.

[qusp]

that replaces the si570 board does it? you seems to have completely missed the point....

[andrea_mori]

Since a member ask "What are the best clock to try", I try to reply to the question, I never said something replaces something.

BTW, you could build 2 or more of this "3 USD oscillators", for any frequency you are looking, the only limit is the suitable crystal you can get from the market.

[qusp]

sure, but you have suggested building the clock board plus trimmed cmos buffer as a cheaper solution...and doing 50 uncontrolled listening tests on 11MHz crystals to select a suitable unit as something that compares to 45/49/90/98MHz clocks that are already cheaper than the cost of the parts for your suggestion alone (not NDK or si570 of course)... one is a multiple speed unit, all with well designed boards and interfaces to connect to fifo

you have made all of these suggestions as a solution for a clock to try in the fifo clock module?.

(i'm interested in how you select these by ear, you just remember, take notes on 50 parts? again tick tock tick tock)

it just seems like straight up trolling to me, just like most posts you make in this thread, they are off topic. your posts contain the subject of clocks, but other than that have zero to do with this thread, its just OT noise IMO.

anyone who might take on building their own clock board with diff CMOS amp as a solution, is probably already aware of the fact it could be done....

[qusp]

I get the whole DIY as enjoyment thing, ive done my toils of matching discrete parts etc. ;but these are serious undertakings time wise. sorry i'm just getting frustrated with all these oddball interludes, many lead to suggestions the fifo is pointless/superfuous, or harping on about 44.1 as the only valid audiophile pursuit. All in a thread where people are pretty clearly willing to try new things and mostly want much higher speeds

[andrea_mori]

If you are looking for a dual clock frequency, most of folks in this thread does not need more than 2 frequency (I believe), it's a cheaper solution (I wrote 5 good from 50, so 1 from 10: 10 listening tests and 3 USD), but also IMHO (starting from the specs of the above devices) a better solution.

No limit for higher speeds, you can reach easily 100MHz with overtone crystal.

And again you forgot the question: "What are the best clock to try"

[Blitz]

I appreciate to learn from you and I was not aware that I could build a clock myself which outperforms the Crystek...(I am wondering if there is no project going on at DIYAUDIO on this or even a group Buy...this for sure would be on the list of many)..

...and the FIFO does its job, it is a must have, but why not than being hardcore as well on the Clock-Quality and PSU-Quality ?

DIY is for me about learning and get crazy and have fun...Andrea, your recommendation fulfills all of that fur sure...but selecting by ear from a bag of 50 is too crazy, even for me...

So...if you can recommend a better clock (which manufacturer ?) and if you have a link, I will build it...just for fun and to see what the FIFO can achieve in the Dual-Board-Setup ....and than I will do the shoot-out against the S570...just for fun.

[qusp]

Originally Posted by andrea_mori

If you are looking for a dual clock frequency, most of folks in this thread does not need more than 2 frequency (I believe), it's a cheaper solution (I wrote 5 good from 50, so 1 from 10: 10 listening tests and 3 USD), but also IMHO (starting from the specs of the above devices) a better solution.

No limit for higher speeds, you can reach easily 100MHz with overtone crystal.

yes of course you can, which will put pressure on the layout/termination and you will still need another clock for the 44.1/48k material since the board doesnt support 2048x FS, so needs auto switching between 4 clocks to emulate the si570, with all of them needing some trimming in the divider to get a close multiple

thus my statement of you missing the point of the si570

And again you forgot the question: "What are the best clock to try"

no I didnt forget...

its far too general a question to answer in this context and without a budget or aim. if you actually remember my query was regarding your minimization of the effort required vs quite a small money saving overall.

if just building a single standalone clock for a dac board and a single fundamental frequency is all thats desired, sure I might consider it, but I wouldnt be listening to 50 XOs to pick a few, sorry I have better things to do with my time. when you produce something that actually does reach the theoretical potential, (harder than you claim and you know it) then I might start listening a bit harder, these are not new concepts to me, but myself I know that the buffer already adds more, or at best about the same jitter as the NDK and crysteks sans buffer, if you have to multiply the fundemental, then...

so you may be able to equal or slightly better the common clocks in use in this thread for a tenner less, or really about the same, if you just lash it up and dont care about your time.

so remind me of the point aside from the adventure? which is fair enough as well as long as thats how its presented, rather than an easy way to save money and get a better result with no experience.... along with a very likely lack of test gear to verify the result

[andrea_mori]

As far as I know (maybe I'm wrong) the only clock project in DiyAudio is the Discrete low jitter clock from Anton Clark. I got a couple of PCBs, but I have not yet tried them. I own a bag of Citizen crystal, so I'll try them in Anton's circuit. The great issue is that cheap crystals are not selected and therefore constant quality is not guaranteed. A good crystal costs 100 times and more than the Citizen, and in this case the manufacturer tests all crystal and selects only those are conform. I bought 5 sample of good quality AT-cut crystal from Laptech, they came with a paper listing the specs of each crystal. Reading the document I found they produced 12 crystals to send me 5, 7 was throw out because they don't reach the minimun spec requested (Q below 100K). That's the reason whereby they are so expensive.

At this moment I'm working about a low phase noise oscillator using the above crystal, but my time is limited, so this project is far from the end. The prototype works, but I have to design the PCB to test its performance at a professional lab.
Unfortunately the most popular distributors such as Mouser, Digi-Key and so on, do not supply good oscillator