I wonder if it'll do DSD/SACD...?
It — by the very definition of DSD — can not.
Massively* oversimplified:
A delta-sigma DAC, which this is, takes an LPCM data stream and converts the information it contains into a pulse width-modulated output signal, that then gets cleaned up to varying degrees and through varying means by an output filter—like Mike's Mega Combo Burrito—before it gets sent out the back of your DAC.
In slightly more detail,
but still very much* oversimplified:
(L)PCM is a continuous stream of integer values that represent the strength of the audio pressure wave at any given moment in time relative to a certain baseline. The higher the integer, the higher the sound pressure.
Do this often and fast enough in a row, and you end up with a wave form, "painting by numbers"-style.
The more often you take a sample and save that integer during any given second, the higher is the resulting resolution, and the more accurate the digital representation of the original pressure wave form. Take too many samples, and you waste space and run the danger of introducing high-frequency noise that wasn't there in the original pressure wave. Too few samples, and you lose information you need to accurately represent higher frequencies.
The pulse width-modulated signal that the DAC puts out is one of a fixed voltage, but it gets switched on and off at varying degrees to "fake" a variable output voltage. The higher the desired "apparent" output voltage, the longer that actual voltage signal stays "switched on." The lower the desired "apparent" output voltage, the longer that actual voltage signal stays "switched off." The frequency at which this switching happens remains constant with the sampling rate of the supplied (L)PCM data stream, but the length at which that fixed output voltage stays switched on within that fixed time frame varies in proportion to the integer value of that sample. (The proportion is linear in LPCM streams—which is usually what you have in audio—and as a function of amplitude in other PCM formats that are used in a bunch of other fields where analog data needs to be encoded into a digital stream of values.)
The more accurate this super fast switching of the output voltage can be done, meaning in terms of regularity in time as well as in steadiness of voltage, and with as little of a voltage spike and bounce during and immediately after the on/off state changes, the more accurate the resulting wave form will be that comes out the back of your DAC.
But before it can leave your DAC, this pulse width-modulated voltage gets fed through a filter. You can think of this filter as somewhat of a water basin where the modulated-in-duration-but-fixed-in-pressure water pulses come in at one end, and a smoothed-out and steady stream of varying voltage levels drains out the other. The better the filter, the less noisy your output. This wave form is what eventually comes out the back of your DAC. (There might also be an amplifier in this filter. Depending on the voltage the DAC puts out, the signal may need to be amplified to be closer to the voltages that are expected for RCA and XLR outputs respectively. Which is why you can't design a good DAC without also knowing how to design high-quality audio amplifiers.)
DSD skips over the (L)PCM and DAC part altogether. DSD is simply a stream of 1-bit samples taken at a constant sample rate: 2.8224 MHz, or one sample every 2.8224 millionth of a second. The DSD converter simply reads in one bit after the next, one bit every 2.8224 millionth of a second, and will either switch a fixed output voltage to an "on" or "off" state, depending on whether that bit happens to be a 1 or a 0. The result is—for all intents and purposes—the same pulse width-modulated fixed voltage signal that a "normal" DAC implementation puts out. And as such, DSD is merely a digital representation of that pulse width-modulated signal that can very easily be converted into the voltage pulses.
The only differences are that DSD skips one step during encoding, and one step during decoding—namely the conversion of the pulse width-modulated signal into an integer value, and back—and that it is fixed in its sample rate to 2.8224 MHz, and PCM can represent pretty much any sample rate you like.
On first glance, this would mean that you should be able to just take that DSD stream of bits, make sure that the "1"s in your bit stream are at the desired voltage level, and feed that more or less directly into your output filter, just as you would with the pulse width-modulated voltage that you get out of your DAC chip.
But it's not quite as simple as that.
DSD's accuracy lives and dies by the steadiness of your stream of bits. Any form of jitter this stream of bit contains coming in will simply be forwarded into your filter more or less unchanged.
And since the laws of physics are a thing, every single digital stream has some amount of jitter to it.
But with USB, it gets even worse, as there IS no steady stream of bits to begin with. USB delivers the data in small packages—including overhead information like the address of the intended receiving device, and some other information like package checksums—at any somewhat random point in time that the USB host sees fit—but never as a steady stream of bits. This is how a single USB port can deliver data to a whole number of different devices that are connected to it, by talking to one device at a time, cycling thought them one by one depending on the amount and priority of information that needs to be transmitted. This also means that any device that is currently not being "talked" to has to sit around and wait its turn.
To deal with the inevitable jitter from inputs like SPDIF, and to reorganize the USB packages with their overhead into a steady stream of pure DSD bits, you would have to implement an entire new controller just for that purpose. In addition to that, you will also need an entirely separate buffer that you can feed the incoming jittery DSD stream into, and re-clock that buffer's output to remove all jitter and gaps from that DSD stream that you just fed into it. Only then can you send that signal on to your filter.
As a side note:
The benefit of DSD is that it has fewer "moving parts," so there's less of a chance to introduce errors/noise into your information chain. But your result is only ever as accurate as the DSD devices that you have in your chain, as any node that your DSD signal passes through can introduce additional jitter into that signal, degrading the contained waveform information each time, which can only be mitigated by a dedicated and hopefully highly accurate buffer and re-clocker as the last step before your signal gets fed through to the output filter.
The benefit of (L)PCM is that you are infinitely more variable in terms of sample rate as well as dynamic resolution, meaning the detail you can catch and reproduce in the time-domain as well as the detail you can catch and reproduce in the dynamics of your original air pressure wave form. This comes at the price of more moving parts, and with that, more chances for things to go sideways if any one of those additional parts wasn't implemented with care.
Personally, because of the above-mentioned limitations, I prefer PCM over DSD any day of the week. At least as long as the people who implement my LPCM DACs know what they're doing. But I also know that people much smarter than I—such as Paul McGowan, for example—beg to differ.
All of Schiit's DACs put their entire focus on converting your (L)PCM integers into a pulse width-modulated signal as accurately as physically possible. To add all the additional stuff that would be needed to properly process a DSD signal would mean quite a bit of additional complexity, more failure points, higher cost, and—most crucially—less time to focus on getting the PCM side of things just right. So you will likely end up with Schiit DACs that may be able to do more, DACs that will cost more, but also DACs that won't be doing any one thing nearly as well as they can do them now. And all that just to play a DSD data stream that—if done right—won't very much resemble what you've fed into the box to begin with.
Personally, I'm not sure that this would be worth it. But I'm always open to be proven otherwise.
* Seriously, though, this is nothing but a tiny glimpse into an incredibly complex topic, one that I don't fully understand yet myself.
