Melodious MX-U8 USB Interface 8 Core XMOS chip

[rb2013]

 
I lost track of the post NK2520sSD vs crystek cchd 957, how is crystek better ?

The comparison of the NDK and Crystek.
 
 

The NDK ad on DIYinHK incorrectly states the NDK has "For easier comparsion(lower is better), we overlay the two phase noise graph at the same scale, under 25khz(human listenable range) nz2520sd(red line) is better, over 25khz(human unlistenable range) crystek 957 is better."
 
Unfortunately it seems these 'engineers' can't read a log scale.  The crossover in phase noise is at 2500Hz not 25000Hz.  2500Hz is definitely IN the human range of hearing. Ugg!
 
If you can't read a log chart I can't help you.

 

[rb2013]

   
Hi and sorry but are you referring to the stock unit ?
 
I have another general question.
At the output of the LT1963 i see a 220 uF cap ... could it be beneficial to put a bigger and better quality one ?
like a 1000 uF cap of the usual Nichicon FG type ? 

this mod i could do even myself

Thanks a lot and kind regards,  gino

It looks like Melodius is now shipping their units with the same value cap but with a higher quality one.  And heat sinks on the LDO and XMOS
From Sergey's stock unit:

 

[rb2013]

But speaking of NDK Crytals (which I do think are very good - certainly better then std XOs) DIYinHK is now using them on their USB to Spdif board:
DXIO PRO3A - (they have an i2s board available as well).
 
And ultra low noise regulator - not bad for $168.  I tried ordering one  but they are out of stock - they told me they had more coming.

 

 

1) Newest XMOS chip and uses 48MHz oscillator to asynchronous reclock usb audio data to SPDIF line, old XMOS uses only 13Mhz oscillator, over 4x better jitter rejection theoretically.
2) C0G AC coupling capacitor (All SPDIF output equipment requires AC coupling capacitor to avoid the isolated transformer saturated by DC bias component, if no grey color(C0G) capacitor is found in the PCB, that SPDIF equipment can be considered as low end)
3) Ultra low phase noise NDK NZ2520SD oscillator, thin film resistor and Murata isolated transformer (Be care the very common China made Pulse transformer in many 192khz SPDIF equipment is only capable of maximum 7Mbps(i.e. 96khz) according to the official datasheet.) Components are sourced from NDK and digikey directly (guarantee no fake product from China)
4) 1.0uVrms Ultralow noise linear power regulator (use of decade old LM317/1117 regulator is kidding for audiophile)
5) Solid ground plane (a must for high speed digital circuit)
6) No Via in active circuit (via inductance always create jitter problem)
7) Compact size 25mmx 59mm x 63mm
8) PCB can be USB powered or auto switch to external power 5-6V when connected(Please notes over-voltage can damage the component immediately, any modification void item warranty) External power socket has LED shows which power supply is in use. The external power socket uses the most common plugs size 5.5 mm (0.22 in) in outside diameter.

 
http://www.diyinhk.com/shop/audio-kits/54-xmos-192khz-high-quality-usb-to-spdif-with-ultralow-noise-1uv-regulator-wauto-power-switch.html

 

[ccschua]

We never compared Crystek's with NDK's, but, if reading my shootout between Hydra-Z and my modded MX-U8, they sound 100% identical,
maybe one could conclude there's not much difference between them SQ wise.

I am finalizing order at mouser. Can u advise what other parts i can get at mouser, e, g,

Nichicon fine gokd 16V 220uF x 4 ? Etc.

 

[ginetto61]

Originally Posted by rb2013 /img/forum/go_quote.gif
It looks like Melodius is now shipping their units with the same value cap but with a higher quality one.  And heat sinks on the LDO and XMOS

Thanks a lot indeed for the very helpful information.
I would like to see a version of the Melodious with external DC power supply like the Hydra Z.
I like the solution a lot indeed.
Regards, gino

 

[ccschua]

It looks like Melodius is now shipping their units with the same value cap but with a higher quality one.  And heat sinks on the LDO and XMOS
From Sergey's stock unit:

Agree. I open my unit and saw the better cap.
So the 4 blue caps are 2200uF 25V ?

 

[rb2013]

Agree. I open my unit and saw the better cap.
So the 4 blue caps are 2200uF 25V ?

Yes - does yours have the heat sinks?  If you could post a picture of your board that would be helpful.  Trying to stay on top of the design changes.

 

[ccschua]

I bought from ebay and paid with Canadian dollars. 

 

[ccschua]

   
Yes, stock unit has 1x LT1963 and 9x ADP150.
 
You can try a better quality cap, but DON'T CHANGE TO BIGGER VALUE, you could FRY LT1963, and it will definitely running hot!

 
my mistake, it should be 8 ADP.
 
I am finalizing the order. it should be 4 x 2200uF 16V nichicon fine gold. I cant go for 25V as it will be too high right ?

 

[abartels]

   
my mistake, it should be 8 ADP.
 
I am finalizing the order. it should be 4 x 2200uF 16V nichicon fine gold. I cant go for 25V as it will be too high right ?

 
Yes, right, 25V will be too high.

 

[abartels]

I am finalizing order at mouser. Can u advise what other parts i can get at mouser, e, g,

Nichicon fine gokd 16V 220uF x 4 ? Etc.

 
I also replaced the stock bridge rectifiers, but those are not easy to replace
 
CREE C3D02060

 

[ccschua]

I just realised that the C3D02060F is no longer possible as its full bridge now as shown in my picture. So I will drop the CREE and go ahead with 4x nichicon fine gold and 2 x crystek. makes sense (any other thing to add, as shpping to my country costs more than the components alone)? 

 
note : added back the CREE as backup plan. dint like the U8 low cost power plan.

 

[abartels]

 

I just realised that the C3D02060F is no longer possible as its full bridge now as shown in my picture. So I will drop the CREE and go ahead with 4x nichicon fine gold and 2 x crystek. makes sense (any other thing to add, as shpping to my country costs more than the components alone)? 

 
My MX-U8 had same diode bridge as yours,,,,,,,,,,,,,, it's just a hassle to get them in, i didn't use other components besides the FG's, CREE's, NDK's, seperate double psu 0.8uV from diyinhk, 2 R-Core transformers, copper shielding, high quality rhodium plated red copper IEC320 socket, and tellurium copper rhodium plated rca, but from mouser or digikey just the caps and the diodes

 

[ccschua]

Plan on hold. If I were to add your MTO, it will cost around a new Tanly unit. 
 
the NDK is supposed to have better phase noise below 25kHz. but its real challenge to solder that tiny XO in.
 
wait for some report on the DIYINHK usb ddc.
 
side note : Today I power up my PPA OCXO (motherboard ) and OCXO (USB 3) with twin tekdevice TPS7A4700 ULN LDO (2 unit in parallel) and it works. The music was more transparent. I strongly believe in ULN for digital. 
 
ps. : I have started to read back post 1. sorry for all this repeat questions.

 

[rb2013]

   
Try setting LOWEST buffer and latency in xmosusbaudiocplapp, mine works ONLY flawless if I set these values to the lowest.

A little info on buffer size and pops and clicks:

Setting the correct buffer size is crucial to achieving optimum performance from your audio interface: if it's too small you'll suffer audio clicks and pops, while if it's too large you'll encounter audible delays when performing in real time.

http://www.soundonsound.com/sos/jan05/articles/pcmusician.htm
 
So the issue of smaller buffer is greater chance of clicks and pops (not less or lower chance) - but increased 'latency' - which means the time delay in the Audio player to respond your command changes.  The buffer, in this case FIFO buffer, acts a store of the digital bit flow.  It's fed from the hard drive or SSD drive, which can have uneven data flow.  To the CPU processor that converts the WAV,FLAC,MP3 data to the audio output data stream fed out of the USB (in this case) or to the OS codec (if the OS is doing the digital to analog conversion).  The buffer helps smooth out the data flow to the CPU processor.  The bigger the buffer less chance the CPU will run into a data feed interruption during the music file playback.  The latency comes from the fact the buffer has to empty before the audio program can respond.
 
Just setting the record straight.