I'm working on my signal chain and thought I could add a few data points here. I don't fancy sweeps, as I was using an oscilloscope to look for clipping/distorted waveforms. However, that saves me having to worry about my loopback measurements and a questionable soundcard.
Here's the signal generator from REW out via hdmi into an Onkyo TX-NR809 receiver.
screenshot of 5 hz:
The yellow boxes at bottom are displaying period, frequency, Vpp, and Vrms.
With AVR at maximum possible non-clipped output, I saw 22.5 Vpp (!) at the sub out. In the screenshot you can see that 5 hz was 20.2 Vpp, and also see the min values from 2.5 hz where it dipped to 15.3 Vpp.
I make that 20 * log10(22.5 / 20.2) = roughly -1 dB at 5hz
and 20 * log10(22.5 / 15.3) = roughly -3 dB at 2.5 hz.
Interestingly, I could push the AVR crazy hot without clipping the output. The above measurements were taken with the master volume pegged at +18 and sub out at -2.5. Adding another half dB chops the waveform unmistakably. Because I assumed it would clip much sooner (my Onkyo TX 535 does), I had the sub out previously at -15 with lots of amp gain, since I regularly go above 0 master volume with music. It looks like I can get a bunch of SNR back. Unless I'm missing something and the AVR is not routing all the input signal to the sub preout. My AVR was reading multipcm in with REW running, but I couldn't tell exactly what channel(s) the signal generator widget in REW was outputting to. I'll next try to see if I can grab an actual full sweep into the scope from REW where I know I'm outputting on exclusively the sub out channel. That leads me to my next concern, which is that I don't clip my minidsp inputs. I can adjust my AVR down a bit to not exceed the balanced minidsp 4 Vrms max in, but I don't know if/what effect bass management has. More testing required...
Also, I measured my Onkyo TX 535 upstairs and saw similar rolloff results --I think it was 1.5 dB down at 3 hz. Its preout was even hotter - up to 27.4 Vpp or 9.6 Vrms before clipping.