|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
85.19.92.6
In Reply to: RE: An insolent provocation .... if a usb dac is very well design (and built) the pc quality should not matter posted by Thorsten on December 10, 2014 at 03:08:10
Hi and thanks for the reply
What i meant is that if the usb dac is designed and built correctly and therefore provides, among other things, full isolation from the PC, changing the pc source upstream, with any other thing unchanged from the usb dac down, should not affect the final sound significantly.
Maybe only an extremely resolving system could highlight small but not very significant changes in the sound.
I think tha this could in some way tame all the debate about pc quality and so on and we could put more energy and resources on the rest of the chain from the usb dac down to the speakers
Just a decent pc able to run a 64 bit OS should be enough.
I have understood by the way that the usb interface is a very critical point and not that easy to execute rightly.
Of course i would be very happy to hear the opinions from someone who has tried in a high quality system with a very high quality usb dac to swap the pc source
Thanks again.P.S. i am sorry for the very bad English but after having founded Londinium in 50 A.D. we have in some way lost the control of the situation
Kind regards,
bg
Edits: 12/10/14 12/10/14 12/10/14 12/10/14 12/10/14 12/10/14 12/10/14 12/10/14 12/10/14Follow Ups:
Hi,
> What i meant is that if the usb dac is designed and built correctly
> and therefore provides, among other things, full isolation from the
> PC,
Good luck finding one.
Common digital isolators in the Digital Audio section can easily add several nanoseconds of jitter, so they are rare.
Readily available USB Isolator systems remain resolutely stuck at 12MBps, which limits them to 96kHz 2-channel audio (good enough for CD standard Audio but little else).
Another option is to isolate the ULPI Bus if available, this means bidirectional isolation for 8 Bits, Clock, direction and other housekeeping at > 60MBps, also no cakewalk.
So isolated USB is usually rare. As another thread showed, even some gear that should be isolated reads 0 Ohm between USB ground andAudio Ground (read no isolation whatsoever).
> changing the pc source upstream, with any other thing unchanged
> from the usb dac down, should not affect the final sound
> significantly.
Lot's of if's and butt's (more than in a Vegas Strip Joint) here.
Like you must also keep noise from the PC out of the mains, unless your gear is designed to block such noise.
But yes, IN PRINCIPLE, if money was no object, if any agency requirements could be safely met or ignored if they cannot be met and if all parts of the audio systems were designed for sufficient EMI/RFI etc. Resistance, then yes, your contention should hold true.
Just remember that in theory there is no difference between theory and practice...
Ciao T
At 20 bits, you are on the verge of dynamic range covering fly-farts-at-20-feet to untolerable pain. Really, what more could we need?
"Common digital isolators in the Digital Audio section can easily add several nanoseconds of jitter, so they are rare. "
Not that rare, and they don't need to add any time delay at all. An experienced digital designer can do it.
However, even with complete galvanic isolation, the power supply on the PC and the playback software still matters. Ripper software still matters. Don't ask me why...
Steve N.
Steve,
All the isolators for digital signals I have encountered are subject to delays, delay skew between channels and most crucially delay skew on each transmitted bit. This is due to their fundamental principle of using a high frequency carrier and some kind of magnetic or capacitive coupling. Even optical systems are subject to some delay and skew.
But good to you if you have something that isolates (even at RF) and is not subject to the usual limitations.I hope you use it extensively to gain a competitive advantage.
Ciao T
At 20 bits, you are on the verge of dynamic range covering fly-farts-at-20-feet to untolerable pain. Really, what more could we need?
Certainly isolators insert delay, but so do flip-flops and clock periods are delays too....
Hi and thank sincerely for the kind explanation unfortunately way beyond my knowledge
You say " isolated USB is usually rare " i say ... why ?
Is it really that difficult ? is it really that expensive ?
Just asking
When i read of people who use usb power supply and say that there is a big improvement i wonder who has designed the usb dac ... he knows what he is doing or what ? That should not happen at all.
If you see my point.
Thanks again.
Kind regards,
bg
Hi,
> You say " isolated USB is usually rare " i say ... why ?
Because it represents drastic engineering challenges that so far have not been overcome in a manner and fashion that easy to implement, never mind cost.
> Is it really that difficult ?
If someone brings out a suitable Isolator Chip that works at High Speed USB2 rates (480MBPS), no, until then it is trivial if all you need is 12MBPS, or should be. If using 12MBPS the maximum standard sample rate supported is 96kHz - 2-Channel.
> When i read of people who use usb power supply and say that
> there is a big improvement i wonder who has designed the usb
> dac ... he knows what he is doing or what ?
Probably yes, he does know what he does.
> That should not happen at all.
> If you see my point.
I do not. Particularly.
