I’d love to see measurements from DAC output showing the difference.
It won’t make anything “easier”, as signal transitions are either translated correctly or incorrectly. In case of incorrect translation, you get error counters incremented and you can see. But so far, error counters in my switches are showing 0 errors despite having many terabytes worth of data having been transferred over.
I don’t use any audiophile networking gear, I trust Cisco and Hewlett-Packard Enterprise more on networking gear.
I am unsure myself why it matters, but I am equally sure that it does matter. Whether it shows up on measurement is another issue, but it definitely makes a difference to for example sound stage and clarity.
For example, it’s not at all hard to hear the difference between a Cisco network switch and etherRegen from UpTone (I have tested that one myself). And the 25Mhz clock upgrade to Crystek in the streamer did even more difference.
etherRegen is limited to 100 Mbps? While you will need 1 Gbps copper or preferably 10 Gbps optical connection for optimal performance. If you use 10 Gbps ethernet, you will get extreme tightly accurate clocks.
You also have to make sure you are not using shielded ethernet cables as these will spoil galvanic isolation.
The clock is for the USB protocol, and has nothing directly to do with the actual clocking of audio (which is done in the DAC anyway for async USB). Its the same with the Innous USB device btw, it clocks the USB protocol communication and not the actual audio.
As I mentioned, I am not sure why this matters, but you don’t need golden ear or high end equipment to hear that it does matter.
Here is the specification for the XMOS XU208 USB audio interface. You can see that it has it has its own buffers and gets its clock signal from surrounding circuitry, so just about the very first thing it does is “reclock” the incoming data.
It’s the clock that connects to the XMOS (or another USB interface) that’s important, and upgrading the USB clock on a USB streamer, or using a re-clocker like Innous USB, will improve this clock.
And as usual in audio, it’s not the long-term accuracy of the clock that matters, but its phase noise.
Here is a picture I took when I switched the rather bad clock in my microRendu to a Crystek. The bottom chip is the clock, the top chip is the USB interface. The difference in sound was not subtle, instead, it was much bigger than I anticipated.
I just can’t understand how that can be the case. The XMOS has its own clock input and buffer. It doesn’t take a clock signal from the USB cable. So how can the clock at the sending end of the USB - ie the one that’s in the Phoenix “reclocker”, have any effect?
All digital transmissions need a clock (USB, Ethernet etc), and a better clock leads to a more perfect signal (more square-shaped, less rounded, etc). When I started my HiFi journey I was sure things like this would not matter, as long as it’s bit-perfect. But once you learn to compensate for bias (for me its listening at least a couple of days), you can trust your ears and since in this case, the effect is not at all subtle it’s hard to disregard just because I don’t understand exactly why.
I think that a better signal leads to less “work” for the receiver, which means less internal noise. And of course, the less noise on the input line the better. This leads to these observations:
A “better” digital signal leads to better sound
The sender of a digital signal should remove as much electronic noise as possible (noise has many negative effects, for example it lowers the accuracy of the oscillator).
Can you point us to peer-reviewed publications in digital electronics that support this claim?
I could understand a claim that poor quality USB receivers are corrupting the incoming data stream if its clocking is below spec, but that’s not a question of more or less “work.” I could also understand that the USB receiver circuit might be poorly laid out in a way that the incoming signal would affect the timing of the outgoing synchronous (often I2S) signal to the actual DAC circuitry. In either case, reclocking of the incoming signal could conceivably reduce those effects. But the real problem in either case would be a poorly designed or implemented USB receiver. Which would not surprise me too much, even in expensive DACs
I base my findings on my own personal experience, maybe it can show up in measurements as well (it certainly can show up on digital measurements) but that is beside the point. 3D sense or width of soundstage don’t show up on measurements either, but you can certainly hear them (and yes, soundstage is not only dependent on room and acoustics).
XMOS or other USB Audio Class receiver will have two clocks. The audio clock for I2S, sourced from the D/A conversion stage. And USB/MCU clock. It could even have three clocks, with a separate clock for the MCU (processor).
Like in a computer you have separate clocks for CPU and USB.
No it won’t, the other end of USB will have it’s own clock. And this clock is not related to audio clocks, it is only used to transfer USB data packets between two RAM buffers. Then audio clock will fetch data out of another RAM buffer that contains audio samples unpacked from USB data packets by the MCU.
USB interfaces don’t have two clocks, there might be one later though for audio but for example a USB interface in a streamer has one clock (and there will also be a 25Mhz clock for the Ethernet receiver). I know very well how async USB works on a DAC, and that in theory, the USB clock should not have any direct influence over the audio clock.
What I am saying though is that better clocks for USB interfaces (and ethernet interfaces) have an impact on sound quality. For example, the Innous USB mentioned in this thread makes a big difference to most DACs from what I have read, and definitely makes a difference on the RME ADI-2 DAC I own. Having said that, I don’t own the Innous USB reclocker myself but I have heard it “in action”.
This has been known to listeners for years btw, ever since the first USB decrapifier was released. Yes, even cheap ones make a difference to most DACs, but shows little or no difference in measurements.
But feel free to check the 24Mhz USB clock in the DACs you own, I am willing to bet that they are all of high quality and with little phase noise (Crystek, NDKs best, Accusilicon, etc). Then think about why DAC manufacturers would use more expensive clocks for USB if it in fact didn’t make any difference.
Heard it in action in a different system or your own?
I assume you are talking about the Phoenix Reclocker.
I think there is a mechanism for this to make a system sound “different” - you are adding another AC power supply to your system which is another source of mains borne RF added to your system.
More mains RF is likely to have more of an impact than USB re-clocking.
RME ADI-2’s USB input is not galvanic isolated.
Intona isolator company’s website shows (measurements!) the effect their isolator has and the DAC used to show this is… an RME ADI-2…
So the ADI-2 is objectively shown to be a little bit sensitive.
One may question if it audible but the measurements show a different at least.
At the analogue output of ADI-2. We don’t have such measurements with the USB reclocker.
If Phoenix also had built in USB ground isolation that would be more interesting. Still better to buy latest Intona model in my opinion, or use a 5Vdc battery powered NAA like RPi4. No connection to AC power on the DAC’s USB input.
You use a lightweight NAA normally or direct USB connection to HQPlayer server?
The best example to use is the microRendu streamer and the clock I upgraded, which I still use in my system (with LiFePo4 batteries it sound very good). Previously that clock-upgrade was available to buy from Sonore (v 1.4) as an upgrade, but they stopped probably because the difference between upgraded microRendu and ultraRendu became almost nonexistent.
The difference with the new clock was not subtle, more improvement than any filter in HQPlayer does for example.
I know this will not convince people that puts an = between measurements and sound quality, but the best advice I can give is try yourself and listen for a few days to get past bias.
The more days (minutes) that pass, the more unreliable our memory becomes.
Proper way to overcome bias is have both v1.3 and v1.4 together and be able to switch between the 2 and reliably pick which is which, while blind.
Anything less and you haven’t eliminated bias at all.
Actually to say that you need more time with the upgraded 1.4 to get past bias is completely backwards. The more time that passes, the closer you get to pure raw bias.
Everyone’s favourite Amir was not convinced it does anything for the RME at all, but it did clean up the Schitt Modi, so it does work on DACs that have issues it seems. But whether that’s actually audible is another thing.
I have the RME it works perfectly without any messing about with anything like isolators or power.
