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

352?

That's interesting that Si570 enables the SPDIF board to pass 352/384k music. I'll look into why that may be, could just be luck and may not be working optimally. May need Ian to run a test with that config if he has time?

I think Ian has a solution cooking in his multichannel FIFO project. Alternatively I think you could have an i2s switch like TPA's OTTO-II to select i2s data from amanero or from spdif board and feed it into FIFO input. I do not know if that board would cope with 352/384k but I think it should do ok.

Chris

Hi Chris,

Just checked, with my tidier cabling, again my playing of 352Khz files, using or not the Dix9211 in the input board. Seems to play rather nice during a second listening..
Isn't playing 352Khz in stead of 215Khz(max for 9211) not a too big margin?
Curious what Ian can tell about it....
Regards,

Ed
 
385Khz

That's interesting that Si570 enables the SPDIF board to pass 352/384k music. I'll look into why that may be, could just be luck and may not be working optimally. May need Ian to run a test with that config if he has time?


Hi Chris,

Forget about what I said about 570 playing 384Khz!
I did put the 570 again in the system and this guy will not go higher than 352 as well! I am afraid my memory did let me down on this.

Switching with Brian's mux will not work I am afraid. I need switching from the three clocklines in one strike to SPDIF(One dig.channel). I will have to figure out something. The way Ian's input board works is fantastic and exactly what I need, besides the bandwith.

I will wait untill a solution pops up.
Regards,
Ed
 
Hi Chris,

Forget about what I said about 570 playing 384Khz!
I did put the 570 again in the system and this guy will not go higher than 352 as well! I am afraid my memory did let me down on this.

Switching with Brian's mux will not work I am afraid. I need switching from the three clocklines in one strike to SPDIF(One dig.channel). I will have to figure out something. The way Ian's input board works is fantastic and exactly what I need, besides the bandwith.

I will wait untill a solution pops up.
Regards,
Ed

I thought that was strange, is actually surprising that 352k works but like you say, its possible that there are tolerances on the ICs and the IC sample rate limits are conservative.

That's what Brian's mux would give wouldn't it? [assuming that it allows the higher sample rates - it should].

Use it to switch between the i2s output from Ian's SPDIF board and the i2s from the amanero/waveio/USB source that runs at 352/384k. Toggle the mux then use Ian's input selector to select opt/coax input. Two button presses/switches instead of one but still functional and ticks all of the boxes today, rather than waiting for a solution tomorrow.

Chris
 
Mux

I thought that was strange, is actually Use it to switch between the i2s output from Ian's SPDIF board and the i2s from the amanero/waveio/USB source that runs at 352/384k. Toggle the mux then use Ian's input selector to select opt/coax input. Two button presses/switches instead of one but still functional and ticks all of the boxes today, rather than waiting for a solution tomorrow.

Chris

Hi Chris,

Yes,
this one:cheers: should work! I was looking at the 4:1 Mux/demux but I need the later OttoII.
Thanks,
Ed
 
Hi Ed,

Just to confirm WaveIO Un-isolated output > FIFO > Original Universal Isolator (Ian) > Dual XO > I2S to PCM > DAC chip and now 384kHz success?..

.. and that the only physical change has been to drive from the non-isolated outputs (uFL) on the WaveIO?.

If so.. thats my next move !! :)

I wonder if the Universal Isolator is better before or after the FIFO in this case?, any thoughts?. Ian, perhaps if you're not swamped with work you can comment?.

Really good stuff !! :)

Regards,
Shane

The function of an isolator is to cut the ground loop and make the DAC has clean ground. So, it should be placed as close as possible to the DAC, I would prefer place it after FIFO.

However somebody might be worry about the PC ground, that's why they are trying to isolate the USB in advance. In this case, if you can isolate both, why not?But please keep in mind, if you have more than one isolator in your system, only the one close to DAC can improve the SQ.

Regards,

Ian
 
Update the new XMOS USB running at 384KHz with isolator

I received the new XMOS USB on weekend and did a quick test.The result was quite promising.

It runs perfect at 384KHz with isolator and 45/49 MHz clock board!

The new XMOS chip is in BGA package with 48Mhz system XO, the old version was in QFP package with 13MHz system XO.

The reason could be that the XMOS USB is running Async mode, therefore CPU clock is different from audio clock, but the I2S timing is generated by CPU. If it is true, higher CPU clock or CPU clock with less jitter result in higher quality on timing, especially for high audio Fs like 384Khz.

I would suggest try this new XMOS USB or Amanero 384 (which has an external CPLD dealing with I2S timing issue by hardware).

Ian
 

Attachments

  • S_IMG_DSC05296.jpg
    S_IMG_DSC05296.jpg
    302.9 KB · Views: 393
  • S_IMG_DSC05298.jpg
    S_IMG_DSC05298.jpg
    290 KB · Views: 387
  • S_IMG_DSC05302.jpg
    S_IMG_DSC05302.jpg
    237.6 KB · Views: 385
  • S_IMG_DSC05305.jpg
    S_IMG_DSC05305.jpg
    255.3 KB · Views: 373
  • S_IMG_DSC05306.jpg
    S_IMG_DSC05306.jpg
    225.8 KB · Views: 372
I actually tried diyinhk xmos board(not the latest version) with 48mhz clock... On board 45/49 clock with u.fl MCLK out... Could not work at 352/384. Only up to 192... Have to revert back to the joro xmos which works at 352/384.. But the diyinhk sounds better though...

I received the new XMOS USB on weekend and did a quick test.The result was quite promising.

It runs perfect at 384KHz with isolator and 45/49 MHz clock board!

The new XMOS chip is in BGA package with 48Mhz system XO, the old version was in QFP package with 13MHz system XO.

The reason could be that the XMOS USB is running Async mode, therefore CPU clock is different from audio clock, but the I2S timing is generated by CPU. If it is true, higher CPU clock or CPU clock with less jitter result in higher quality on timing, especially for high audio Fs like 384Khz.

I would suggest try this new XMOS USB or Amanero 384 (which has an external CPLD dealing with I2S timing issue by hardware).

Ian
 
384Khz and WaveIO/Fifo

I received the new XMOS USB on weekend and did a quick test.The result was quite promising.

It runs perfect at 384KHz with isolator and 45/49 MHz clock board!

The new XMOS chip is in BGA package with 48Mhz system XO, the old version was in QFP package with 13MHz system XO.

The reason could be that the XMOS USB is running Async mode, therefore CPU clock is different from audio clock, but the I2S timing is generated by CPU. If it is true, higher CPU clock or CPU clock with less jitter result in higher quality on timing, especially for high audio Fs like 384Khz.

I would suggest try this new XMOS USB or Amanero 384 (which has an external CPLD dealing with I2S timing issue by hardware).

Ian



Hi Ian,

The WaveIO directly into the Fifo, isolator between Fifo and Clockboard, shows to have no problem with 384Khz!
The culprit is the fact that if connected via the SPDIF input board the restriction is 192Khz.(PIX9211) Though my board has a margin to play 352Khz.
Do you prefer the SQ of the DIYINHK Xmos?
Regards,

Ed
 
Hi Ian,
Do you prefer the SQ of the DIYINHK Xmos?
Regards,
Ed

I'm interested in the reply to this as well.. I was thinking the FIFO shouldn't render any difference in front end. IIRC Krell transport via SPDIF vs generic all sounded the same (Ians post_I don't know where theres so much but.. it was all to do with verifying the functionality of the FIFO)

Shane
 
Hi Ian,

The WaveIO directly into the Fifo, isolator between Fifo and Clockboard, shows to have no problem with 384Khz
Regards,

Ed

Hi Ed, this is to say (again) that by directly, you are meaning the non-isolated outputs from the WaveIO.

And therefore, one could assume that the isolator on the WaveIO isn't up to 384kHz repro?.

Regards,
Shane
 
Last edited:
Hi Ed, this is to say (again) that by directly, you are meaning the non-isolated outputs from the WaveIO.

And therefore, one could assume that the isolator on the WaveIO isn't up to 384kHz repro?.

Regards,
Shane


Hi Shane,

With directly I meant using the three WaveIO U.fl outputs (no isolator) directly into the three U.fl inputs of the Fifo board.

The 'problematic' connection was WaveIO isolated output to the 'backdoor' of Ian's SPDIF interface input board.
This board, the SPDIF interface board, has the 192Khz (DIX9211) restriction. So I do not think the WaveIO isolator has this 192 restriction too, but I did not try that. But you could trust the isolator for being able to throughput 384 Khz.
Clear now?
Regards,
Ed
 
Last edited:
I would suggest try this new XMOS USB or Amanero 384 (which has an external CPLD dealing with I2S timing issue by hardware).

Ian

Hi all, if someone can explain please :

The two clocks of the Amareno are 22 & 24 Mhz and the device play synchronous ?

Does the Amareno not need to be feeded by an external clock to a better result via its input MCLK pin (not UF.L unluckily!) ?

Or is it better for you to let it play synchronous then the FIFO for reclocking at higher speed then isolate (because Amareno is one of the rare device to have not embeded isolator) then the PCM or not...

In the case of no FIFO used, does people need an external MCLK to feed both the AMARENO and the PCM board ?... my understanding with the answer of Ian about the PCM is it could be possible with a MCLK of 22/24 Mhz (because the limitation of speed clock of the Amareno) ?

