Another alternative to consider is the Simply Sound SS-1. I used these for some recordings done with my home-made ribbon mikes and they worked very well. I published a measurement suite for these in AudioXpress December 2017,
but the folks at AX were kind enough to put it on-line for free.
Hmm. If I'm not mistaken, equivalent input noise level come out as somewhere around 0.6 µV or 0.44 µV (A). Not exactly super great for a mic (pre)pre, is it? This is, if anything, at best on par with or slightly worse than a sub-$100 Behringer mixer (with two mic inputs and a gazillion other features). A mere 42 dB of CMRR seems a bit concerning, too (even if it probably was to be expected for the kind of circuit used and should be of little consequence if the device is used right at the microphone). Not to mention that the device is advertised with "Boosts the signal at the source for long
cable runs" but then sports almost 3 kOhms of output impedance... mind you, that's still good for maybe 20 m / 60' with a 3 kOhm mic input before cable capacitance intrudes on the high end.
That's a good set of measurements for distortion, but noise wise it all becomes much less clear. Quite honestly, I'd have trouble reproducing the test setup. You write you were using the APx1701, but what the SS-1 was fed from and what kind of source impedance it was seeing remains unclear. It almost sounds like you were just whacking the APx515 output into the input. From the instrument's specs it is not clear whether output noise is even low enough to permit doing this. Clearly it has to be pretty quiet, but I suspect it may still be in the same order of magnitude as the SS-1 itself, which definitely is not what you want for obvious reasons.
I would have suggested making a low-impedance passive attenuator - since output impedance of the generator itself is known to be an accurate 100 or 600 ohms (selectable), you would just need an adapter with two low-value resistors (like 5.1R) between hot and shield and cold and shield, giving you a selection of -20.6 dB or -35.5 dB, respectively. Theoretical best-case input noise density for a balanced input using LSK170s is about 1.3 nV/sqrt(Hz), that's the equivalent of a good 100 ohms, so <10 ohms of source impedance should be fine.
Figure 7 is another that raises some question marks here.
"The noise spectrum of the SS-1 compared to the mic preamps in a Focusrite Scarlett 2i2 audio interface show a slight increase in noise floor in exchange for the gain boost."
What exactly were you comparing there? Was it
a) 2i2 vs. 2i2 + SS-1, gain settings unchanged
or
b) 2i2 vs. 2i2 + SS-1, with gain readjusted for same total preamp gain?
That would be the difference between an equivalent input noise delta (b) and
Captain Obvious for anyone who remembers Friis' formula and knows a thing or two about noise vs. gain setting in typical mic pres (a), hence the distinction.
---
Now on the device in question in this thread, the DM1.
Specified output noise is 9 µV (JIS-A), minus 28 dB of gain that's -126.7 dBu(A). Hardly
spectacugreat but seemingly par for the course for these FET amps. Output impedance is given as 135 ohms, this device should be quite robust even with long cable runs then. Input impedance does not appear to be specified, even though it can make a noticeable difference in frequency response with dynamic mics. (It was quite high on the FETHead if memory serves. This can "open up" some mics that on regular 1-3 kOhm inputs might sound a bit constricted.)
If you have spare mic inputs you can double amp it. Put it into one mic pre and put halve your gain on it. Then feed that into a second pre for the rest. That doesn't add as much noise as you might think.
Basically a good idea but I would much rather suggest using the first mic pre at close to full tilt (maybe 10-20 dB shy tops). The reason being, equivalent input noise on typical mic pres will degrade substantially at gains substantially less than maximum due to increased feedback network impedance. Consult the datasheets of typical INA/SSM chips for examples.
So yes, the second stage would have substantially higher input noise. That doesn't matter, however, since output noise from the first stage is going to absolutely obliterate its contribution. Again, a case for Mr. Friis. I've done the equivalent math for noise voltages
right here (scroll down to
Gain staging for dummies).