If you are willing to spend unlimited amounts of money you can of course demand the kind of things you suggest. In a cost limited application, it is a different story.
Ciao T
At 20 bits, you are on the verge of dynamic range covering fly-farts-at-20-feet to untolerable pain. Really, what more could we need?
Hi and thank you very much again for your very kind and helpful reply.
Even of course the all stuff is quite beyond my ability to understand i have realized that, as often i do, i was very trivializing the issue
I thought was a much easier task but clearly it is not.
No more ramblings on the subject. Promised.
Thanks a lot again.
Kind regards,
bg
Hi,
To give you a "datapoint", you can but commercial USB Extenders that send USB over optical fibre and handle 480MBPS (High Speed) and of course also isolate the signal.
All the ones available I have seen so far cost around 1,000 Euro Retail and there are not many options out there.
The Owen Corning Optical USB Cable starts around 300 USD per piece (likely 300 Euro if sold in Europe) and if employed correctly also isolates a single, self powered peripheral (e.g. DAC).
So if you are willing to pay that much you may isolate your DAC from the Computer without penalty (other than financial). I am waiting for one of the Owen Corning cables (they seem in short supply) to try.
Ciao T
At 20 bits, you are on the verge of dynamic range covering fly-farts-at-20-feet to untolerable pain. Really, what more could we need?
Hi and thanks again for the very interesting advice
I guess that in a very high rez/quality system this devices can really make a difference
As a start i will listen for any noise or audible click/pop and distortion
Personally i think that digital recordings with a soundstage particularly well captured are a very good tool to assess the quality of a playback system.
If the system images very well is a " really damn good system " as Mr. Doug Sax used to say. And i agree of course.
Thanks a lot again.
Kind regards,
bg
Don't even consider isolating the USB signal. Were you to do so, the packet processing overhead on the DAC side of the isolator would still nail you as John Swensen discovered. Treat the entire USB section as "dirty". Do the isolation at much lower data rates, e.g. I2S rates. Send the clock from the DAC through its isolator to the USB converter section, synchronize it with the packet buffer, and send back the audio data in the DAC clock domain through a separate isolator. You can easily reclock (flip flop or shift register) in the DAC clock domain. Repeat the isolate and buffer mantra as necessary...
A single stage flip-flop will eliminate some noise (phase noise and amplitude noise) but not all because signals will leak around open switches and closed switches because the flip flops aren't completely "hard". However, multiple stages can be ganged in a shift register given appropriate clock phasing to get more isolation. The limit will be power rail noise shared by the stages of the shift register. Hence, the possible need to use a second stage of isolator, and reclocking to remove its noise.
Note that multiple stages will not produce a perfectly clean digital signal. There will still be noise. However, multiple stages of isolation will provide as much decorrelation as desired from the original signal. The problem of getting good analog to digital conversion will remain. :-)
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
Correct, I2S is where you isolate USB interfaces and any other interface for that matter. Works like a charm.
Tony,
There are many options to isolate things.
One of the problems is that the common RF based isolators often cause more RFI than they remove. And if you need 50MHz bit-clock few isolators will even work acceptably.
All options have substantial challenges to resolve and are not cheap.
I tried something using in principle Ethernet targeted parts (magnetics and LVDS drivers), but even this in the end is neither cheap nor that good.
If you have any suggestion of a simple and reliable low jitter solution that does not create massive amounts of RFI of it's own, I am looking, a PM will be fine though I would not mind if you can share it in public.
Ciao T
At 20 bits, you are on the verge of dynamic range covering fly-farts-at-20-feet to untolerable pain. Really, what more could we need?
No, I have no magic solution, just using multiple stages of isolation. I think the bulk can be done with optical isolators. I suspect pulse transformers will work as well, depending on the signal coding due to the problem of passing DC. The last stages of buffering will require hard flip flops and very very careful power and ground distribution.
The first problem is to characterize the situation, i.e. to see what people are hearing and how sensitive they are. Then the second part of the problem is to figure out how to measure this reliably, including measuring isolation from the source.
If I were you, I would look for people who design high speed crypto equipment (red black separation). They probably have this problem pretty well nailed, since their goal includes preventing any information from leaking other than the intended bits.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
Tony,
Common optical isolator are way too slow for the sample rates I run my stuff at.
Pulse transformers that are off the shelf suffer bandwidth limitations that are not easy to overcome.
The bit at the end is not so difficult... I had that for ages in most of my DAC's.
Ciao T
At 20 bits, you are on the verge of dynamic range covering fly-farts-at-20-feet to untolerable pain. Really, what more could we need?
"Common optical isolator are way too slow for the sample rates I run my stuff at."
Here you go. Even a model where you can run the clock in the reverse direction without requiring a separate chip. (Of course this is only a spec sheet, may be useless marketing BS.)
