Though you may have already found this link somewhere in the forum, how about this? (I have no experience on it.)
It is developed for USB 3.0. However, it seems to be OK for USB 2.0, high speed.
SuperSpeed USB 3.0 Over Fiber Optic Expansion System:Amazon:Electronics
By the way, I think a challenge for realization of 8 or more channels on DSD512 is valuable.
why not just use something like the i2s fifo? I dont see any advantage of this over that, particularly when he finishes the multichannel version. you would of course have to add upgraded non switching power supplies to the above isolator to realize any real advantage
No, I havent found it yet 🙂
Nice, its USB3! And costs much less than the Icron boxes.
The card is pci-e, which is needed for usb3 superspeed, which i dont need at the moment but I will eventually do in a couple years.
Did you try it? Or you want me to plug the plunge?
I'm still looking for something do DYI cheaper than this, and I can use usb2 still.
There are two different things... I want to isolate the computer USB from the digital interface. The i2s fifo should only reduce jitter (I didnt read the complete thread).
If Berkeley and Ayre spent a lot on research to isolate the usb input, I tend to believe that is the best approach.
If you remember the posts in the exa thread, the i2s isolators add about 100ps of jitter. THEN you would need a(nother?) FIFO buffer to counter that.
As I have written in my previous post, I have not tried yet.
My main field of interest is not USB DDC but SD memory card player. I am not sure whether I will try it or not.
At least, I do not want to connect the ground of PC to upper stream digital audio devices (A ground is a common source of electron flow).
The optical isolation method seems very attractive for me.
Oh sorry Bunpei, I missed that.
My interest is in computer audio, which i use in two systems (2ch one and 8way the other).
I have found some nice optocouplers rated 25Mbs, which seems pretty good and probably is what RayCtech used in his mods. 25Mbit limits it to 2x 352khz, which will do for me in the 2ch system (headphone).
I havent decided yet how to handle USB ground, but for sure I dont want it to go into the DDC. 🙂
My interest is in computer audio, which i use in two systems (2ch one and 8way the other).
I have found some nice optocouplers rated 25Mbs, which seems pretty good and probably is what RayCtech used in his mods. 25Mbit limits it to 2x 352khz, which will do for me in the 2ch system (headphone).
I havent decided yet how to handle USB ground, but for sure I dont want it to go into the DDC. 🙂
My speculation is rather on the use of a pulse transformer device popular in the field of giga-bit Ethernet physical link isolation. A USB dataline must be bi-directional. A pulse transformer can be bi-directional, can't it? One problem is the control of pull-up, pull-down control of D+, D- lines at an initial speed detection phase. I don't know how RayCtech made his work-around on the issue.
I'd agree.
1. A GB ethernet pulsetransformer on the datalines to attenuate common-mode noise.
2. Cutoff the 5V and shield
Q:
3. How to clean up the ground ?? I think there some discussions ongoing over here at DIYA. But I lost that thread.
The fiber solution looks interesting though. And It can easily replace an ethernet streaming client.
A guy called Ackcheng uses PCIe over fiber and is more than happy with it. He also had to replace the power supply on the receiving unit though.
I'd really like to see who that unit compares to the >400$ SOTM USB interface.
The FIFO solution mentioned is not a standalone solution. It needs something like the WaveIO interface. And that needs again USB.
The WaveIO seems to be a real good interface. However optimizations on the PC side still seem (from what I read) make it all the way through to the speakers.
No idea if the FIFO would solve that issue. Nobody answered that question in the FIFO thread.
What do we need then:
USB-fiber solution+Waveio+FIFO+I2S-DAC(synchronous clock)
Things are getting really complicated!!!
1. A GB ethernet pulsetransformer on the datalines to attenuate common-mode noise.
2. Cutoff the 5V and shield
Q:
3. How to clean up the ground ?? I think there some discussions ongoing over here at DIYA. But I lost that thread.
The fiber solution looks interesting though. And It can easily replace an ethernet streaming client.
A guy called Ackcheng uses PCIe over fiber and is more than happy with it. He also had to replace the power supply on the receiving unit though.
I'd really like to see who that unit compares to the >400$ SOTM USB interface.
The FIFO solution mentioned is not a standalone solution. It needs something like the WaveIO interface. And that needs again USB.
The WaveIO seems to be a real good interface. However optimizations on the PC side still seem (from what I read) make it all the way through to the speakers.
No idea if the FIFO would solve that issue. Nobody answered that question in the FIFO thread.
What do we need then:
USB-fiber solution+Waveio+FIFO+I2S-DAC(synchronous clock)
Things are getting really complicated!!!
I have a waveIO coming this way shortly, and i'll test how stable is on my platform.
I have no need for audio over ethernet, or any other proprietary solution. USB3 next year will cover all my needs.
The pulse transformer will only attenuate not isolate completely My buddy doesnt like them. Optocouplers/fiber optics seems the better solution to me.
Any DDC, whether it costs 100$ or 400$ will benefit from galvanic isolation.
I have no need for audio over ethernet, or any other proprietary solution. USB3 next year will cover all my needs.
