Matrix X-SPDIF 2 Alternative - I2S vis LAN with the Mano Ultra MKII

Right now, I am using a Pi4 with the FifoPi and HDMI transmitter from Iancanada to try out direct i2s as it seemed like the most “direct” route. I also have a dedicated music PC running Windows and Jriver that I am using, but the only reasonable way to get i2s in LVDS out of that is via the Pink Faun Bridge, but it going Pi was cheaper. I would still try the Pink Faun if I come across a used one at a decent price.

The caveat to using the Pi, though, is I had to set the dtoverlay to hfiberry-dac to make it work with the Iancanada stuff. On top of that, unless there is some kind of dither between tracks, the Iancanada boards would stop between tracks when switching (it did not occur with gapless continuous tracks like found in Pink Floyd). It also only occured when streaming, not playing local files.

Because of this, it would drop the first few milliseconds at the beginning of any track when manually switching @ high res. For example, if I played Low Rider in 24/192, that very first hit of the cowbell would be dropped.

I tried all the usuals - Volumio, PiCorePlayer, etc., and they all did that with no way to control it. Using Jriver, however (I have the Master License to use across OSs), there is a setting called “play silence at startup for hardware synchronization.” Setting it to even the lowest setting (0.25 seconds) took care of that problem. I could not find any other OS/software that could do that. I keep it at 1 second just to be safe - hey, what is 1 second in life?

The Iancanada HDMI transmitter is set to the PS Audio standard. It DOES put out the 5v on pin 18 though. It wasn’t a problem, and even Ted said it shouldn’t be a problem, but just to make sure, Blue Jeans cable was able to make me a cable (using a high quality Belden cable) without pin 18 for $15 more than the base cost. They said they will make it for anyone who asks - just email them. I didn’t want to do it myself because I hate fidgeting with HDMI cable pinouts.


I guess it’s a YMMV situation. Some feel that the Bridge doesn’t sound as good, and some feel it sounds great. The people who feel the Bridge doesn’t sound as good tends to believe it’s the rf/emi noise caused by the Bridge being within the enclosure.

The USB input goes through an XMOS chip, which is then converted to i2s, so it’s doing the same thing as the Matrix, but it’s true that it’s an older chip, since the DS has been out since way before the current crop of XU208 and XU216 chips (though I’m not sure which chip off hand). The Matrix uses the XU208.

But some feel the Matrix sounds better because of better implementation (and more up to date chip). Ted stated that there is some benefit for the activities of the USB - i2s translation is better outside the box, again because of noise. But Ted did also state the caveat that more cores can also mean more noise without being better.

I can’t say for sure if one method is better than any other. MAYBE the Pi4 with Iancanada is a smidge cleaner than USB or Bridge, but there is so much to pychoacoustics that I cannot discount what might be happening in my brain versus what is ACTUALLY happening.

I have a track of “digital silence” in 16/44.1 (nothing but zeros) that I use to test for intrinsic noise (like from power supplies) because some dacs actually shut output off if there is no actual signal (ie. my comment above about dither). Playing the digital silence keeps the output circuits active so I can use my cans (Sennheiser and Audio Technica) at full volume to see if I can hear any low level background noise being caused by rf/emi/power supply noise/etc.

I have heard it occasionally with cheap dacs - almost sounds like a high pitched wine with Morse code sprinkled in - but I haven’t been able to hear anything under any implementation with the DS. Mind you, my ears are 60 years old, and I can barely hear a mosquito unless it’s hovering near my year. I used to be able to hear them across the room in my younger days…

1 Like

I forgot to mention. Using BubbleUpnp to stream from Qobuz to the Bridge does NOT result in the dropped first few milliseconds.

In fact, if you stream direct, I found the Bridge to be the most reliable/least problematic method (any “sound” preferences aside).

For example, when streaming from Qobuz to Pi4 to DS (via i2s), there is a slight click/pop when manually changing tracks, especially with different bit/sample rates. This does not occur when going from Qobuz to Pi4 to DS via USB, or Qobuz to PC to DS (via USB). There is no clicking, popping, dropped milliseconds when streaming to the Bridge - at least that I’ve come across.

If you stream direct, the Bridge is the easiest, problem free, method to do it (if you already have the Bridge - if you have to buy one, the calc is different). It also eliminates the need for any external component - Qobuz—>Bridge II (via controller such as BubbleUpnp). You don’t need a Pi, Mano, PC, or anything else.

In fact, even for local files, if it’s stored on a NAS, just use Bubble to stream NAS–>BridgeII. The caveat being if you like the UI of something like Roon, you can’t get that anywhere else as far as I know. I’m not a Roon user - I’m too cheap - but I know for a lot of people, the UI is where it’s at.

1 Like

The Bridge II does not sound good and is quite limited. The best thing I ever did to my DirectStream was to remove that card from it!!


I seem to recall something about Pi3 being better than Pi4 in this application. I wonder if this is part of your issue. Or I could have been dreaming…

1 Like

No, the issue with the Pi4 has to do with how the stream is handled between the Pi and the Iancanada FifoPi. I haven’t been able to test this out with other cards, since there’s only a few cards that actually take the GPIO output and translates it to i2s in LVDS. For example, the Pi2AES is capable, but the current iteration is not in the PS format (switched L/R), which is easy enough to compensate for, and they are supposedly fixing that in a new version coming out early this year. But @ $200+, I might as well buy a Pink Faun for my PC. Plus I’ve been able to fix that with the JRiver setting as noted above anyway.

Also note that I haven’t tried it, but you should be able to just run the HDMI Transmitter without the FifoPi since the DS does not really need the MCLK signal, so I haven’t tried it without the FifoPi to see if the behavior is the same. I got the FifoPi in case I decided to use it with the TransportPi to output S/PDIF via coax or toslink to another DAC in the future because the FifoPi allows you to swap clocks. Right now, I have the CCHD957 clocks in there.

I posted a couple of pics of the setup here:

The value of high $$$ clocks may be questionable for the DS, however, since as long as the jitter is decent, the DS has to reclock everything anyway as part of the conversion to DSD.

1 Like

+1. IMO, comparing I2S and USB signal paths into a DS Sr. DAC with a Bridge II installed is a flawed experiment. Only by removing the noisy Bridge II network card beforehand can one truly know how good an I2S (or USB) signal into the Sr. can sound.

Note: I do not have a dog in this hunt, as I have long since sold my Bridge II (first/separately) and then my DS Sr. (second).


1 Like

Ooh…want one of those…

Care to share some details…? That would be a handy test track.


What is happening is when manually switching tracks, you can see that the signal stops and restarts by the LEDs on the FifoPi. I have searched for why/remedies, and have tried various settings, including setting parameters in the dtoverlay, but the only thing that has worked is the above setting in JRiver.

That restart is what is dropping the first few milliseconds of the the track. Also note that it’s only noticeable on certain tracks. I suspect a lot of tracks actually start of with a few milliseconds of silence before any real audio signal, but with some tracks like Low Rider, the track begins with the first hit of the cowbell (which is why I used it for testing).

So I threw everything in a mini-itx case so it looks better - though I’m kind of liking the “raw” look of it all so maybe I’ll pull it out.

Also, the FifoPi REQUIRES a separate 3.3v supply since it’s built to isolate itself from the power in the Pi. In addition, you can add your own 5v supply to power the whole shebang through through the FifoPi. So what I am currently doing is powering everything from two separate supplies (one 5v and one 3.3v) to the FifoPi. The HDMI Transmitter gets it’s power from the FifoPi’s isolated power supply. If you use the Transmitter directly on the Pi, there is a method to use an isolated 3.3v power supply so as not to get the 3.3v from the Pi.

I have done comparative listening with and without the Bridge installed.

That’s cool …

(My post was not directed at you.)

I got it on GitHub. It’s in MP3 format, but you can use your software of perference to reformat to wave (or whatever you want). Even reformatted, a zero will still be zero for this purpose. As stated, I use it because some DACS actually switch off output if it sees no incoming signal.

Note - I don’t know if the DS does that. I assume it does, but the silent track keeps it “alive.”

I have the track saved, so will have to dig through my bookmarks to find the link. The list includes silent tracks of various lengths up to one hour (I think). It’s primarily used by people who make play lists that loop so they can insert silence where needed/wanted.

You don’t really have to convert it either, since the DS can play MP3 - I just converted it some time ago for my purposes. As far as I know, zeros in MP3 sound pretty much the same as zeros in other formats, and even if it doesn’t, as long as I don’t hear anything other than what I’m checking for, it really doesn’t matter.

I’ll post the link later.

1 Like

I2S, USB, OPT, BRIDGE in my testing.

Although I never got around to removing the card. Thanks for the reminder @bootzilla


Also note that if you go FifoPi, it comes with 2 OK clocks to test/make sure it’s working properly. To swap out the clocks, you have to either source clocks in full size 14 pin dip packages (note that there are only 4 pins that actually stick out), or have smd soldering skills enough to solder smd clocks to adapter boards (which Iancanada sells).

I haven’t been able to find good clocks in 14 pin dip packages lately - at least from sellers I trust. A lot of fleabay sellers are actually selling outright fakes of questionable quality or used clocks. I have even read of people buying dip package clocks on fleaby only to find out it was actually cheap clocks with stickers on them. I went with the CCHD clocks from Digikey and bought the adapter boards. The CCHD at least used to come in dip packages, but I don’t know if it’s a victim of supply chain issues, or they just don’t make them in that format anymore because there really isn’t a market for that package these days.

Here’s the link to the “files of silence”


Thanks a bunch.


I started with my Sonore and then tried the card. No comparison. Sold the card.

1 Like



Yes but that is not the same what we are talking about with etherRegen or any other good “audiophile switch” and then with better ocxo clock to fight against ground and phase noise,emi/rf in ethernet path. Same madness as cleaning usb signal. And differences are very easy to hear.
Ted doesn´t use clocks with digital inputs but does that mean that any signal you feed there is perfect then and sound the same? nope.
You may want to try some ethernet toys yourself too and get pleasantly surprised :slight_smile:

1 Like

I am still waiting for John Swenson to prove his switch does anything with phase noise and how does this phase noise affect the DirectStream when none of the clocking information is used?

I have listened to a setup with the EtherREGEN and heard no difference. In fact, the guy that had it was sure he could he a huge difference until I had him try to tell the difference between the EtherREGEN and a NetGear switch properly grounded. He picked the NetGear 6 times out of 10.

1 Like