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

Znarfomat,

it is hard for me to look for manual now, but all your questions are answered there. Crystek 957 type clock will be the best choice. You need two of those with frequencies 45.1584 and 49.152 Two of those will get you any sampling rate up to 192KHz. For particular set up, check manual, it is all there. After reading, if it is still unclear, post your question.
 
I would say its never too many I2S inputs :). But it might be more practical to have some kind of switcher instead of too many inputs. In my mind, that multi channel I2S input could serve as stereo input as well, right? Than we have right there:

A. Two Stereo I2S or One Multichannel and one Stereo
B. S/PDIF, AES/EBU, Toslink I would say optional with your existing input board?

If I am just selfishly thinking, for my own situation, this would be the ideal scenario at the input: Multichannel USB to I2S, S/PDIF and on a second I2S input I would love to se ADC based on this:
ESS Products - ADC

The reason having a ADC is for anyone having digital crossover, has to go through ADC if desired to play analog source.
So two I2S and S/PDIF. Plenty would be three I2S or switcher instead.

Hi Vladimir,

I agree ADC input to a DSP system is important and that's a very nice ADC chip [maybe next year will see some news on PCBs for that too, I know some have been looking at designs for it ...] ... however ... I don't see much point for a stereo ADC input to the FIFO in that use-case. Because the FIFO would typically sit in the chain at the DSP output you need to connect the ADC to the input of the DSP and have that switching happen at input to the DSP. There might be other usecases where it would be handy though.


Hi Ian,

I've been thinking about multichannel FIFO interfacing ... one question though ... with the DIX9211 board do you need a 3rd port? I think the DIX9211 can switch DSD or i2s streams so with a revision to that board you could cover i2s/DSD and SPDIF/TOSLINK/AES inputs.

For people using the multichannel FIFO I'd almost expect that their input switching would be taken care of elsewhere (input to a DSP or in some sort of HDMI/toslink/spdif DD/DTS/LPCM decoding board) so one multichannel port on the FIFO would cover everything that group of users would want.

I sent you an email last week too btw :) [not sure how often you check that email account]


Cheers,
Chris
 
OCXO review: 90.3168MHz/98.3040MHz, -135dBc/Hz@100Hz

I was tied up with GBV and couldn’t do any research job until last week I did some carefully comparison testing on this pair of OCXO with CCHD957 and Si570.

Good news and bad news. The good news is I have to say those OCXOs are really good. The bad news I’ll tell you later on.

The test was done on my system with FIFO KIT, ESS9018, Pass pre amplifier and power amplifier and B&W speakers. Clock board was powered by single cell of A123 LifePO4 battery directly.

The overall impress was that those OCXOs really sound different. Unlike CCHD957 which makes music more musical, and Si570 which has some sweet spot on high range, those OCXO make the music real. This makes me remembering a couple of years ago, when my wife and I had wonderful dinner in hotel restaurant. A violin player was playing at the quiet lobby. I was standing there, closing my eyes, carefully listening to the solo, trying to distinguish what the different with my audio system. My audio system sounds a bit rich, but the real sound comes with solid density and all the details. The music was exactly there that you can touch, no matter you close of open your eyes. Now, with those OCXO, my system plays music more close to the real than before.

With those OCXO, the sound stage was huge and every instrument sounds from its position like a point, you can distinguish both position and distance as a vivid 3D picture even you open your eyes. But when you close your eyes, it seems like the music is something you can touch.

I still need take some time to get familiar with this kind of sound. And now, I start to miss the sound of CCHD957 and Si570 :).

It seems that clock jitter not only affecting the sound quality by Fs but also by effecting the D/A (DS DAC) conversion process itself.

The next, I may do some jitter test on my LeCroy. But I fairly believe that it’s already beyond the ability of the jitter measurement noise floor of my oscilloscope.

Please see the pictures for more details on testing setups.

Ian
 

Attachments

  • DSC00272.jpg
    DSC00272.jpg
    340.5 KB · Views: 726
  • DSC00278.jpg
    DSC00278.jpg
    464.5 KB · Views: 705
  • DSC00276.jpg
    DSC00276.jpg
    423.8 KB · Views: 681
  • DSC00287.jpg
    DSC00287.jpg
    458.2 KB · Views: 671
Last edited:
Ian,

that is a great news. Hard to believe it is still possible to improve on the set up you have. I am just wondering what is the bad news?
Regardless, a few questions. I see you have some kind of isolator after Amanero. What kind of isolator is that one, since that is a different one from the one used before the clock? Also I see you use Legato, just like I did until recently.