The pulse transformer will only attenuate not isolate completely My buddy doesnt like them. Optocouplers/fiber optics seems the better solution to me.
Any DDC, whether it costs 100$ or 400$ will benefit from galvanic isolation.
I'm afraid you have mixed up the bit clock rate of I2S with USB high-speed data transfer rate. USB high-speed is 480Mbps though it does not straightly mean 480 MHz base-band modulation pulse signal.
I read RayCtech's post again as quoted below. He disclosed no details about his isolation method. No indication on optical or pulse transformer. However, I understand his report has been still asserting the effectiveness of employing a USB Isolation.
Today I studied the USB specifications and cut a USB cable in two and implemented galvanic isolation.....
As the 5 volt already was solved I started with ground as this is the easy part.
This worked at the very first attempt.
Then it was the D+ and D- signals..
At the first attempt the downstream functioned and the exaU2I received data and filled the FIFO at all rates,
but the upstream signals was not working properly and the ASIO driver failed.
I tweaked some of the component values and now the galvanic isolation of the USB works reliable, but some more work are needed to perfect the solution.
I have asked exa065 for his help as the component values on the exaU2I and the programming of the ASIO driver together with the component values of the galvanic isolation can for sure be optimized as my implementation are not fine tuned.
The galvanic isolation of the USB does not appear to affect jitter much or at all in my setup, and if it does I was at least not able to quantify it.
But it affects the audio quality in my setup.
The differences are easy to verify in all resolutions from 44.1k/16bit flac and up to 384k/32bit upsampled DXD...
For me there will never again be played music without galvanic isolation of the USB
Bunpei
No, I havent.
2x32x352.800 bits = 23Mbps circa < 25Mbs (Mega-baud per second, equivalent to bit in digital domain). The optocouplers that i have found are rated for 25Mbs.
It's you that talked about pulse transformers 🙂
no, you have misunderstood the purpose of the fifo, yes it reduces jitter to the level of whatever clock you are using by local buffering of the data, but it also completely decouples the MCLK from whatever you are feeding it from by replacing it with its own clock, rather than reclocking it and IMO the rest is not of huge/much/kinda any importance
if you still find that ground is causing issues after all that (only thing left), then I think you need a better designed dac, or perhaps you could try the SOTA USB card with a battery or very high quality linear supply, which is a better way to go about it IMO; tackle it at the source
Last edited:
Still the FIFO reclocker costs 200$ + the need of a dedicated psu. A pair of optocouplers to the USB pins costs <20$.
A separate psu for the USB controller is a thing i'm considering, but i have yet to see how to deal with the usb power and ground taken from the pci-e slot.
the fifo has pretty darn good onboard regs really, if concerned about ground just use LiFePO4 batteries to supply it.
I dont think we are talking about the same things with the PCI card, the SOTA USB card has ground lift and allows external power, which could also be batteries. ive linked this thing a few times already and people will be getting sus... I havent got any experience with it, if I had a mac with a PCI slot I would probably be considering it, no affiliation.
Actually I mispoke its SotM (sonore) and its TX-USB I guess its an issue many are struggling with, or its just one of the last areas to tackle, but it seems to be the solution to a number of thread topics ive seen lately so i've linked it. bit pricey IMO, but anything at the pointy end is I guess and you can bet they realize they are one of if not the only company making such a product.
I dont think we are talking about the same things with the PCI card, the SOTA USB card has ground lift and allows external power, which could also be batteries. ive linked this thing a few times already and people will be getting sus... I havent got any experience with it, if I had a mac with a PCI slot I would probably be considering it, no affiliation.
Actually I mispoke its SotM (sonore) and its TX-USB I guess its an issue many are struggling with, or its just one of the last areas to tackle, but it seems to be the solution to a number of thread topics ive seen lately so i've linked it. bit pricey IMO, but anything at the pointy end is I guess and you can bet they realize they are one of if not the only company making such a product.
I never had the need for isolation on USB. No noise/hum whatsoever.
Probably because I always use the PC/receiver with grounded outlets...
I am considering a normal pcie USB 3 controller, such as the buffalo, which costs 35€ and then work the psu. The sotm card is way overpriced, but they are releasing a pci express version soon. I want to have a dedicated USB controller anyway.
Sonic, I do not have any ground loop nor I hear noise, but I'm sure isolating the USB will improve the sound.
ok, well I guess i'll just leave you to it, personally the only reason I would consider it would be to satisfy my audio neurosis, I dont feel if a system is designed properly that its actually needed. just for completeness sake
Why are you so sure? It cannot do nothing good but add jitter. If anybody has issues with ground loops and hum, it's better to address that issue than trying to band-aid it with an insulator.
The only improvement that I can see to an USB connection is... not using it. SPDIF, i2s are better suited for digital audio. Heck, even firewire/IEEE 1394 is better!
The FIFO reclocker is also the only other improvement that I can see is needed for a digital audio connection. I just didn't go around building one... don't think that I have that much jitter anyway in my system to justify the $200.
Last edited:
- Status
- Not open for further replies.