It looks like no one so far is that keen on the regulator-based circuits previously referred to here.
https://www.goldpt.com/virtual_ground_circuit.html
I will go with Brad's switching circuit above and breadboard it.
There is still the issue of 16VDC being the maximum allowable supply voltage for an LM555. I have come up with two solutions:
1. Dropping 24VDC to 12VDC with a regulator is a bit steep. Instead, I could feed the switcher with 15VDC using this 2A plug pack from Jaycar.
https://www.jaycar.com.au/Power-Pro...ut/Switchmode-Mains-Adaptor-15VDC-2A/p/MP3492
With a bit of luck, I will wind up with a 10Vpp swing from each amp, one powered between the positive rail and VGND and the other powered between the negative rail and VGND.
2. I could feed the 24VDC into an 18VDC regulator and add a 1N4007 diode or two in series to drop it a bit further. This would allow for the full 18VDC rails specified in Brad's schematic.
Before starting, just a few questions please with regard to Brad's circuit.
1. It appears that applying a voltage less than 2/3 Vcc is intended to shorten the charge cycle and increase the output frequency. Why is this necessary?
2. If it IS necessary, and assuming the 15VDC supply, can I feed 9VDC to the LM555 control pin via a 9VDC regulator? I could drop it a bit more with a series diode. Or would a voltage divider off the supply be preferable?
3. I would use a BC327/337 pair for the left-hand transistors, and a TIP41/42 pair for the right-hand ones. 10W ceramics for the 12R resistors. And increase the output caps to 2200uF. Does that sound OK?
4. The rail voltages in the simulation are slightly different. What would be the best way to provide an adjustment to make them equal?
BTW here is a page on running an LM555 from 24VDC.
https://www.555-timer-circuits.com/555-on-24v.html
Thank you for any further input.