Yes, you will have to reclock it.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
Tony,
I am familiar with NVE's isolators (and most other from any major manufacturer). Please enquire into the operating principle, then you might see why having these can be worse than having no isolation.
Ciao T
At 20 bits, you are on the verge of dynamic range covering fly-farts-at-20-feet to untolerable pain. Really, what more could we need?
Most XMOS designs only need 24.576MHz across the I2S interface. This is easily isolatable. Why do you need 50MHz?
Hi,
> Most XMOS designs only need 24.576MHz across the I2S interface.
Well, I need more than that, ok?
> This is easily isolatable.
Depends on your definition of "easy" and "isolate".
> Why do you need 50MHz?
Now that would be telling, no?
Ciao T
At 20 bits, you are on the verge of dynamic range covering fly-farts-at-20-feet to untolerable pain. Really, what more could we need?
Look, I also use 45 and 49MHz MCLK, but I don't need it in the I2S XMOS interface.
I'm sure some of the crypto circuit designers have a handle on this problem. I never got close to those particular spooks, however. The spooks that I did meet with in person were scary enough.
As you pointed out, there are optical transmitters and receivers that work at multiple gigabit rates. There are probably other suitable design techniques to deal with power/ground noise affecting signal transitions, or somehow decorrelating the noise from the musical signal, thereby rendering it benign. I'm a systems guy, not a circuit designer, so I'm afraid I can't help you out with details.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
Hi,
> I'm sure some of the crypto circuit designers have a handle on
> this problem. I never got close to those particular spooks, however.
I suspect they do, from a different angle. But Deep Crypto scares me, the people into in are a bit odd.
> As you pointed out, there are optical transmitters and receivers
> that work at multiple gigabit rates.
But not necessarily ones optimised to carry 50MHz/50MBPS signals with low jitter, or at prices that make inclusion in relatively inexpensive commercial equipment feasible.
> There are probably other suitable design techniques to deal
> with power/ground noise affecting signal transitions, or somehow
> decorrelating the noise from the musical signal, thereby
> rendering it benign.
Yes, but that does not really help with things that can only isolation (or breaking the noise loops in other ways) can fix.
What we (Computer Audio) need are commoditised USB2.0 High Speed isolators OR good isolators for I2S style digital signals up to 100MHz that do not operate based on RFI producing principles and have low jitter.
Ciao T
At 20 bits, you are on the verge of dynamic range covering fly-farts-at-20-feet to untolerable pain. Really, what more could we need?
The crypto people are very much concerned with jitter on their signals. Jitter provides a "covert" channel that can leak secret information, bypassing the encryption algorithms. I'm sure they work with signals that are vastly faster than 50 Mbps these days. As to how fast they want to go, I was told , "ten times faster than the fastest that is possible".
You may find this article interesting.
Tony Lauck
"Diversity is the law of nature; no two entities in this universe are uniform." - P.R. Sarkar
Tony,
Interesting, but fails to answer "commoditised solution".
Now I remember precisely why I find Deep Crypto so scary...Thank you not for reminding me.
Actually, I gave up on anything serious crypto in the 90's, switched to everything in the open and use cypher instead. Cypher cannot be broken mechanically, though it can be compromised (e.g. the Midway affair in WWII).
In fact, Crypto did not work well even in WWII, Enigma anyone?
In the presence of near turing capable analytic engines almost no cryptography is uncrackable.
If you ever spend time in China ( I get to go one or two times a year - married a chinese gal), the level of deep package inspection the government there employs is something else. They leave my VPN connections alone as long as I do not do anything "naughty" (meaning I am trying to look up something their censorship system think is, not Porn or the like, not even politics mostly).
But as soon as I do - my VPN connection is choked off.
0kbps. Not cut, mind you, choked off.
Then I have to change the server and reconnect. And get choked off 3 minutes later again.
And I am not using ordinary or free VPN's either (nor am I one of the usual gibronis, I use near mil spec stuff!). Because of the high annoyance factor of this I plugged about any possible leak on my machine and they still get through all the ICE, somewhere...
They must have machines that eat 256 Bit keys for a light snack, at ISP level too, to monitor what is inside my VPN Tunnel. Scary. Expensive.
Back to music.
Deep Crypto is too scary. WAY TO SCARY.
Commoditised Isolation Solutions please.
As in "Fast(ish ~100MBPS), Low Jitter ( < 20pS would be nice), Cheap (say 5 USD for EIAJ Dual Mono)".
Ciao T
At 20 bits, you are on the verge of dynamic range covering fly-farts-at-20-feet to untolerable pain. Really, what more could we need?
Post a Followup:
FAQ |
Post a Message! |
Forgot Password? |
|
||||||||||||||
|
This post is made possible by the generous support of people like you and our sponsors: