Gustard U16 - the first USB Interface featuring ESS USB chip
Sep 19, 2019 at 12:33 AM Post #1,337 of 1,697
I am using a linear power supply I built myself. It is not the best I have, but this one can output 2A. My other so called better power supplies can only output 1A.

Switching between the internal and this OCXO external clocks, I feel that the external one sounds louder (more solid?), even though not that louder, but noticeable. The instruments in a small jazz band sound more distinct. Vocals are enjoyable too.

Once again, I would like to mention that the OCXO metal box is quite hot, I do not want to put my fingers on it for more than a few seconds. Should we use a heat sink around this metal box? Is this because it is described as a "Double Oven" OCXO in its Ebay AD?

This OCXO came as a package better than I expected. It contains two BNC to SMA adapters and the board has 4 metal standoffs already installed. It is not hard to install this OCXO in a metal case, I only need to do some drilling.


IMG_20190919_000824[1].jpg
 
Sep 19, 2019 at 2:42 AM Post #1,339 of 1,697
Should we use a heat sink around this metal box?

No, but put it/mount it into a box to further stabilize the temperature on it. The thermal impact of the ambient temperature is one thing that its electronics compensates for. The double oven is less sensitive but still benefits "no" or at least slow changes.

Some guys put neoprene isolation on the OCXO to acheive maximum stability. (The UCT is a clone of the Oscilloquartz OCXO (=Double oven)).
IMG_8600.jpg IMG_8601.jpg
 
Last edited:
Sep 19, 2019 at 8:10 AM Post #1,340 of 1,697
I still have occasional dropouts of signal from U16 after the installation of the 1.61_nfb firmware. On all outputs - I2S, SPDIF and AES/EBU. On higher rate files the dropouts are more frequent. The Gustard is fed from WB external HDD to Bryston BDP-1 to SOtM tX-USBUltra. Singxer SU-1 behaves exemplary in the same configuration. The problem is that the Gustard sounds better. I tried several USB cables, to no avail. I do not want to go back to 1.4 firmware as it sounds worse and cannot be used simultaneously for AES/EBU and I2S outputs. It is either or. Any suggestions on what I am doing wrong? Thanks.
 
Sep 19, 2019 at 9:35 AM Post #1,341 of 1,697
I never consider heat in such a beneficial way. This is educational, thanks for explanation.

No, but put it/mount it into a box to further stabilize the temperature on it. The thermal impact of the ambient temperature is one thing that its electronics compensates for. The double oven is less sensitive but still benefits "no" or at least slow changes.

Some guys put neoprene isolation on the OCXO to acheive maximum stability. (The UCT is a clone of the Oscilloquartz OCXO (=Double oven)).
 
Sep 23, 2019 at 11:30 AM Post #1,346 of 1,697
I don't know if I'm doing it right or if I'm just a lucky person who never had any issues with the U16...I've it connected to a Mac Mini Server via an AudioQuest Carbon with ifi ipurifier and wyred 4 sound usb reclocker. The U16 is connected to the C16 external clock via a 50 ohm BNC cable. Firmware version is 1.61 (sounds better than 1.4 & 1.71 and is more stable than 1.61 NFB on my system), I2S configuration is for HoloSpring DAC. I've never experienced any pops, clicks, dropouts, restarts at ANY sampling rate - using it as a Roon endpoint and I've used all possible sampling rates from 44.1 KHz to 705 kHz PCM and from DSD64 to DSD512 without any issues. The only thing I've noticed is that it may restart by itself at the beginning of a song after being turned on for over 72 hrs without playing anything. So I'm not sure if the USB line hygiene I've implemented helps in maintaining the U16's proper functionality (which i personally doubt) or if it's just luck (which I believe).
 
Sep 23, 2019 at 12:11 PM Post #1,347 of 1,697
My theory, which is probably total hogwash, the U16 does not have enough buffer memory, poorly implemented buffer memory, or not optimized Windows driver. In Windows I have better performance with the ASIO driver where I can set butter sizes using the Thesycon applet to change buffer size. Have to use ASIO driver in Windows which is not dedicated playback. But I get better operational performance with ASIO. Using WASAPI the dropouts at higher sampling rates return.

The U16 may work great paired with the right equipment and set to a certain mode of operation. Just seems one size does not fit all. The Singxers (all three - F1, SU-1, SU-6) work better in all my systems.
 
Sep 23, 2019 at 12:32 PM Post #1,349 of 1,697
Seems like 128MB is a healthy amount of buffer memory. With the Aurender I can hear tiny ticks (small dropouts I guess) at the start of 96K tracks. This stops after a few seconds until the next song is cued up for play (back to back). Seems the buffer memory doesn't have enough data at the start of tracks. 176K and above is unusable with the U16/ Aurender. Again the Singxers have no issues. I even pulled out the old Off Ramp 5 DDC and it works better operationally versus the U16 with the Aurender.

DSD64 works great though with the U16 and Aurender. Solid performance.
 
Sep 24, 2019 at 12:01 PM Post #1,350 of 1,697
We should not use a qualifier like 'only' for a ring buffer of size 128 MB - that's about 47 seconds of 32 bit 352.8 kHz data!
We need to understand that USB receivers like these using ESS / Amaenero / XMOS use USB 2 Isochronous interrupts of 8kHz. So ideally you might want to keep driver buffers as high you might want to, the optimal results are obtained only when 8000 samples are tranferred per interrupt request.
This is an issue for redbook CD as 44100 is not a multiple of 8. For anything higher or mutiples of 48 this is alleviated.
Further for computer systems, you want to align the samples in block sizes which are multiples of 64 in memory. Which means you will ideally get optimal performance with highest buffer size polled at 8192 samples per transfer. This translates to 0.0232 seconds for 352.8 k and 0.1857 seconds for redbook.
If you know how to configure ALSA or Kernel streaming with these parameters, you can extract a lot of performance from USB receivers.
 

Users who are viewing this thread

Back
Top