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

TNT

Member
Joined 2003
Paid Member
Ian!

Re: the dual clock board vers II:

- just running it on the bench using +5V, is there a way to switch clocks. One is only enabled at a time, rigth?

- is there an disadvantage to have both clocks running enabled simultaneously? Interference? If one plays songs using mixed Fs the clocks would need to warm up after every switch? Or is the enable just letting the clock signal out of the unit and its internally always oscillating?

- can I feed a local 3,3V but still rely on enable and ground from the clock board? Are the SOTA 3,3V already onboard? :)

//
 
Last edited:
Dual XO verII - SMA connectors

Ian,
we found the UFL connectors very effective, but very fragile too. Would you consider to add two, maybe unpopulated, SMA connectors for clock distribution on the DualXO verII under development?
We know that SMA are space-consuming, but they are stiffer than UFL.
Thank you Ian.
 
Ian,
we found the UFL connectors very effective, but very fragile too. Would you consider to add two, maybe unpopulated, SMA connectors for clock distribution on the DualXO verII under development?
We know that SMA are space-consuming, but they are stiffer than UFL.
Thank you Ian.

Yes +1 for SMA connector.
The UFL cables are fragile,sometimes the connector detached from the cable after slightly bend.
Pulsar,on the XO 1 remove the UFL connector and put a SMA
An externally hosted image should be here but it was not working when we last tested it.

19bff0dd0ef16b923e873c15be02cf8ad partagé sur ZimageZ
 

TNT

Member
Joined 2003
Paid Member
# On board pads for the NDK oscilators (NZ2520SD). Best value for money and delars that provide selected and measured individuals exist.

# A new take on physical isolation (thinking the whole board here - boxing?). Requirement: -30dB acoustical isolation 20-20kHz?

# hyper low noise DC feeding to oscillators - especially low freq 0-100Hz.

time, time and timing again :)

//
 
Hi Ian, suggestions for dual XO board II...please make it compatible for dual mono output (or at least an easy solder job like the last board). An SMA option would be welcome too...I get through too many u.fl connectors!!

Very excited by the forthcoming board...my system sounds better than it ever has before when using your project...good luck with development.
 
Dual XO II PCB layout

It's double side mounted four layers PCB. The new design optimized to low noise high frequency clock signals up to 100MHz. All clock traces are impedance controlled.

Dimensions, hole positions, connectors are compatible will previous Dual XO. Direct upgrading is no any problem.

Thank you so much for all of those suggestions. You can see some of them already been accepted. I'm still open for new suggestions at current stage.

Just hope you like my new design.

Regards,
Ian
 

Attachments

  • DualXOIIPCB3Dlayout.png
    DualXOIIPCB3Dlayout.png
    470.5 KB · Views: 525
Ian, UFL are very good connectors but loose performance after a limited number of mate/demate cycles also using the dedicated tool. We change often our setup for test purposes so a couple of SMA on the clock lines are better to us.
We were going to tell you this when you published the images of the new board and it seems that you already found room for 2 SMA on independent clock outputs, is our interpretation correct?
We had a look at the preview of your board and it appers to be as usual a great job with professional solutions and care for any detail. Congratulations Ian, we look forward to have it in our hands.
 
Hi,

This is a double posting from over at the Pulsar Clock Forum, hopefully further exposure helps to generate discussion/opinion etc.

I'm trying to get a better opinion re: sampling rate vs phase noise inherent to clock speed, all with regard to Ians Dual XO and a tda1541A running in simultaneous mode (up to 384.xxkHz).

I can afford to replace one clock, not both.. and so figured to use the clock that allows for direct multiples of the 44.1kHz sampling frequency.

Question is.. with the 45.xxMHz clock I can get to Fs = 352.xx kHz , but I see that phase noise is higher than 22.xxMhz which can get me to Fs = 176.xkHz. Of course both will allow for playback at 44.1kHz if no upsampling (via PC) is employed.

General subjective preference seems to be for lower clock speed, given that functions at the desired sampling rate, although I have noticed that Ians preference is for the higher speed clocks with the classic multi bit D/A chips - and after all this is with his Dual XO Board and I2S>PCM module, so I'm a bit lost.

Can any informed person please shed some light on which is likely the best way to go with regard to clock speed, Ians Dual XO, I2S to PCM board, and sampling rates.

Thank you,
Shane
 
Shane, I'm in practically exactly the same position, but I'm feeding an ESS 9018 DAC (BIII), so if anyone has experience with that setup I'd appreciate the information too.

I'm feeding the DAC with 96KHz material only - with Quad ESLs and Pass amps I cannot hear any difference if the conversion (from 44.1 or from 192/176 etc rates) is done well.

Thanks,
BK
 
BK,

Is that to say that you cannot detect any meaningful difference between the sampling rates so long as it feeds 96kHz to your device?

I have noticed an improvement, but really only very slightly in the PC upsample from 44.1 to 176.xx and it could either be in my mind, or due to the alias frequency being further removed from the audible band with Non-OS. I am un-convinced.

What clock speed are you using, and which clock board are you using?


Shane
 
Shane,
I have tried feeding material from all sample rates to the DAC, and I cannot detect _any_ difference (blind testing) between the original at, say, 176KHz and the version I've down sampled with Izotope to 96KHz.

I have also not detected anything negative comparing original 44.1KHz material and the up sampled version at 96KHz. There is plenty of variation in sample rate conversion routines as can be seen here http://src.infinitewave.ca

I have the si570 board and I'm waiting on the new Dual XO board for the new oscillators.

BK
 
Thanks BK,

My experience is limited to the CCHD-957 Crysteks and have tried (22.xx, 24.xx, and 45.xx, 49.xx). (numbers could be incorrect but you will know what I mean). I have not played with 176/192 recordings just up sampled via Audirvana (Izotope engine?) from 44.1.

The thing I've noticed is that the phase noise figures for the CCHD957 are about the same, regardless of clock speed. But then, these are specified only at 10Hz.

I truely believe that OCXO is the way to go, and especially if they are measured individually - each unit for sale as finished product, and at 1Hz. Pulsar Clock fits the bill.

Interesting that with the Pulsar Clock, the phase noise increases with clock speed, to the point where some went back from 90MHz+ to 45.xx (again with the number inaccuracy, but but you would get the idea) as the sound apparently became more strident, or less natural.

Yet, Ian prefers the high speed clocks in his XO board.. but then Ian has a new higher speed XO board for sale soon :)

Best I can tell, and from reading between the lines.. maybe best to stick with direct multiples of original sample frequency (44100) and keep the ability to playback 176.xx high res files.. and use the lower phase noise clock at 22.xx from Pulsar, use the 49.xx Crystek in the other position in the Dual XO Board.

Expensive.. but so is therapy :)

Any further information, insight, experience or considered opinion would be very welcome. Including of course Ian and the Pulsar Clock representative/s


Good Listening,
Shane
 
Last edited:
Rubber bands on your clock board

Hi there everybody!

Just a quick message to say one little thing...

Rubber bands are important to a very large degree.

If you haven't already put rubber bands on your clock board pleases do so, it seem that it is rather imperative.

Break those four little connectors and attach small rubber bands as i have tonight and be prepared to experience a rather large increase in sounds quality. Tonal quality, more detail yet softer sound should be experienced. I can't quite believe that such a small change can make such a large improvement. Yet it makes sense (crystals are microphonic).

Although i must point out that this may have been a weak link in my system. Depending on your implementation your mileage may vary. So easy to try anyway. If you've tried everything but rubber bands, it should be night and day.

Cut the 4 links and give it a go.

Ryan
 
Hi there everybody!

Just a quick message to say one little thing...

Rubber bands are important to a very large degree.

If you haven't already put rubber bands on your clock board pleases do so, it seem that it is rather imperative.

Break those four little connectors and attach small rubber bands as i have tonight and be prepared to experience a rather large increase in sounds quality. Tonal quality, more detail yet softer sound should be experienced. I can't quite believe that such a small change can make such a large improvement. Yet it makes sense (crystals are microphonic).

Although i must point out that this may have been a weak link in my system. Depending on your implementation your mileage may vary. So easy to try anyway. If you've tried everything but rubber bands, it should be night and day.

Cut the 4 links and give it a go.

Ryan

This is an implementation imperative for quality clocks,has already had reported by Pulsar....;)


In the end I just mention two details of our configuration that in our experience demonstrated to be effective to improve the sound quality:
a) Absorber feet reduces the vibrations that reaches the clock. It is well known that any clock is sensitive to the mechanical vibrations that are transformed in jitter on the output signal and this becomes relevant when it is added to a significantly low phase noise.

http://www.diyaudio.com/forums/vend...clock-ultra-low-noise-ocxo-4.html#post4116980