Computer based audio and I2S?

You got a typo in there? DSD64 as DoP requires 176.4kHz 24-bit PCM. DSD128, being double the amount of data per second, requires 352.8kHz 24-bit PCM to be carried as DoP.

If you’re using a 192kHz PCM rate, you are not sending DoP.

Hi @routlaw . You’ve asked a fascinating original question that’s incredibly challenging to answer succinctly. I can’t do it. So you get a wall of text instead. :smile:

Any numeric value can be represented in an infinite number of equally-valid ways, and transformed an infinite number of times from one form to another, without damaging the information or losing any sense of it. We are able to choose the representation that’s most suited to our purposes – both in terms of its abstract encoding (eg 24-bit PCM data represented as two’s complement binary integers or as a single-precision IEEE754 floating-point value, occupying 32 bits of storage as different binary patterns but meaning exactly the same thing), and in terms of its physical manifestation (eg charge states of transistors in RAM, magnetic dipoles on a hard disk platter, the presence or absence of a physical divot on the metallic disc of a Compact Disc, or hand-written pencil marks on a piece of paper).

All of those things are ways of storing bit patterns so that they can hopefully be retrieved later, provided you understand the scheme that was used to encode the information and have the physical means to extract it. The thing about extraction is that (almost always) involves reading bits one after the other, in sequence. There’s a time component to it.

I2S is just a scheme designed and optimised for transferring stereo PCM audio data (literally just a stream of numbers, most often but not always in two’s complement integer form) from one section of a circuit to another in real time using a single serial data line and two clock lines to delineate the timing of each individual bit and the start of each alternating left-channel and right-channel sample. Electronically the expectation for I2S was low voltages and short distances, but there’s nothing about the information content that precludes the same scheme from being used over longer distances or at higher voltages if the electronic engineering works out OK. Turns out that the HDMI connector and its differential signalling does in fact work great in that role.

The physical storage of your audio changes countless times during its production, distribution and playback. So does the abstract representation of that audio in binary form. Provided that we do it losslessly in the appropriate places, meaning that the numeric value going in to the system is identical to the numeric value that comes out the other end, it is utterly pointless to care about what happens in between.

So why do we care about USB vs I2S in that last stage of connecting to the DS DAC? It’s because of physical, or more correctly electrical matters arising from how the data is being transferred in that exact moment. Specific to the DS DAC, we can say that I2S has two advantages over USB:

  1. By having the incredibly complex USB receiver chip sit idle, we avoid generating electrical noise inside the DAC which impacts the resulting analog output
  2. The HDMI cable is really well grounded and otherwise resistant to picking up electromagnetic noise from the environment nearby the DAC, again avoiding contamination of the analog output

You should also know that if you are using the USB input on the DAC, the entire job of that USB receiver chip is to extract the audio data from the USB data packets, then steadily ship those over to the FPGA… using one of four I2S buses present on the DirectStream DAC’s digital board. (The 4th one goes to the Bridge.)

No mistake, I recommend that you google for the dCS DOP spec and read that document. It works a I described.

What, like this?

The most formal version of the spec is published here:

And I did learn something new from that spec. It also supports cramming stereo DSD128 into quad-channel 176.4kHz PCM packing. That would be quite doable over USB, so maybe that’s what your system is up to. 192kHz makes no sense to me though – it’s a multiple of 48 whereas all DSD is a multiple of 44.1.

Yes, indeed that’s where I got it from.

Bit Perfect on my Mac packs it into 24 bit 192 kHz DoP which the Stellar Gain Cell DAC and assumably can perfectly read. It’s really good.

Hope it works on your system as well, good luck and happy listening.

Oh, I think I understand your confusion now.

BitPerfect is very clever software. I used it for a few years before I switched from having my music library on a Mac to running a Roon core on my NAS. BitPerfect was designed to let you use iTunes as the player interface while BitPerfect took care of actually moving data from your computer to your DAC.

iTunes has no support for DSD, at all. But the file format (not audio encoding format, the actual file format for the thing you move around in Finder etc) that iTunes uses for Apple Lossless tracks is an extensible container which can hold more than just what iTunes is expecting to see. So the BitPerfect team created a second app called “DSD Master” which reads DSD audio files (.dsf or .dff?), creates an Apple Lossless PCM version of the track which can actually be played by iTunes and synced to your iPod or iPhone etc, then also crams the original unmodified DSD audio into a separate section of the same file on disk. So one file contains both PCM and DSD versions of the track, but iTunes only sees the PCM version.