I have been experimenting with different I/Vs and output devices, and I went back on experimenting with transformers. After few, I settled for now on Lundahl LL1674 but used in reverse as 2:1, in order to provide a proper loading to the DAC. I hope you will try the same or similar, in my case is day / night difference to Legato. Highs completely opened, the depth and soundstage just exploded. There is nothing else, just Lundahl between DAC and my 2A3 tube amp, and when I tried for a first time, I just could not believe the difference. Imagine, I am not new to transformers in the output from DAC, but I was just astonished with the difference. Once you hear it you would never go back. It is not necessary to use the one I used, but high impedance on primary is what is needed.
 
My setup:
Waveio latest firmware, Fifo+Isolator+dual XO board with 45.x and 49.x crystek clocks. LiFePO4 2x for fifo, 2x LiFePO4 with belleson 5v reg for the clock board. My DAC ist the 1794NOS from Doede, 4 decks. It is a 1794 dual mono NOS DAC, fed with i2s from a special motherboard.

http://www.dddac.com/dddac1794_circuit.html

My problem:
I´m only able to play something upsampled to 352khz but the music is completely distorted and with a lot of noise (pink noise). The jumpers on the clock board are set: tp4+5 and tp7+8. On the input side (left) is the 45.x clock, on the right side is the 49.x clock.

without the Fifo etc. everything works fine.. Have no idea where to start looking for the failure...

apoo
 
Last edited:
Apoo,
according to the DualXOclockBoard Manual, your jumper settings (tp4+5 and tp7+8) are for 11.2896 and 12.288 MHz Oscillators.
Maybe wrong jumper settings are tha cause of your problems.
45 and 49 MHz are not mentioned in my Manual.
Do I have an obsolete Manual? It is the newest I found, from GBIV post.
 
Apoo,
I think your jumper settings are wrong according to the Double Speed mode of Dual XO Clock Board - Manual.
Qoute:
U1 frequency Jumper settings
45.1584 MHz TP3 – TP4 short
22.5792 MHz TP4 – TP5 short

U2 frequency Jumper settings
49.1520 MHz TP6 – TP7 short
24.5760 MHz TP7 – TP8 short

Maybe the Double speed option is not optimal or actually dangerous for your PCM1794.
Quote:
'Some applications need higher MCLK frequency. For example running SARBE32 DAC at synchronize mode. Dual XO Clock Board has a reserved double speed mode which could allow XO frequency going up to 45.1584 MHz and 49.1520 MHz to meet this kind of requirements. The double speed mode is not recommended to classical DACs and S/PDIF DITs to avoid any risk of possible damage. Because I don’t know what gonna be happened if
feed MCLK with up to 1024*Fs into those devices.'
 
Last edited:
zn, there is an extra PDF in there with the details for V3.8 of firmware where double speed mode is discussed which I think will hold the answers you're after :)

Ian made adapter boards available that suit CCHD957 did you get any of those?

I'm not familiar with deode's board, can it accept 1025x fs input? Might need to ask him or someone here might know perhaps
 
Last edited:
Thanks a lot! Where did you find the new manuals, wiki?

With the jumpers at 3-4 and 6-7, all sampling and FS LED's are flashing very fast on the dual xo board!?!

With the jumpers at 4-5 and 7-8 ,from 88khz to 192khz it plays some music but with distortions, especially the higher frequencies..

I will do some lecture ��

Apoo
 
Znarfomat, I have one adapter for you! :) thx for the link! Had them already but didn't open the second PDF...

By the way, I also use the isolator board between FIFO and dual xo board. Do I really have to remove L11 in this case for using J5 for separate DC?
 
Last edited:
apoo, I don't have the docs of a FIFO board handy to check the L11 question, I'll try to have a look later today if someone hasn't chimed in already.

I wrote up that wiki for exactly this reason, it's hard to find some of those docs between these two long threads .... the wiki links to the manuals probably should be updated to more recent revisions, I think I put them up when GBII was running, for the most part their hasn't been any firmware revisions or major changes so they should be okay for FIFO and DualXO board.

Cheers,
Chris
 
AR2, what impedance is your DAC chip seeing?

It is hard to say exactly, but LL1674 is reversed so it is 2:1 where at that DAC sees 4x impedance of the next stage, in my case tube amp. I would say DAC sees very high impedance. Definitely DAC is used in voltage mode. There is a thread going on, about transformers used in output of ESS 9018 /12 and their approach is different. There is suggested to push ESS to current mode with very low resistance and than use transformers with very high primary / secondary ratio. I did not try that, but I am little bit worried about that approach.