Hi Robert:
Your enthusiasm in this thread and elsewhere is wonderful--and infectious.
I mean no offense, but I've been watching this thread for some time, and that enthusiasm and your posts often run well ahead of the facts. I really would not know where to begin in correcting some of what has been written here.
There is a LOT of misunderstanding about audio-over-IP for sure, and if eliminating USB and putting Ethernet into DACs was so simple you would have seen much more of it long ago. I could write a lot about that, but mainly people need to think about the OS and player side, as lack of truly free and open "virtual sound card" drivers--which should just be built into the OS--to allow Ethernet ports look to apps look like a sound card is what keeps the requirement for end-point devices to be full 'renderers', which aside from potential convenience and galvanic isolation does not otherwise change the world.
But today I can at least correct what you wrote above: The ICRON units--and all USB and Ethernet connection products--absolutely have oscillator clocks! All processor chips need them, as do all PHY interfaces (USB, Ethernet, etc.). For the Icron units they are 24Mhz and 25Mhz, the latter of which gets used at both 25Mhz as well as multiplied to 100Mhz (for FPGA use).
The other misconception here is with regards isochronous versus async transmission.
In the USB
audio specs (UAC1 and UAC2) ALL audio is isochronous, isochronous means "constant time", the host reserves time on the bus for the stream so no other device can use it, the stream is guaranteed its slice of bandwidth. So both of the common protocols used for audio--adaptive and asynchronous--are isochronous, the stream from the computer is EXACTLY the same in both cases. The difference is that asynchronous protocol adds a feedback path from the DAC
to the computer than can be used to tell the computer to speed up or slow down. The samples that come from the computer go into a FIFO as they come in, the local (presumably low jitter
) clock is used to take them out. The DAC
keeps tabs on how full the FIFO is and tells the computer to slow down or speed up if the FIFO is getting too full or too empty. The feedback happens every so often as needed to keep the computer in step with the local clock. Adaptive protocol does not have the feedback, so the clock in the DAC
has to adjust itself to match the rate coming over the wire.
For the ICRON devices, only the faster Gigabit LAN versions have enough speed and buffer to handle an isochronous stream. There are other aspects of Icron's ExtremeUSB and Simultaneous User tech which are quite interesting, but I am under a non-disclosure agreement with Icron and thus not sure what I can or can not reveal.
While we are more interested in what could be done with their SO-DIMM modules, if I was PS Audio, one of the first areas I'd look to for customizing/improving the REX end of the Icron RG2304GE-LAN boards would be the power supply, as stock it just uses DC-DC switching regulators to take that 24V down to 5V, 3.3V, and likely 1.2V for the core.
I'd say something about improving the REX's USB output signal integrity with some popular devices, but then the mods would probably delete this post.
BTW, the PS for the REX end does not have to be 24V. The spec sheet says 10-24V is fine. Just figure that a 12V supply with it ought to be capable of at about 2A. They provide a 24V/1A supply just because it is smaller and cheaper--and because their device's internal switching regulator can handle that.
Have a great weekend all!
--Alex C.
UpTone Audio