When Emotiva stated 4.5v peak (unbalanced) and 9v peak (balanced), did they mean max?
Peak to me indicates the peak voltage of a sine wave.
Caution: You have entered the world of AV/Home Theater specifications.
An example: You'll often see specifications such a 120W RMS. Watts RMS is of course a meaningless term, and it unlikely that the 120W is actually calculated in in RMS terms, although it certainly can be calculated.
So how about peak? First note, peak is often not used in the physics/E.E. sense. Often peak means maximum V RMS.
In the real world marketing/sales people have contributed some of the specifications, which actually may be high or low. A call to tech support often gets the marketing numbers. Several calls to tech support often get different numbers. In addition the rating may be for some level of clean output, (not specified) when the maximum, say 1% or the like distortion is higher. Of course the rating could be 10% distortion and a reasonable person would use a lower number.
Let's make our own guess. For the Emotiva unit as described in a post above:
2V is the nominal output of the AK4490 DAC IC. That's 2V per leg. Assume 0dB gain for the output filter, 6dB gain for using the channels in differential format. This means the differential signal is at 4V RMS.
The above means 2V RMS per leg is the input to the CS3318 volume control That volume control is rated at low distortion to over 5V output with 2V input with 9V DC supply rails. That makes a maximum of 10V RMS for the differential signal. Of course the input to the volume control can be less than 2V and still get 5V output based on the gain available from the volume control. We don't know the actual gain structure that Emotiva has implemented in the unit.
Self, at noted in the post above, has tested the 5532 opamp. He uses 10V RMS into 500 ohms, likely with very high rail voltages. Clearly the 5532 can handle robust signal levels. Using the datasheet for the 5532 an estimate is that it can do 7V RMS with (+/-)12V DC power rails, which is about as high rail voltages as you'll typically see in consumer level AVR/AVP's. This implies 14V for the differential signal. Sadly many AVR's/AVP's use 7V DC rail voltages, which restrict headroom and also distortion performance.
Based on the assumptions of the gain structure above and the components involved, it appears the volume control would be the limiting factor and it can do a bit over 10V RMS differential. Based the 10V RMS calculation, 9V RMS seems reasonable. The distortion at that level is of course unknown. That said, it would really take an upfront agreement on what level of distortion should be used, then measurements of an actual unit to be sure of the actual value, as where to find a truly authoritative source is not obvious.