What's the difference between the new XMOS USB and the DIYinkh one ?
The Ian's one does not have crystals and play synchronous instead the DIYinkh play 45/49 with two NDK Crystal and provide a uf.L connector to output mclk to feed an external device (like the PCM e.g. ?) ?
My understanding is the DIYink has its Fs programmed to 128 Fs and for people like me who plan to use PCM board and TDA1541... it needs to be reprogrammed at 32 or 64 Fs (because the TDA1541A) via an external and not free USB device (look at DIYinkh website....)

Well what a pity than the SPIDF protocol is limited at 192 Khz ! I would like to try the higher speed with TDA1541A step by step with a NOS at 48 Khz to make an idea by myself !
 
Last edited:
Hi Shane,

With directly I meant using the three WaveIO U.fl outputs (no isolator) directly into the three U.fl inputs of the Fifo board.

But you could trust the isolator for being able to throughput 384 Khz.
Clear now?
Regards,
Ed

Thanks, Ed.

Its clear if you're meaning Ians Isolator is good at 384k - and not the WaveIO Isolator which seems to be limited to 192k. :)

These u.Fl cables are hard to locate. Local source has 6-7wk delay, ebay has some out of china, price is reasonable but unsure about quality. Any ideas?

Regards,
Shane
 
Thanks, Ed.

Its clear if you're meaning Ians Isolator is good at 384k - and not the WaveIO Isolator which seems to be limited to 192k. :)

These u.Fl cables are hard to locate. Local source has 6-7wk delay, ebay has some out of china, price is reasonable but unsure about quality. Any ideas?

Regards,
Shane
Hi Shane,

The limit was that Ed was routing i2s via Ian's spdif board and using the spdif interface IC as an input selector. The IC is an DIX9211 by TI which has a specified limit to sample rate of roughly 200kHz. Was just luck that manufacturing tolerance allowed function at 352kHz. By bypassing that board and feeding i2s into the FIFO input directly Ed has said everything works correctly but he has not tested the Isolated output of WaveIO in this config. I think its reasonable to assume that WaveIO isolated output will work fine to 384k. I have a WaveIO here with old firmware, I will be able to test it out and confirm its behaviour in a few weeks most likely (would need to rearrange a few things and other projects taking up space on the bench at the moment).

Does that clarify?

u.fl cables are probably best sourced from mouser or digikey ... I agree local sources in Aus are lacking (or very expensive) for this sort of thing.


Chris
 
Hi Shane,

The limit was that Ed was routing i2s via Ian's spdif board and using the spdif interface IC as an input selector. The IC is an DIX9211 by TI which has a specified limit to sample rate of roughly 200kHz. Was just luck that manufacturing tolerance allowed function at 352kHz. By bypassing that board and feeding i2s into the FIFO input directly Ed has said everything works correctly but he has not tested the Isolated output of WaveIO in this config. I think its reasonable to assume that WaveIO isolated output will work fine to 384k. I have a WaveIO here with old firmware, I will be able to test it out and confirm its behaviour in a few weeks most likely (would need to rearrange a few things and other projects taking up space on the bench at the moment).

Does that clarify?

u.fl cables are probably best sourced from mouser or digikey ... I agree local sources in Aus are lacking (or very expensive) for this sort of thing.


Chris

Thanks, Chris.

Isolated WaveIO > FIFO > Ians Isolator > Dual XO 45/49 > DAC is limited to 192kHz. Thats all I know.

Need 3x u.Fl cables to go direct from WaveIO and bypass its isolator - hard to find.

Hopefully the direct connection will be ok to 384k.. as its essentially the same as what Ed has done, only thing that stands out is the isolator on the WaveIO.

None of its really clear, just need to try it and see.. but $48 shipping (digikey etc) from the US for something that weighs 100g and costs $9 is unreasonable IMO.

Regards,
Shane
 
Last edited:
Member
Joined 2003
Paid Member
I've been wanting to ask this too...

I'm interested in the reply to this as well.. I was thinking the FIFO shouldn't render any difference in front end. IIRC Krell transport via SPDIF vs generic all sounded the same (Ians post_I don't know where theres so much but.. it was all to do with verifying the functionality of the FIFO)

Shane

Can any of those who have used the FIFO with or without the isolator board comment on how well the FIFO reduces or eliminates the sonic differences between different digital sources?

Inquiring minds & such!!!

Greg in Mississippi
 
Thanks, Chris.

Isolated WaveIO > FIFO > Ians Isolator > Dual XO 45/49 > DAC is limited to 192kHz. Thats all I know.

The impression that I got was that config hadn't been tested by Ed. Or have I joined the conversation late and you're having a problem with the config described in your post? What DAC exactly?

