XMOS-based Asynchronous USB to I2S interface

unusual distortion --Still troubleshooting !

I match it with the discrete dac from diyparadise malaysia.
It works but there is a constant low level distortion with soft music . the distortion is absent in between tracks or on pause.
I am using window 7 in a fanless ATOM based pc and Foobar .
The digital isolator is powered at 3.5 V via a salas shunt regulator that was proven to work well in another Tenor based converter.
Any idea what the cause ?
thanks

kp93300

I build a linear trans based PS for Wave io and down regulated to 4.95 V with LM 317.
Problem persisted .

The distortion noise is heard in the mid range of the speaker only and sounds a bit like tape hiss with Dolby noise reduction off.. It is not heard when music is paused , stop or between tracks. It is heard when the music is of low volume.It is heard in both channels with equal loudness and seems to be independent of loudness control. This appear constant in loudness.

The wave io isolated i2s signal has individual grounds for the data, clock and word select lines. The discrete dac i2s input has only one ground.. Could this be incompatibilty issue?

I hope to get some suggestions from other users .I have tried at the dac forum without any success.

My set up is attached for review.
Link to the discrete dac :
I Found Miss Right!

thanks

kp93300
 

Attachments

  • wave monica .JPG
    wave monica .JPG
    136.8 KB · Views: 658
  • monica .JPG
    monica .JPG
    107.6 KB · Views: 644
  • wave i2s gd.JPG
    wave i2s gd.JPG
    107.2 KB · Views: 636
@ kp93300: you have PM.
There's other thing that I want to say about your setup: I2S lines seems to be longer than expected.
m.massimo was the latest person I know that runs into the same issue: too long I2S lines. Please make them shorter if you can.
In addition to that, I notice that there's only one trafo+diode bridge+caps bundle for two PSUs: one for WaveIO (LM317) while the other is for Salas. Are the grounds between these two PSUs connected together (WaveIO & Salas)? If yes, the you do not have any isolation at all and the noise from your PC/MAC source is getting into your DAC through this common ground connection (you have bypassed the isolator this way). Not to mention the ground loop. Just make sure to keep WaveIO's supply only for WaveIO while the output side of NVE's isolator be powered from different PSU (with dedicated trafo+diode bridge+filtering) or, at least, from the digital side of your DAC board (if possible - still not recommended!).
Hope it helps,
Lucian
 
Last edited:
Yup...

The first thing I would do is to re do the I2S lines. Use tightly twisted pairs, and get them all the same length (as short as possible, with your layout it looks like you only need 3 cm of wiring or so). Use 24 or 28 awg solid core wire, and run all of the grounds. Then, at your DAC board, run an exposed wire from the ground connection along the + input points, and attatch all the ground wires right next to where the positive leads terminate: this way you keep the tightly twisted pairs together for as much of their length as possible.
 
In addition to being able to withstand some noise, digital also creates a lot of noise on the supply lines due to basically passing around square waves all the time.
(...)
I'm sure you're right that clock jitter may need filtered supplies, but I still wouldn't feed a clock supply to an analog circuit, so separate is good.

Separating analog and digital supplies is obvious. But among the digital lines, how much should we separate, in a Sota setup?
Xmos for once, then the oscillators, and then the i2s output (isolated or not)

but what about the ground?
 
I wonder if you could describe the 3.5V salas shunt and its relationship to the DAC supply. Is there a possible ground loop condition between those two? Have you confirmed that the output of the salas reg is stable?

The salas shunt is powered from a 9V smps regulated supply. The 3,6 v from the salas shunt supply the IL715 only.
The wave io board derives the supply from a 1A trans and regulated with LM 317 that provided 4.95 V . I have not confirm the stability of the shunt but it sound nice powering an Tenor usb spdif converter.
There is only 1 ground connection from the salas shunt to IL715 and one ground connection from LM 317 to wave io.
thanks

kp93300
 
Follow up

Hi All,
Please see the attached image.
I did the changes for the ground arrangement and change the i2s lines length as suggested. Hope i have interpret correctly. I use some thin solid core magnet wire.
The problem persisted .

I did a few experiment of the different types of ground arragement and none of this helps.I left the isolated ground unconnected and wire the salas power supply to the star ground ie blue terminal block. Distortion still there and there is music.

thanks to all for the input.

I am entertaining the possibility that it may be the discrete dac problem.

regards

kp93300
 

Attachments

  • IMG_7329.JPG
    IMG_7329.JPG
    133 KB · Views: 573
no they cant; well not as loose cells as this is against the reseller agreement and they would lose their resellers rights, maybe old or rebadged, but certainly not new stock.

But put into motorbike starter package they can :)
Motorcycle Starter Battery - A123 ANR26650
How do the price looks (its 4x 3.3v cells inside)?

I need this to compare the output quality of the PSU(s) i'll build.
 
I have now tested thesycon DPC latency on five different platforms running XP. Two to three hundred micro-seconds is the range for any platform with a processor clock greater than one gigahertz.

It's not related to the cpu clock at all.
I have a 4ghz core i7 and I have 60ms average latency, and that is with the onboard audio active (and playing wasapi in foobar).
XP it's terrible for audio, unless u strip it naked like cics does for cPlay.
I cant do any of such since i use w7 x64 to work on this machine and it's not my dedicated music PC.
 
Indeed I would ideally use that solution also, havent thought up a suitably low noise way to drop the voltage within the tolerance level of the onboard regs yet. I did discuss some options with Lucian by email some time back and have ended up going for a LT1085 regulator (not yet finished building that though).
 
Last edited:
WaveIO is very good for me

I've been listening WaveIO for about 20hrs and I'm enjoying a great step forward compared with my previous setup. I used a Teralink X2 (adaptive mode with a single 12.000Mhz crystal), ASIO, I2S output into a stereo TP Opus 8740.
My actual WaveIO setup is: non-isolated I2S, usb through tybe B connector (same 3m usb cable as before), power by usb. Opus powering is the same.
The very first impression was a clear improvement in the high frequencies region. I always suffered from harsh highs, for example cymbals and female voice sibilants. Highs are more natural now.
Another improvement I noted was with hd tracks. It seems to me there's more quality difference between SD and HD now.
The last improvement was in the bass region and I started to note it in the last hours. Bass are firmer and present. For example, while listening some classical music (full orchestra) I feel an involvement with the music I didn't experience with the previous setup.
I'm very satisfied with WaveIO. I think it's a real bargain. Very good job, Lucian.
External power, isolated I2s and diy usb cable are on schedule.