Think I'll just get a JDS Atom at some point down the road and hook that up to a SU-1/Kuang Pai/E30 III Lite/similar.
You could get just an Atom or similar amp (e.g. Sabaj A20h, Topping L30 II or Schiit Magni) and run that off of the Scarlett's headphone out with a stereo 1/4" (or 3.5mm with adapter) to 2x RCA cable, you wouldn't be missing out on much. Going by
@Julian Krause's 2i2 gen 2 measurements, you can expect a maximum dynamic range of about 106 dB(A), which is just fine for a line-level source (and, considering that, I would probably keep its output level setting 3-6 dB below undistorted maximum, which should still yield about 105 to 102 dB(A)).
The Dynalo seems decent enough as long as you're not trying to drive planars, I just don't know about output noise level with IEMs given the relatively high gain of 11 (with no explicit loop gain compensation in sight, I would not advise reducing it either). Since the input circuitry ought to be fairly low-noise by itself, you could swap out the feedback network and volume pot to reduce its thermal noise contribution instead:
(each channel)
R27 (10k) --> 2.2k to 3.3k (ish)
R28 (1k) --> R27 / 10, i.e. 220 to 330 ohms (ish)
C6 (10 pF) --> no change (this is only supposed to counteract response peaking due to input capacitance and as such is only a function of desired gain and input capacitance itself)
(both channels)
Volume pot --> 2x 10k log (if previously higher, like 50k or 100k)
Resistors should be 1% 1/4W metal film much like the originals and can be oddball values from higher E-series as well. If you just happen to get a good deal on 274R and 2.74k, by all means.
In light of the substantial
out-of-band noise of typical CS4272 implementations, I would also contemplate giving the amp some input lowpass filtering (pre-pot). Here's a passive 5th order 50 kHz inverting Chebychev that
this tool spit out:
(L: E6, C: E12)
Now I have no idea what kind of inductors are available in this range that would be suitable for the audio band. It would probably be smarter to use a few opamps. I used the
Analog Filter Wizard to come up with this 4th order Butterworth with a ca. 29 kHz cutoff:
I used the ADA4625-2 as a substitute for a fairly typical audio opamp. Being the inverting multiple feedback topology, an NE5532 or NJM4580 ought to work fine in this. (R film, C film PP or ceramic NP0.) I won't be surprised if there are finished PCBs out there that could be used, it's hardly an exotic circuit.
Ideally you'd even want a 6th order filter.
