Innuos Zenith Mk3 Ultra-low Noise USB Audio Output, Really?

Mega Byte?
I prefer Giga.

MB = Motherboard

Listen to it and see what you think.  Lots of people are thrilled with its sound.  Or just go buy the MB and LPS and build your own and see how that works out for you.  I can buy the same ingredients as Wolfgang Puck — doesn’t mean I can create what he can. 

USB is inherently a compromised interface. That said the Innuos version really is quite accomplished. It does though substantially benefit from galvanic isolation and superior cables.

USB is nothing but unclocked data. Computers do not pay attention to noise only ones and zeros. You guys probably don't remember the old analog cell phones. Needless to say, our modern digital ones were a breath of fresh air. It is all about the clocking and there are any number of great clocks in various formats to do the job. I have not ever had a jitter problem.

Motherboard, Mainboard, it’s 6 of one half a dozen of another to me - Dad joke?

@soix +1

 

How do we suppose these invisible ones and zeros magically pass along conductive materials to arrive perfectly arranged at the other device?
After the computer, which essentially all devices that unpack the music file information in the manner the Zenith does, processes the file into what will become the bit stream. A bit is like a placeholder for the one or zero value, high or low, on or off (if you want to read more, look up JK-flip flop, fascinating stuff).

So the computer makes electrical noise as it processes through these calculations , the amount of electrical noise that is inherited by the chain of events that transfer the information over an analog alternating electrical signal which has a specific pattern to be measured and read in by the end device, is generated by multiple factors.
Electrical noise that is introduced to the alternating electrical signal can effect the transfer of the data read in, the baud rate (the rate of switching change) does have a significant impact on the integrity of the data (encoded information in the series of ones and zeros), music playback by all accounts is low frequency.

Most current DACs do not have onboard computers, which could read in the packet (a specified number of zeros and ones) and determine that the checksum, the number of zeros and ones that are read in at a baud rate are all there, in which case it could request it to be resent, if there was something amiss.
Failing that, the DAC processes what it has, and just keeps on trucking along.
It can be measured with the right equipment, and in the case of audio playback, it can be heard, particularly if more accurate transfer is achieved.

And how does the other device know that it’s supposed to be ready to magically receive the perfect ones and zeros?
There is generally a 5VDC current transferred down one of the four USB conductors, which acts like a trigger for a handshake between the two devices.
Now I am aware that some clever designers have used a relay for the DAC which keeps the actual power inside the DAC isolated from the source. Engineers, such clever people.

But the electrical noise inherent in all things electrical and alternating, can be minimised (there he goes with British spelling again) with engineering measures in the transfer conductors (two of them, and it’s the measurable potential between these conductors that is read from).
Also, apparently Innuos have been making updates to the program, that have significantly lessened the electrical noise, and resulted in superior sounding music playback. I did mention the processes in the computer create electrical noise.
People are evidently impressed, that despite thinking they don’t have a jitter problem, have experienced less jitter problem that they never realised they had!

If all devices were perfect, they would all sound the same: perfect.

Reports are that the two software changes have lifted performance as though they had replaced it with a totally better source. If you have the entry price at your disposal, giving one a listen in your system would not at all be foolish, by all accounts I have read and discussions on the phone. I don't currently have the spare entry money, it's tied up elsewhere - priorities huh?

@martin-andersen This is exactly my issue with many off the shelf servers, motherboard is noisy environment. A top tier server will have usb coming off nice card like Pink Faun or JCAT, or perhaps some proprietary scheme.

 

I modded my server for second ethernet port, this port goes out to FMC, far superior to prior usb off motherboard into dac. Ethernet out was also superior prior to fiber conversion scheme I use now.

@sns I have the Jcat Femto USB card on my server.
I use a 300w HDPlex LPS, an 8watt embeded commercial grade fanless mainboard with a couple of SSD, it’s a few years old now. I’m sure there’s better solutions, seems you’ve found some already and put them to good use.

I run Winserver 2012 R2 in core mode with Audio Optimizer 3.0, Fidelizer Pro and JPlay Femto as well.
I think next time I will buy a good server, Zenith Mk 3 is absolutely on my radar for such an upgrade.

Oh what is FMC?

@rixthetrick You're present usb out is likely superior to Zenith usb port. The Innuos Phoenixusb exists because Innuos knows their Zen usb out can be improved upon.

 

Sounds to me like you have a pretty nice server already, you'd have to reach higher than Zen to beat what you already have. You have top line USB and optimized Windows set up, you need to set your sights higher to ensure you're making upgrade. Sounds like you have some form of atx motherboard server, atx motherboards are basis of some of the highest end off shelf and diy servers.

 

FMC is fiber media converter, TPLink MC200 is example of one, my Sonore Opticalrendu is optimized FMC.

A follow up.

I build a pc using HDPlex faneless and a Matrik Audio USB card with an external LSPU. Running Linux roon core

@martin-andersen This sounds very close to my planned diy music server. I'm mostly waiting to determine how much processing engine I'll need. If I decide on using HQPlayer with high levels of dsp will need much, if not this not much concern.

 

I don't like Roon dsp. Tried pcm to dsd 256  again last night, smoother than pcm, but that smoothing sounds artificial to me. Even worse that smoothing is in fact loss of resolution, the smoothing is loss of transients, also lose sense of recording venue. If HQPlayer anything like this, won't be for me. Reportedly, it plays at much higher level than Roon dsp, we shall see.

@sns
@martin-andersen
I went a different path, instead of higher processing power, I went lower wattage.
My thinking is that by lowering the power needed the lessening of the electrical jitter, and I also went commercial grade low power fanless CPU.
ASRock N3150DC-ITX - up to 2.08 GHz dual core.
There’s probably better out there to lessen the electrical noise, but I went lower wattage not higher processing speeds. 2GHz is already enough to unpack music at high res DSD.

Fidelizer Pro assigns a CPU and RAM for expressly computing the music, the other is for the operating system, well that’s how I understand it.

I am very much open to experimenting with Linux as my OS, I use AO to turn off a lot of superfluous Windows threads, which is another way. Wouldn’t it be great if there were an AMD or Intel music player that is the OS, the core of the machine language not simply layered on top of it?

HQPlayer is great, there’s also another player from Japan Bug head emperor player, which is fantastic.

I would like to know what is best? better? a way to move forward?
Do I just buy a very expensive server, or find ways to improve mine?

I had a bump in performance when I went to the HDPlex linear power supply, but not as much as when the DAC and the HDPlex were fed power from my Puritan PSM-156 power conditioner.

I don’t know where you guys are, if it’s feasible or not, last year I went from Texas to Washington state to listen to some gear with an invitation. I’d really like to know what works, that isn’t just insanely priced? Can audiophiles willing to do the leg work, build devices as servers that actually compete with the big names?
I really do think that clean power, better software controlling the hardware and good use of current or forward moving technologies are an answer.

I had the Innuos PhoenixUSB reclocker with the Zen MKIII and judging from memory my home build server is as good or better.

And the cost is only half. Also no SSD only M.2

https://www.matrix-digi.com/en/products/333.html

400W NanoATX Combo

@rixthetrick Audiophilestyle forum is my go to for all things in streaming realm. Guys there have much experience in everything from beginning to sota streaming. I'm using them as resource for upcoming diy server build, one of the builds over there trying to mimic Taiko Extreme.

 

Yes, two courses on processing front, lowest would theoretically inject less noise into motherboard, but negate ability to use dsp. High level processing more noise,  but dsp capable. I'm not sure which way to go at this point, depends if I care for dsp. I'll soon experience high level HQPlayer processing on a diy high end server build on another guy's system, hopefully, I'll be able to try in my system. This will inform as to which direction I go on my own build. Thus far I don't like Roon dsp,

 

I don't doubt you can improve on your setup, but how and at what cost? I'd say in general terms we can't diy servers at Taiko Extreme or Wadax level, but we can do better than vast majority of off the shelf servers. The better diy servers I see over on Audiophilestyle, which may cost in vicinity of $5k-$7k should compete with much higher cost off the shelf. Austinpop over at that forum has been mentor to me, started at my level of server, now owns Taiko Extreme. He made incremental moves with diy, finally went all out with Taiko Extreme, too rich for me! '

My plan, depending on processing power is diy in realm of $5k max. Per your statement above, forward moving technology may make any present server at least somewhat obsolete in future, just depends how long in future.