The Sabre-based DACs MUST have an excellent design with high-end clock to just work at all. With a mediocre set-up (such as with the Calyx DAC 24/192) the music is detailed, but sound flat. Using better power than just USB and the Audiophilleo + Pure Power it sounds excellent IMO, far better than when using the in-built USB does.  I think most people who review it spin CDs for the most part and are behind when it comes to computer rigs (unlike myself, where I refuse to use CDs if at all possible) and very likely even their highly expensive transports are rubbish compared to the top USB converters now, as every one commented that the USB input to the Calyx sounded better than their multi-thousand-dollar transports.

I personally find almost everything I've listened to that uses the latest Sigma Delta BB chips: PCM1792, 3 and 5 to sound unnatural and awful. I've been a Luxman fan for a long time, but when I tried their DA-100 and DA-200, even numerous times, I was shocked that they sounded like total shit to me. I know it sounds crazy to be talking about DA converter chips alone, but in every case we're talking about equipment from companies who know how to do build a component properly. Regardless, the DA-100's output is directly from the chips, which have built-in OPAMPs, so there is no question that the issue has something to do with the digital conversion.

Add $600-$2.5k (or not) as you wish for a high-end USB converter to the Metrum Octave and higher (in price) for best results (except maybe the NAD, as it is too radical a design to know if it would be improved by one).

The Sabre is very tricky to implement properly. It definitely needs good power, good clocks, and a good output stage. Cut corners anywhere, and you end up with mediocre results. I'm unsure of just how good the Calyx's clocks are, but I think a major part of the problem is there is no isolation. Driving it straight from the USB port is just asking for poor sound. It allows the Calyx to technically be portable without the need for batteries, but you end up with the worst possible DC power. The included wall-wart is better, but you're still feeding the ground noise for the source straight into the DAC. Even the optional linear supply can't fix that problem. Something like the Off-Ramp or the Transient MKII where the USB power line can be cut off entirely and the ground fed through a choke like Empirical's Short Block should improve the performance substantially.

The Anedio's USB port connects to a U2 converter that sits inside the case. On the positive side it has galvanic isolation, on the negative side it has to be converted internally to S/Pdif which connects to the D2's WM8805 digital receiver with a short length of unshielded wire. Assuming it's literally just a U2 with the BNC connector removed, that means the USB circuitry is bus powered. The clocks used are decent, not amazing. Still, there's only so much you can expect for $1500.

Totally agree on the BB PCM179x chips - crap. The AKM4399, AD1955, and WM8741 all seem to produce consistently good results in well designed DACs, and I would take any one of those before anything with a BB chip in it. The Sabre is much more "try before you buy". I also don't like Cirrus DACs, and I frown on anything with the CS8416 receiver.

There's two DACs I would suggest considering in addition to the ones already on your list. Neither of these have USB inputs, but both are VERY highly regarded in professional circles. Combined with an Off-Ramp or Transient, these could be considerable performers, although admittedly the investment is then $3500-4000, not chump change by any means. That's Invicta money, so they'd have to be extremely good to be worth it.

Quote from: longbowbbs on September 22, 2012, 11:46:17 AM

Currawong, what about the PWD or the W4S DAC2?

The W4S is crap. The USB input is some sort of proprietary tech that's WAY inferior to XMOS, M2Tech, and other common asynch implementations.

What are your favorites?

The Forssell and the Burl are both great, but in listening tests the Crookwood was almost on a par, so I went for that, as it is considerably cheaper. The only downside being you need to order it as part of a monitor controller system which can be expensive. The Crookwood DAC is what's used in the DACS Clarity Headmaster.

Forgive my incompetence, but what other technologies are there other than sigma delta on modern DACs that could potentially be implemented in here?

As far I know there's either serial (one bit) and parallel (multibit) processing of digital signal. Multibit DACs must use resistor ladders for signal processing and therefore are more expensive and difficult manufacture. A special case are multibit DACs that use discrete resistor ladders and those are even more difficult to build. Imagine something like a dozen stepped attenuator boards with custom made precision resistors that must not drift under t°.