I've personally no evidence that the professional audio market is any less susceptible to changing vogue or irrational expressions of machismo than the high-end consumer market. This fashion of escalating rail voltages has been evolving over at least a decade, IIRC a big influence here was the design and marketing of Solid-State Logic consoles with high-voltage (+/- 90v?) rails.
While I have great respect for SSL and no specific opinions on the SPL gear you mention . . . here are some very specific reasons why I chose to optimize a recent discrete opamp design around +/- 24v rails:
- The requirement for higher voltage imposes some severe constraints on semiconductor manufacturing, and consequently the selection of small-signal parts available to circuit designers really plummets above the 40-60 volt range. For bipolars, lower gain and higher noise is a given to withstand voltages above this range, and there are virtually no matched precision pairs or small-signal JFETs of any type at all.
- The methods for using lower-voltage devices in higher-voltage circuits (i.e. variations on cascoding and bootstrapping) make up some extremely well-trodden ground, and in the best situations the only drawback is complexity. But frequently there are also penalties in noise performance or open-loop response (adding another response pole), the latter means increased distortion, reduced bandwidth, and/or reduced stability.
- Increasing the rail voltages is pretty inefficient at creating additional dynamic range (each doubling of the voltage gives you 6dB). On the other hand it's very effective at increasing heat dissipation, and with higher signal voltages the distortion mechanisms of passive components become much more significant.
- Maximum voltage output with +/- 24v rails can easily exceed 13VRMS, even without special attention to "rail-to-rail" circuit design. This is enough to b!tch-slap the input stage of most any piece of equipment that's likely to follow a preamp, ADC, DAC, EQ, or whatever . . . and that's assuming either an unbalanced, pseudo-balanced, or 1:1 transformer-balanced design. For active-balanced outputs (or a 1:2 transformer) the maximum output is doubled to 26VRMS . . . which for perspective would drive a 4-ohm speaker to 170 watts (if it had enough current of course). By my logic this is more than sufficient for any line-output stage, and if substantially more is required to keep something from clipping inside the equipment, then I need to re-think the equipment's internal gain structure.
- Speaking quite conservatively, a well-designed discrete-opamp line-level architecture with +/- 24v rails can achieve a -110dBu noise floor up to maximum output of 24.5dBu (13VRMS), meaning 134.5dB dynamic range. This exceeds that of a low-noise dynamic microphone (let's say 210 ohms and 1.8mV/Pa based on the industry-standard AKG D112 kick-drum mic) suddenly being moved from an extremely quiet room (15dBa) to a position 50 feet from a jet engine at takeoff thrust (150dBa).
In the end, I felt that setting by the rails at +/- 24V, I could design a simpler circuit that achieved better performance while drawing less power and producing less heat, with no real-world loss in usable headroom. Other designers may of course come to other conclusions for their applications . . . and if they happen to enjoy bragging about how high the rail voltages are on their designs, then they certainly win on that point.