My understanding of the conversation so far:

Isolated WaveIO > SPDIF Boad DIX9211 'backdoor' > FIFO > Ians Isolator > Dual XO 45/49 > DAC is limited to 192kHz (may work at 352kHz sometimes according to Ed's experiments)

Direct WaveIO > SPDIF Boad DIX9211 'backdoor' > FIFO > Ians Isolator > Dual XO 45/49 > DAC is limited to 192kHz (may work at 352kHz sometimes according to Ed's experiments)

Direct WaveIO > FIFO > Ians Isolator > Dual XO 45/49 > DAC works at 384kHz




Need 3x u.Fl cables to go direct from WaveIO and bypass its isolator - hard to find.

Hopefully the direct connection will be ok to 384k.. as its essentially the same as what Ed has done, only thing that stands out is the isolator on the WaveIO.

None of its really clear, just need to try it and see.. but $48 shipping (digikey etc) from the US for something that weighs 100g and costs $9 is unreasonable IMO.

Regards,
Shane

$48?!? last few times I've ordered from Mouser/digikey (March and June 2014) I had orders that were $100 - $300 and I checked the 'free shipping to Aus' options which require payment in AUD against paying their shipping charges (I recall it being something like a $30 flat rate parcel) and it worked out cheaper overall to pay their shipping charges. I'd avoid doing an order for just u.fl cables and order parts for a few projects at once, but that's mostly because I've got a Platinum Membership to diyAudio Addicts Annonymous and can't stop myself finding and building more projects all the time.

I will be putting an order in to Mouser/digikey a month or two after the time that the AYA II 2014 boards arrive and can add some u.fl cables to my order then if you'd like and put them in an envelope to you if you'd like? I just had a look and I don't have any spare ones. I have a couple, but they're earmarked for projects I'll be working on in next month or so, otherwise I'd post some to you.

Cheers,
Chris
 
Confirmed I2S backdoor of S/PDIF board works for 384KHZ

Hi Ian,

The WaveIO directly into the Fifo, isolator between Fifo and Clockboard, shows to have no problem with 384Khz!
The culprit is the fact that if connected via the SPDIF input board the restriction is 192Khz.(PIX9211) Though my board has a margin to play 352Khz.
Do you prefer the SQ of the DIYINHK Xmos?
Regards,

Ed

Hi Ed,

I did more test today on my system with feeding I2S signal into S/PDIF board backdoor, I confirmed S/PDIF board I2S backdoor works for 384KHZ without any problem! But not for all USB :).

1. Amanero Combo384: works for 384KHz, with/without isolator, both FIFO and S/PDIF board backdoor.
2. DIYinHK new XMOS: works for 384KHz, with/without isolator, both FIFO and S/PDIF board backdoor.
3. WaveIO old XMOS: works only up to 352KHz, without isolator, into S/PDIF board backdoor;Noise starts at 384KHz.

My system for this test:

iMac – iTunes – USB – S/PDIF board – FIFO – DualXO clock board (45/49Mhz) –PCM board – TDA1541A DAC (simultaneous mode)
50ohm coaxial u.fl cables between FIFO and S/PDIF board

There is no 384KHz limitation on DIX9211 I2S input. I think that 192KHz limitation was only for S/PDIF receiver.

Some tips running 384KHz,
With this Fs, BCK will run at 24.576MHz, while Data up to 12.288MHz, so we have to tread them as high speed clock signals
1. Use coaxial cables and u.fl connectors as possible as could.
2. Making cables as short as possible.
3. Termate signal sources with 33 ohm chip resistors to mach impendence with cables

Good luck to your project

Ian
 

Attachments

  • S_DSC05318.jpg
    S_DSC05318.jpg
    171.1 KB · Views: 190
  • WaveIO 2014-08-12 at 9.28.57 PM.png
    WaveIO 2014-08-12 at 9.28.57 PM.png
    572.7 KB · Views: 211
  • S_DSC05315.jpg
    S_DSC05315.jpg
    275.7 KB · Views: 444
  • DIYHK 2014-08-12 at 9.34.24 PM.png
    DIYHK 2014-08-12 at 9.34.24 PM.png
    504 KB · Views: 446
  • S_DSC05311.jpg
    S_DSC05311.jpg
    280.7 KB · Views: 443
  • Combo384 2014-08-12 at 9.35.51 PM.png
    Combo384 2014-08-12 at 9.35.51 PM.png
    528.3 KB · Views: 457
  • S_DSC05310.jpg
    S_DSC05310.jpg
    300 KB · Views: 460
Last edited:
Thanks for doing those tests Ian, interesting result!


I was going off the datasheet spec for 'routing'

attachment.php


So assumed that 216kHz was the limit for i2s signals also. It may be that different batches are built on different processes so all ICs may not come out exactly equal after fab tolerances perhaps?


Chris
 

Attachments

  • www_ti_com_lit_ds_symlink_dix9211_pdf.png
    www_ti_com_lit_ds_symlink_dix9211_pdf.png
    35.8 KB · Views: 745