I hope I don't bore anyone to death with this long post... only read on your own leisure, if at all...
(A detail I left out first is that I didn't have all the boxes outputs at +7 dB, that was a first (and too high) rough estimation made listening with one pass through, at the last outputs of the 8 times loop I had a smaller value to correct. I now have +6.2 dB in each output, which is a tiny bit too low. I still have to do a precise measurement. But at least going from 8 to 1 loop the error is divided by 8 also.)
DSP 4x4 Mini input specs: Level +12 dBu, Impedance 1 MΩ (stereo), 500 kΩ (mono)
According to this calculator:
https://www.analog.com/en/design-center/interactive-design-tools/dbconvert.html#
+12 dBu means 3.08 Vrms / 4.36 Vpeak
Would that be (1) the voltage difference between hot and cold, or (2) between hot and 0?
Then inputting an unbalanced signal with 1.54 Vrms / 2.18 Vpeak would be
In case (1): +6 dBu
In case (2): 0 dBu
Right?
if you're applying DSP, say a +6dB EQ boost, reverb or some other process which adds to the signal, then effectively you're not really throwing anything away.
But to minimise the degrading effect of the ADC wouldn't it be better to input max level, and lower the input gain - assuming this works in the digital domain after the ADC - to create headroom for the processing?
(Internal processing is 32 bits by the way.)
Actually I have listened to the boxes in 2 different situations:
1. the boxes after the volume control (between the pre-outs of my receiver and the main-in power amp inputs of an integrated amp with removed pre-main bridges). Which is of course the worst possible situation.
2. the boxes before the volume control (between source and a normal input of the receiver).
As a "source" I used the zone 2 output (fixed volume, so full line level) of my Yamaha RX-V771 receiver but it is: 200 mV (1.2 kOhm); which according to the calculator is: -11.76 dBu, so almost 24 dBu too low!
Still very far from optimal.
[Edit: and this was in case (1), but now I know case (2) is "the case", so the situation is even worse: I am inputting almost 30 dBu too low! Which actually increases my respect for these boxes because except the added noise I can not hear anything wrong with 8 loop-backs.]
My intended use for the boxes is between the 16 analog outputs of my Smyth Realiser A16 (on order) and 16 channels of amplification in the form of 2 obsolete Yamaha RX-V750 7.1 receivers with analog 7.1 inputs and an old Yamaha AX-592 integrated stereo amp. For use of the A16 as a 16 channel decoder with real speakers, and for PRIR measurements both with PEQ, level, and delay in the chain which is otherwise problematic to achieve with only the analog outputs available on the A16 (most av receivers can not apply any dsp to analog 7.1 inputs).
Ideally I would have to do the volume control after the boxes (set A16 volume to max, or max -2.24 dB, see below), maybe I should find a third RX-V750, because their volumes follow the remote control beautifully synchronous (with digitally displayed discete steps, whereas the AX-592 has a motorized potentiometer that of course doesn't follow the others correctly and even by hand would be difficult to set right).
The A16 16 channel analog output specs are a bit more promising:
Impedance 10 Ohms
Peak Output 2 Vrms
Which would mean +8.24 dBu.
So in case (1) at least I can reach the +6 dBu on the DSP 4x4 Mini inputs.
But in case (2) I can reach only +2.24 dBu.
Correct?
But I am wondering if in case (1) it would be possible to input +8.24 dBu in the box unbalanced? If the input was implemented using a transformer (very unlikely I guess), or a differential amp that doesn't care about the absolute voltages (if that is possible at all?)...
By the way [Edit: back to the 8 loops test situation]: if I only had the cables I could do the actual loop-backs themselves balanced, so that only the first inputs and the last outputs of the loops are unbalanced. And I could raise the input gain of the first inputs (and lower the output gain of the last outputs) such that the signal reaches full level and probably stays that all the way through the 7 "middle" DA-AD (from the inside-the-box perspective) conversions. Then I could check the influence of 8 conversions with only one time the (extra "gainstaging") added noise. I have the first inputs and the last outputs in one and the same box, so in that box I could switch (using the "Matrix", just unfortunately not a fast instantanious switch) between including or excluding the additional 7 DA-AD conversions (all full level, balanced connections).