Subjectively I think it’s a better sounding (i.e. more enjoyable) unit than the Red, though as I mentioned earlier that’s going to depend somewhat on taste and what cans it is being paired with.
Fidelity-wise, it’s clearly worse than its predecessors simply due to the choice of filter (other issues not-withstanding). There’s a reason I bitched about the minimum-phase filter and slow-roll off model before ever hearing this thing. You can’t roll-off that early and claim “hi-fidelity”. And “natural sound” is purely subjective anyway.
Still, it is enjoyable to listen to, directly driving appropriate cans, when it’s not at the limit.
I suspect that with most of the headphones that things like this get paired with, a bit of top-end roll-off is probably quite desirable (for me that’d be pretty much anything carrying a Grado or Beyer label, for starters).
No one that’s pairing it with reasonable headphones is going to be running this to the rails, so that’s likely a moot point, but it definitely overshoots it’s spec there and not necessarily with good results. Though it’s behavior at “max volume” certainly isn’t what one wants to see.
I’ve gone on like a broken record, in other threads, about not using these things as pure-DACs into other amplifiers, which is where the “behavior at max volume” is really going to be an issue.
And I still maintain that unless you can only have ONE DAC/amp, then having a different desktop stack, even at lower cost, will get better results, and still leave money on the table for a pure-and-tiny-portable option. Value is a big question mark at this level.
For a bit of fun I fired up the cheap analyzer again … (I could break out the “big LeCroy” but using $900K+ instruments, with no audio-centric-automation, in multi-million-dollar isolation facilities, to measure $300 devices is more of an ePeen-waving exercise than anything truly useful).
So here’s the Red vs. the Cobalt in “inferred” jitter (pure sine wave at 1/4 fs), Red first (excuse me not resetting the X-axis range for the Cobalt):
The differences here are so far below “audible” that they’re ridiculously irrelevant, even though the Cobalt does have a smaller noise band and lower noise artifacts.
And then the same setup for an actual J-Test (1/4 fs primary tone, with the LSB toggled every 250ms - so those spikes at 250ms intervals are supposed to be there. The Cobalt is cleaner and more correct here. Again, Red first:
And, as a final comment … I am returning my Cobalt (a normal, personal, “hell, it’s same day” Amazon purchase). Not because of audible issues. I still think it is subjectively more enjoyable than the Red with what I use “on the go”, but because the “rattling” and movement of the connectors/PCB is pissing me off …