When you play such a file with BitPerfect enabled, BitPerfect reads the DSD data instead of the PCM data, packs it into DoP format (PCM 352.8kHz equivalent if you have DSD128 source material) and sends that off to your USB DAC. iTunes is completely unaware of the DSD stuff – it only knows about the PCM content which could be rendered at any rate DSD Master wanted to use, including 192.

1 Like

An interesting bit of wizardry. Thanks.

Yep, except that for DSD 138 it does DoP over 24 bit 192 kHz. Really.

1 Like

This is all very interesting information… sort of. But does anyone here think that high fidelity audio in the 21st Century has become way too complex? Is it any wonder why the vast majority of music lovers, not audiophiles per se, stick a pair of ear buds in and listen to music via their mobile devices. Home Theater shenanigans are bad enough, but two channel getting this complex!

This is something I have been cogitating on for a long period of time, going back to spinning disc vs fooling around with the computers and their associated never ending obsolescence with forced upgrades. I’m dead serious about this too. The funny thing, I started computer based music servers back around the turn of the century and in those days all we had was iTunes on the first iteration of OS 10. There were no aftermarket juke boxes via software in those days, USB audio did not exist for the most part and in fact Firewire was the chosen connection method in those days.

Thanks again for in depth info which I will probably grapple with for some period.


I have worked in Pro Audio, and IT for the last 30 years.
Nevertheless I agree with you, it does seem a huge faff, with way too many options, and uncertainty about how many layers of translation occur end-to-end which is paranoia-inducing for most audiophiles

The convenience of armchair playing of music appeals though, so I have a very simple set up (one ras pi) playing a library of my own CDs ripped. Even that was a faff, but a reasonable compromise.
Had I no experience with IT in general and Ras Pis in particular I wouldn’t have bothered, though I am glad I did.

The challenge for any system is to make it “just work” but be tweakable to an extent for those who wish to. Here’s hoping PS Audio’s upcoming Octave “ecosystem” can fit the bill :slight_smile:

1 Like

Well, you have a very unique system then. I’d love to see a photo or screenshot of whatever makes you think that the 192kHz sample rate is involved in your DoP playback. But otherwise, I wish you happy listening.

The very problem is that the Stellar Gain Cell DAC own is equipped with a firmware version that does not show sample rates for USB input.

It is a complain I sent to PS Audio and PS Audio refuses to fix. It’s not important it seems.

So sadly, the only thing I can show you are the Bit Perfect settings, when I do get time on my Mac I‘ll show you the settings.

I love the point you’re making there, Routlaw.

Simplicity is one of the things that drives my audiophile process. I have the DS DAC, a power amp, and two-way speakers. Only two of the DAC’s inputs are utilised, one for music via Roon and the other for whatever’s coming out of my TV. Both connections are actually Toslink, albeit one of those via an iFi optical/coax converter because the DAC has only a single optical input.

One of the reasons I love the DS DAC so much is that it delivers such great sound quality while making most of the digital audiophile arcana so much less relevant, in many cases completely irrelevant. Toslink sounds great on this DAC, even with the cheapest of sources. All that futzing around with USB cables and I2S converters and voltage levels on their power supplies… other people are welcome to it but that’s not how I want to spend my time, optimising for a problem (electrical noise) which you can sidestep completely by using optical connections.

And having made that decision, all that really matters is ensuring that I can get bit-perfect playback from Roon (also trivially easy) and then finding good quality digital recordings to enjoy. I really don’t care about streaming vs local files, what file format things are in, what OS my NAS is running, how many digital conversions the data goes through etc etc.

There is still a need to be mindful of AC power quality and nearby sources of electrical and RF noise, but that’s not much different to the situation with an analog-only stereo system.

In the end, while I am personally and professionally fascinated by how digital systems function to a very deep level, the DS DAC makes it easy and relatively inexpensive to create a wonderful music (or TV) experience with a user interface that my non-technical family can handle with ease.

Happy to tell you that my kids can hear the difference between lossless audio streamed from Roon/Tidal and compressed audio formats from other services. :slight_smile:


Hi Dvorak,

first of all I think nothing is wrong with your calculations.

However, like I promised the pictures of my BitPerfect settings and iTunes file information of the DSD Master DSD128 converted file.

  • left Window: in “opmerkingen” (remarks) the file information created by DSD Master.
  • right window: bit perfect DSD settings: dCS DoP
  • top right corner: Playing 24/192.0

Thanks to your comments, I was prompted to play around with the settings of bit perfect again. Whether Bit Perfect or MacOS Mojave had a bug, I don’t know either. When I configured the Bit Perfect system settings in Mojave, the Stellar Gain Cell DAC would not play 24 bit 384 kHz without major blips, neither would any of pure DSD settings work, it resulted in no sound at all. Only the " dCS DoP" setting worked, like indicated in the pictures above, even the DSD128 files, as you can see. So whether bit-perfect downsamples when set to dCS DoP I don’t know that.

In MacOS Catalina with the updated bit-perfect app I can now play pure DSD128 via USB it seems. See picture below. The sound channel must be different, the volume level of the same song was lower when I changed the setting. SQ did not suffer, certainly, I can not conclusively say it’s much better, but it’s not worse and certainly good.

Thanks for pointing this out. I am happy that I can now set for pure DSD rather than DoP.

Even though the DoP conversion does not change sound quality, the conversion adds to the confusion.

Considering the above, I also see the very point of Paul Mc Gowan having hardware and software in one hand for the Octave server/streamer and service.
But, it would have helped a lot if the Stellar Gain Cell DAC would confirm on it’s display what file format it is playing through USB input.

1 Like


That’s why my modem is connected directly to my server and my server is connected directly to my streamer, and the latter connection is a $10 CAT6a cable.

The simple answer is to buy equipment that is designed to be hassle and stress free.

If a protocol for data transmission is good and effective, it will become popular. If it is either bad or ineffective, inappropriate or non-standard, it will not become popular. For me, popular is good. Unpopular is bad.

But then some people just like life to be difficult and complex …


Thanks for following up. If the something in your setup can’t handle 384kHz PCM, maybe it can’t handle 352.8kHz either, which is the PCM sample rate required to carry DSD128 in DoP format. So if you had a DSD128 track but no access to 352.8kHz then perhaps BitPerfect was converting to actual PCM192 on the fly (not DoP) and sending that. That’s my best guess. You could confirm by re-enabling DoP and playing a DSD64 track, which would go out packed inside a 176.4kHz PCM stream.

BitPerfect and other good audio software implement “native mode” where they skip the operating system’s USB Audio Class Drivers (which don’t support native DSD transfer) and communicate directly with the DAC.

The DirectStream DAC can’t handle 384kHz PCM…but it handles 352.8kHz just fine.

He’s got a Stellar GCD. And also I said “something in the system” because it could be a computer or software problem rather than DAC.

Thanks Dvorak.

I come to the the same conclusion as Stevensegal and Routlaw.

Way too complex!

I am glad it works now, however I will really think twice again investing a lot of money for DSD128 files. If it is so complex, the DAC refuses to display what it’s playing and I find out that the software is converting to PCM instead of DoP and I am not hearing what I paid for, I am not happy.

I am going to the record shop today and buy some vinyl and red book CD’s. Much less expensive and not vulnerable to buggy operating systems and software.

Funny enough the Stellar Gain Cell DAC does display the file format it’s playing on all digital inputs, also my optical, except on the USB input where you need it the most.

1 Like

What Rudolf_Appel said!

Literally all these issues and complexity make me want to spin Redbook cd’s and vinyl which is on an order of magnitude much simpler and it can still sound awesome too. And here’s the real conundrum, if you added up all of the audiophiles on this forum and all of the other audio forums in existence we would not make up 1% of the the people listening to music. The rest of the world at large doesn’t give a rats behind about any of this including high fidelity sound.

Outside of my audio friends in person or on forums I know of no one who has even heard of SACD or DVD-Audio for that matter and most who have no idea of a computer based music server. So while Sony was introducing the first SACD player, let alone before DSD came into the fold, Steve Jobs and Apple had already stole hearts and minds of a few generations to come with their mobile iPod. The rest is history. By any measure the vast majority of the people I know only listen to music via a pair of earbuds and mobile devices, and don’t even own a pair of speakers let alone black boxes of various configurations.

We can all talk about the sad demise of the audio industry, but I don’t see this changing anytime soon. And yet there seems no signs of abatement of continued complexity within this hobby be it video or audio. Is it any wonder the masses would prefer earbuds delivering convenient audio.