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

TNT

Member
Joined 2003
Paid Member
I saw a picture of Ians setup and noticed that he used the opto input. I have tried and refused it before. With the FIFO it was an improvement. I suppose due to the galvanic separation. I run my system with to Toslink and it has never been better.

My take on sw-players is that if they don't change information (i.e. they are bit transparent) they are sound wise irrelevant and become a pure usability aspect. This position has nothing to do with the FiFo and I would still have it if the fifo didn't exist.

/
 
Irrelevant, huh? Nice position, but is it purely philosophical, or is it the result of listening tests?

I am asking these questions, as much to my annoyance software players and USB cables continue to sound very different in my system, using or not the FIFO.

Per example Jriver 17/Foobar, asio, exa > fifo > 9018. There is galvanic isolation at the exa's outputs and so far i am not using Ian's isolator but will. There was a huge improvement in sound jumping from the freebie oscillators to Crysteks and also a very substantial one going to 45/49. Will probably try the Si570 today. Overall the sound is very good but alas, not immune to changes in the source.
 
No theory. It makes no sense but the difference is not subtle at all. None of the cables are audiophile or expensive. The exa uses the silly micro USB connector and i've never felt a burning need to buy a proper cable.

I was just curious if anyone else has bothered to test in a similar setup. I normally use the EXA with two external supplies but in this particular build there is simply no room for additional transformers and power comes from USB. 9018 and Ian's board do have separate transformers, so PS coupling seems very unlikely. RF noise coupling between boards? Galvanic isolators leaking too much through?

To be honest i am not aware if any commercial equipment at any price level has managed to completely eliminate the effects of cables/ports/software from the picture. Maybe my expectations to do it on the cheap are unrealistically high.
 
Member
Joined 2003
Paid Member
What bothers me...

qusp said:
I would love to hear about your theory of how its happening, there is no mechanism via which it can possibly happen. its far fetched enough before isolation and then reclocking that lowers any source jitter from any source, to ps levels.

julf said:
I would love to see your ABX test logs.

... about this exchange is how glibly people respond to comments like Analog-SA (and TNT's... and Blitz's) with comments that, boiled down to their essence, say:

"You can't have heard what you say you did. You must be lying to us and to yourself. Give us a technical proof that we are right and you are wrong. Then back that up and prove that we are right and you lied to us and yourself with our preferred double-blind listening test to show that you just imagined what you said you heard."

First, in my world, it is not appropriate to call into question someone's assertion, especially about what they heard in their own system. Asking them about what they heard, the circumstances, and what changed when they changed the circumstances, all are perfectly appropriate. But in my world, where civility is important and people are given the benefit of the doubt first, even if you don't believe what someone said they heard, until you've tried it yourself and heard or not the same situation, you don't get a say! What they heard is what they heard. And no amount of your disbelief changes that.

Even if you try it and hear nothing, all that gives you is the right to say that, not the right to question someone else's credibility!

This forum (AND MANY OTHERS) would function so much more effectively and with much more civility if people stopped questioning other's experience when they listen to their own system!


Greg in Mississippi (<---FLAME SUIT ON!)


P.S. Funny how noone has questioned Ian's comments on the differences in the sonic impacts of different caps used in this fully-digital circuit. Does he get a pass because he is the designer and supplier and has a higher level of listening credibility?
 

TNT

Member
Joined 2003
Paid Member
Your PS somehow indicates to me that you aren't really on top of whats going on as You compare these two things.

All the "questioning" is part of the forum "game". To me its fruitful and educating. If one cant accept that I don't think this is a very fruitful place to expose ones experiences.

Mind You, I did leave a door open for information loss/alteration which is known to go on in some OS lower level mechanisms.

Do we want to submit to woodo or? Some sort of scientific analysis needs to go on, right? Or?

I conclude that the only response SA had was 2 that questioned the result. No one came to rescue for an advice to add more RAM, change some buffer size or invest in a golden USB cable - why?

Did SA check for bit transparency/integrity? I'm not questing what SA expressed that he heard but rather suggested that there might be no technical solution to the situation.

PS: I tink the "ABX" challenge is more of a jargon than an actual advice to do it - but thats me...... DS.

/
 
Not sure if i made myself clear but i only wanted to know if anyone had made a similar, purely subjective test. I am not making any claims nor wanting to start an argument with an engineering type objectivist. Especially one who doesn't own the specific type of kit, or any audio system whatsoever. And sorry, printers don't count. I am also particularly uninterested in proving anything to anyone.

The result of my test brings me mostly annoyance anyway. If i do have a theory it is that the effect of cables has little top do with the processing chain as it is. It is a lot more likely i am hearing a secondary, unrelated effect, per example one of the USB cables might have no screening and thus generate more RF noise, which in turn may affect the amplifiers.

In a similar discussion "engineers" were trying to prove the futility of SATA filtering, repeating their favourite mantra of "bits are bits", yet the filters may indeed affect the sound through RF radiation.

So, instead of coming up with theories i am merely curious to know if anyone is still hearing cables, even after passing through galvanic isolation and reclocking. In fact, it would be even better if anyone is using proper optical isolation of the USB port as well.
 
Last edited:
Greg, yes I tend to become skeptical when it is suggested that there are essentially quantum tunneling intelligent and evil audio only bits, or low level electrical impulses, that are invisible to instrumentation designed to measure artifacts orders of magnitude more sensitive. here they burrow through at least 3 memory buffers that completely reset the time domain (it does not even use any of the time domain information, plus the EXA is bulk mode USB afaik) and jump across >1kV isolation. After this feat, it subtly corrupts bit perfect digital data that has no way of representing subtle changes. All of this makes me more than a little bit skeptical and I will not apologize for that.

In the setup described, there is no electrical connection, no mechanical connection, no time domain connection and the data processing is verified as bit-perfect. perhaps it has something to do with the Pauli Exclusion Principle?

if I snuck into your house, presented to you and told you I had just walked through the brick wall to get into your house, you would require some sort of plausible explanation of how I did that, at a minimum...right? If I persisted with such claims publicly without that proof, I would be classified insane. Why do you have such a low threshold of proof for audio claims and why do you feel that i'm somehow being rude and unreasonable to question such a claim? The most reasonable explanation in this case IS psychological, or at best, erroneous system interconnection. We are all vulnerable to our minds playing tricks on us, I know I am.

actually I did question the tps reg cap thing, but even that is much more likely, it actually does have a direct connection to audio ground when being used for clock power, it actually does have an objectively verifiable impact (small but real) on both the transient response and noise level of the regulator, so while i'm skeptical of it producing differences in timbre, or soundstage depth, I cannot categorically rule out some minor effect in some situations

that is a very different situation
 
Last edited:
I'm not interested in discussing the cable issues, theses conversations come up too often and always end the same way.

Regarding the piezoelectric capacitor issues raised by Ian, I was going to raise it as a question (qusp and marce both did question it, if you read back). However I was reading the LTC6655 datasheet around that time (independently of anything going on in this thread). I found this paragraph on page 10 of that datasheet -

In order to achieve the best performance, caution should be used when choosing a capacitor. X7R ceramic capacitors are small, come in appropriate values and are relatively stable over a wide temperature range. However, for a low noise application X7R capacitors may not be suitable since they may exhibit a piezoelectric effect. The mechanical vibrations cause a charge displacement in the ceramic dielectric and the resulting perturbation can look like noise. If X7R capacitors are necessary, a thorough bench evaluation should be completed to verify proper performance.

Ian has tried to find the best combination of capacitance magnitude and physical size while being aware of the different dielectrics to produce the best theoretical performance. I spent a good while surfing digikey looking for a better SMT capacitor arrangement for the output of these regulators, including multiterminal murata capacitors for better HF performance, I couldn't find anything that was going to represent a substantial improvement over Ian's design. He has left room for people to try a film cap on the TPS7A regulators, I might even do that with one of mine. I've just built one of the TPS7A regulators on the weekend with the default BOM and just fired up my Si570 for the first time and enjoying the music for now.
 
Last edited:
I mentioned the piezoelectric effect, i'm well aware of it and its so often brought up as woo woo, to explain situations where neither vibration or significant enough voltage slew is present. X7R is better than most ceramic in this regard, but still worse than film or C0G. however, it requires some vibration, or largish voltage transients to have meaningful effect. the application has neither and the fifo boards themselves are covered in small x7r, so ... they are viewed as being preferred to c0g or film due to having some electrical damping and the availability of smaller size, larger capacities.

even if it did miraculously produce some meaningful effect, all the point of load caps are there to smooth it
 
Last edited:
I have read the comments on s570 vs dual boards few pages back, but as they are now 4 weeks back: Any news ? I am still on Dual board with 957, have not yet started s570, as other priorities needed my time and I did not see it as a priority if it is not an improvement overall. Currently I use the dual buffalos in asynchron mode with their own 100mhz clocks on board...thanks for your input.
(sync mode will come with the two acko dacs I srill need to fire up here).

By the way...funny discussion about abx testing and finding scientific truth about what other people hear.

I may be too pracmatic for this, but I am always thankful if other share their experiences. Not all of them I can reproduce in my setup, but it gives me oftenly ideas and choices I never considered before.

I test it and make my own opinion. And report back. Your choice, if you than want to do something with it or if you want to debate if something cant be.

But In Audio, the effect of small stuff you oftenly cant explain or which should not do a difference in theory, adds up so brutally, that not the big things are dominating the chain, but the hundreds of small things which add up.

The fifo takes a big portion of errors out of the equation, and allows this refinement much better than before. A big fun and adventure.
 
Last edited: