XMOS-based Asynchronous USB to I2S interface

You can't really set the current the device will draw what it needs.

wlowes, have you checked the voltage once the device is drawing current or only on an open circuit? What is the external supply you're trying to use?

Yes, the 4.9v was when it was drawing current. The USB cable was not attached, so I will try that and measure. I had it set up with aligator clips. Could be when its under load trying to associate with the PC it draws more and the contact was insufficient. I do keep the USB jumper set correctly. I am using a 100va transformer into a linear power supply with a simple 7805ct reg capable of 1.5A. I did have a discrete reg that could not supply enough power, hence the test with the 7805.
 
Hello,

I'm building a Buffalo III DAC an will use the Wave IO board to convert USB stream from my Vortexbox into I2S. The WaveIO board will be powered by a 5V PS regulated by a Bellison superregulator. Should'nt it be better to put the Wave IO board in a screened box in order to prevent interference with the DAC electronics? For connecting the board with the DAC I just got IPX/U.FL Antenna Cable from Lighthorse Technologies in San Diego.
@Lorien: news about my board?
Henri
 

Attachments

  • P1050378.jpg
    P1050378.jpg
    273.6 KB · Views: 884
  • P1050350.jpg
    P1050350.jpg
    277.8 KB · Views: 837
Thanks Lucian for the compliment. Good to hear the board is almost ready to ship.:)
BTW: What do you think about screening it? In a former life I was a radio Amateur and I know what interference can do. To be honest: I'm a little paranoid when it comes to screening and quality of power supplies.
Kind regards,

Henri
 
but what are you screening it with? unless its some sort of annealed permalloy i dont see it being all that useful past perhaps putting your paranoia at ease (at least until my post just now..sorry) better off just keeping the connections as short as possible IMO
 
ok qusp,
you are right. All connections should be as short as possible and in imho, boards that radiate harmonics (like all digital circuits do) should be isolated from each other. One can do that fi.(as you certainly know) by putting that circuit in a tin Can and feed it trough special feed-trough caps to block rubbish coming in or out via feed lines. But maybe I am indeed to paranoia about it.
 
qusp said:
but what are you screening it with? unless its some sort of annealed permalloy i dont see it being all that useful past perhaps putting your paranoia at ease (at least until my post just now..sorry) better off just keeping the connections as short as possible IMO
Qusp, it's DIY! Maybe WaveIO board doesn't offer too many alternatives when it comes to that but putting it in a screened box could be considered as a valid one. Regarding screening, IMHO a metal 'cage' is better than nothing :)
If this screening approach will be of any good for him, only henri47 will be capable to answer! So, henri47 go ahead and leave your 'paranoia' live... at least this is what I've done in WaveIO's case! I don't admit that I understand it completely but I can say that I'm pretty close :)
Regards,
L
 
I did not meant to start a new discussion about the usefulness of screening some circuits and isolate them from other prone to interference circuits. I simply asked if it would be beneficial to do so. Would sound quality be better? I honestly can't say; that's why I asked the forum for advice. But when I forsee trouble, I'll do everything I can to prevent it. In the end, all that matters is the resulting sound quality. I take that's the goal of every diy-selfer. Like Lorien, I also do not onderstand fully what's the influence of fi. HF radiation on sound quality. I only try to make things I build as good as possible.
Kind regards
 
I don't understand fully to what level the radiated EMF may impact nearby circuits either, but I think qusp made a good point that the material for the screening enclosure is critical to its effectiveness.

To my understanding it should ideally be a metallic enclosure with good conductivity, perhaps you or someone you know has trialled effectiveness of screening materials in amateur radio?

Let us know your findings/experiences!
 
well actually that was exactly my point, out of band HF or RFI will have very little effect on digital circuits and well designed analogue IMO, most of these circuits will have very good rejection at high frequencies both by themselves and because local decoupling caps and ferrites/inductors etc should do pretty well here, or much of it is simply out of band depending on the type of circuit, but at very low EMF frequencies, anything but an annealed box of much heavier stuff will do very little and its these frequencies that can and do cause jitter; its these frequencies where your average clock or even above average will have worse phase noise and the surrounding circuits and regulators will have lower PSRR; its these frequencies where caps have a VERY hard time. i meant only to encourage conversation and thought on the matter and also to state that any shielded box still leaves the lines in and out exposed and can only make it more difficult to have a short i2s connection, which is higher on the agenda IMO

lorien, i think you misunderstand me, of course he is free to do as he wishes and i encourage experimental diy, but i would hope he did so with eyes open, wouldnt you? hell i have my own idiosyncrasies and paranoia when it comes to audio, no argument there. what self-respecting audiophile doesnt? wasnt meant as an insult

Wolfsin: I called the main type of material by name already, you want some sort of annealed permalloy like mumetal, but if you bend it, or overheat it it has to be annealed again, not a bad reason for CNC. for HF shielding actually you want a fine mesh
 
Last edited:
out of band HF or RFI will have very little effect on digital circuits and well designed analogue IMO, most of these circuits will have very good rejection at high frequencies both by themselves

For digital yeah, but its really hard to design analog that has great rejection of RFI. In my experience anyway.

and because local decoupling caps and ferrites/inductors etc should do pretty well here, or much of it is simply out of band depending on the type of circuit, but at very low EMF frequencies, anything but an annealed box of much heavier stuff will do very little

Its relatively pointless to focus on the box while not addressing RF on the cables in and out of the box. Ferrites and inductors do indeed help, but they tend not to work at all well at higher impedances (say >2k ohms). So they're fairly good for power wires, much less good on signals.

and its these frequencies that can and do cause jitter; its these frequencies where your average clock or even above average will have worse phase noise and the surrounding circuits and regulators will have lower PSRR; its these frequencies where caps have a VERY hard time. i meant only to encourage conversation and thought on the matter and also to state that any shielded box still leaves the lines in and out exposed and can only make it more difficult to have a short i2s connection, which is higher on the agenda IMO

Yep, you're right to point out the wires. Conducted interference is the biggest bug-bear, not radiated.
 
agreed, but we were talking about the box =) i didnt bring the box up i only suggested a more useful box that still doesnt solve the problem. when LVDS becomes more standard sure, but until then Lorien already uses the best other method of connection, which still leaves it exposed, so the best plan is just to keep everything as short as possible to keep loop area small and isolated from ground right up to the XO if possible
 
ok,that's very usefull comment. qusp has a point: The box preferrably has to be Mumetal but I think the cheaper tin box will do also. It is commonly used when building Ham radio devices. Does it help to place a circuit in a tin box? In my experience I should say yes. A few decades ago I builded a HF receiver for 10, 20 and 40m amateur band. Boxing the oscillator, receiver circuit and antenna tuning circuit, in the same time adding feedtrough caps did make a world of difference. But I must admit not having the equipment to evaluate this. Even of more importance is blocking of hf or RFI on the cables connecting the box(es). Will boxing increase sound quality in case of the wave board? I really dont know, but keeping all cables as short as possble will shurely help avoiding induced interference.
An additional problem in case of the BuffIII and the WaveIO board is the heat generated by these circuits. Boxing them would shurely overheat them.
Reading all the comments It seems to me it's al in all better not to use a box to shield the circuit and go for the qusp-approach: keep wires as short as possible.
Many thanks guys

PS.: Sorry when something sounds strange or even offending in my comments. I do'nt mean to offend someone, but English is not my native Tongue and I know I not always express my toughts correctlly. Sorry for that.
 
WaveIO listening Impressions = Recommended

After getting to know my new WaveIO card for a couple of weeks, I am happy to say I am very happy with the product. I thought I'd map out my early experience for those considering this product, and offer some recommendations on how to get the most out of it.

My overall impression is that of a really good capacitor. It takes a little while to settle in, and it just gets out of the way. The real improvements then have to be made before and after the WaveIO, as it is not the weak link.

I bought it to power a very niceTDA1541a Dac that I built. It has 6n2p tube output, burson regs on the power supplies and lots of BG and Oscon caps in the power supply. It has DEM reclocking and SMT caps on all the decoupling pins. In short it is not perfect, but no slouch. I bought the WaveIO to replace a nice 2706 based converter done by Peter Daniel.. a hard act to follow.. My source is a highly modified CMP/Cplay PC with most of the most recent radical OS mods.
My 2706 rig had died. There was a considerable wait over Christmas and Jan for the WaveIO to arrive, so over xmas break I picked up a Rotel855 CDP to keep the music going. I made considerable mods (6n6p tube stage..lots of BG&Oscons) and all of a sudden this interim device was way better than the PC driven Dac.

When the WaveIO finally arrived, I plugged it in. The Dac had received a few new caps during the wait. I don't know if it was the WaveIO or the DAC that needed to break in. First impression was it was very clear clean sound, but compressed and lifeless sound that often comes with new Oscons. After about 20 hours it opened up considerably but still was not fully engaging. I compared it to the Rotel and it was clearly in second place. It lacked the organic sense of being there that the Rotel had captured.

I thought it might be suffering from being powered off the USB. I was having trouble getting a power supply with enough juice to do the trick (still not done).
I went back to my CMP rig to complete some optimizations. For those who have not experienced a fully optimized CMP rig, you need to give it a try with WaveIO.
The basic premise is to minimize the XP footprint to the extreme to reduce software induced jitter. I was already using the basic optimization plus pretty complete sliming of XP with deletion of dlls. I completed the file deletions with increase in clarity and lowered noise floor. But still no life. No real toe tap. There is a fairly radical step where you go in with Resource Hacker and remove a unneeded resources from DLLs and Drivers. With this change to the PC, the sound just exploded. Fully 3D. Very natural. Most important that uncanny sense of being in the same room with the performer. Feeling the emotion. Again I think that WaveIO simply is not the weak link. It now has overtaken the CDP reference.

My next step will be to get external power supply working. A cleaner power supply always makes a difference in digital. If nothing else just getting the load off the PC power distribution system has got to be significant.
Overall, I have to recommend the WaveIO product. Expect to need patience for delivery. If you can do that, its well worth the wait. Great design very well executed. I'll post additional impressions once I have cleaned up my install with external PS and shorter I2S leads on the the mBNC connectors.

Again, I compare this product to a really great capacitor at really reasonable price. To me like Russian PIO caps bypassed with Teflon or silver Mica. Dirt cheap but almost performs a VCap. WaveIO just gets the info into the DAC. Be prepared to tweek the PC and the DAC if you want to get the full potential.

Thanks Lucien,
Walter
 
--
I bought it to power a very niceTDA1541a Dac that I built. It has 6n2p tube output, burson regs on the power supplies and lots of BG and Oscon caps in the power supply. It has DEM reclocking and SMT caps on all the decoupling pins. In short it is not perfect, but no slouch. I bought the WaveIO to replace a nice 2706 based converter done by Peter Daniel..
Walter

Hi Walter,
so you managed to take off the I2S from the WAVE-IO board and feed that to your existing TDA1541 pins: BCK/SCK, DTA, LE directly. In fact, that would also work for TDA1543?
And that works just without any in-between resistive dividers, buffers, diodes, capacitors or what people have devised?
 
Hi Walter,
so you managed to take off the I2S from the WAVE-IO board and feed that to your existing TDA1541 pins: BCK/SCK, DTA, LE directly. In fact, that would also work for TDA1543?
And that works just without any in-between resistive dividers, buffers, diodes, capacitors or what people have devised?

Yes precisely! It would also work for TDA1543. In my case, I started with a Peter Daniel 1543 DAC. He does very nice minimalist designs using good parts and careful placement. His design takes I2S right from PCM2706. No resistors flip flops .. nada. See pic. He takes care to have very short signal paths. His clock is right on the pin of the 1543 and 2706 to 1543 is a few mm.

I was building a DAC starting with audio bd from an ARCAM Alpha. It used resistors to attenuate I2S. I tried taking I2S direct from Peter's 2706 to the 1541a, and presto worked like a charm. While both are good, I prefer the 1541a, and for the next 2 years, I used Peter's DAC to generate the I2S into my 1541a. Just used very short (3") teflon CAT5 to take the signal. Frankly, even with this length, the 1541a beat the 1543 design. The 6n2p tube output vs BG N caps on output might be a factor. Point is a few inches of CAT5 for I2S did not mess it up.

Now I have replaced the Peter Daniel converter with WaveIO. I am using 3 mBNC connectors that I took from digital cellular devices. These were put in long (6") to test and work fine. Soldered directly to the pins of the 1541a. I will trim these down to minimum length just on general principals.

I have no doubt that guys like ECDESIGN are building more resolved version of the 1541a. I do use DEM reclocking from his site but again, no flip flop. I take the WS signal at 44.1kHz. It is lower than optimal but light years ahead of the stock 470pF cap done spec'd by Phillips. I am going to play around with using the clock feeds from Lucien's board (88.2) to double the speed of the DEM signal without messing with more components. But that is in the future, as I am pretty happy with this initial setup.
 

Attachments

  • audiosector_usb_dac-I2S to 1543.jpg
    audiosector_usb_dac-I2S to 1543.jpg
    25 KB · Views: 785
Yes precisely! It would also work for TDA1543. In my case, I started with a Peter Daniel 1543 DAC. He does very nice minimalist designs using good parts and careful placement. His design takes I2S right from PCM2706. No resistors flip flops .. nada. See pic. He takes care to have very short signal paths. His clock is right on the pin of the 1543 and 2706 to 1543 is a few mm.

I was building a DAC starting with audio bd from an ARCAM Alpha. It used resistors to attenuate I2S. I tried taking I2S direct from Peter's 2706 to the 1541a, and presto worked like a charm. While both are good, I prefer the 1541a, and for the next 2 years, I used Peter's DAC to generate the I2S into my 1541a. Just used very short (3") teflon CAT5 to take the signal. Frankly, even with this length, the 1541a beat the 1543 design. The 6n2p tube output vs BG N caps on output might be a factor. Point is a few inches of CAT5 for I2S did not mess it up.

Now I have replaced the Peter Daniel converter with WaveIO. I am using 3 mBNC connectors that I took from digital cellular devices. These were put in long (6") to test and work fine. Soldered directly to the pins of the 1541a. I will trim these down to minimum length just on general principals.

I have no doubt that guys like ECDESIGN are building more resolved version of the 1541a. I do use DEM reclocking from his site but again, no flip flop. I take the WS signal at 44.1kHz. It is lower than optimal but light years ahead of the stock 470pF cap done spec'd by Phillips. I am going to play around with using the clock feeds from Lucien's board (88.2) to double the speed of the DEM signal without messing with more components. But that is in the future, as I am pretty happy with this initial setup.

which leads me to a plead to Lucien: couldn't you add a connector parallel to the output to your DA chip? [Like in the experimental XMOS board]. Would be happy :happy1:
albert
 
Few adjustments

qusp said:
lorien, i think you misunderstand me, of course he is free to do as he wishes and i encourage experimental diy, but i would hope he did so with eyes open, wouldnt you? hell i have my own idiosyncrasies and paranoia when it comes to audio, no argument there. what self-respecting audiophile doesnt? wasnt meant as an insult
None taken! It' my fault because I forgot to put the smile face at the end of my statement :) hence our misunderstanding. I totally agree with you!

@ wlowes: what can I say more besides the fact that I'm very pleased to see you happy with your WaveIO card. Please come back with your listening impression after you'll wire up your card to external PSU :cheers:

@ triode_al:
triode_al said:
which leads me to a plead to Lucien: couldn't you add a connector parallel to the output to your DA chip? [Like in the experimental XMOS board]. Would be happy
albert
Hello Albert, I can not make any changes to actual PCB and to look at least aesthetically pleasing to the eye. But, few DIY "adaptation" can be done to it:
if you want free holes into which you can solder wires and don't want to use the mBNC coaxial connectors then I could leave the NVE's isolator chip and 10-pin headers unmounted and with few wires your request can be fulfilled. The isolator chip and pinheader wil be at your disposal when want to reverse this operation. Please see the picture below for details:
 

Attachments

  • Isol.jpg
    Isol.jpg
    372.5 KB · Views: 742
Hi Lucien,
I'll be very happy with the isolated output. Isolating serves an important function. I thought the I2S output passed a D/A chip. :eek: - but it just passes through the isolator.

How about the NDK's you mentioned in the first post? That looks good too, and well worth it, the lower phase noise and lower total jitter (maybe 20 dB) does come with a price increase of course but probably well spent, considering the improvement gained.
And well worth 3 months waiting time.

So is this option still available you mentioned in post 1? And against which price?
regards
albert