There are a few constraints for effective load-side phantom power supply design. The ripple needs to be fairly low, as most amplifiers only have -40 dB or so of PSRR. We want to reject noise picked up from the XLR cable, as the power is a common-mode signal and thus vulnerable to interference. We are limited to a few milliamps of current, so keeping power consumption down is favourable. That said, this is plenty for modern amplifier ICs.

Hopefully, most phantom power supplies are fairly clean, but that's entirely an assumption as I have not actually measured any under load yet. The TI reference design for phantom power I've used in Pmod XLR uses the

So far as I can tell, Texas Instruments, Analog Devices (LT, Maxim, etc.), Renesas (Diodes), and onsemi are the the main manufacturers of appropriate op-amps. Perhaps there are some Chinese manufacturers? This is a problem for another time. LCSC has a huge selection of components but they're not as well organized for searching as the manufacturer's websites, Digikey or Mouser. For now, I'm sticking with TI.

Most amplifiers can't handle a 48V total swing. For example:
OPAx202:
OPA164x:
OPA1652, 54:
OPA1655, 56:
INA1620:
OPA1611, 12:

Here are devices that can handle the swing:
JFE150:
JFE2140:
Power op-amps can handle the swing as well.

Zener diodes are simple devices. Past their breakdown voltage, zener diodes have a small resistance, meaning the diode does not conduct exactly the same amount of current. Thankfully, current consumption is not a major concern for phantom power, as long as it can be kept within 5 mA or so. Due to how they work, they can readily be found in configurations that can handle 48V without any issues.

Finding linear regulators that can handle 48V is a challenge, and they're not particularly cheap. Regulator PSRR is load and frequency dependent, though the loads here will be both small and low enough frequency that this is unlikely to be a concern, so long as there is good filtering from the source and RFI filtering at the device.

Charge pumps are a small, effective way to provide an inverted power rail or a small boost - though the latter is not necessary here. I'd like to look into these more and see if the noise is acceptable. They would need to be used in combination with a Zener diode, linear regulator or buck converter as a 48 volt swing is already too large, as noted earlier. This would add extra cost and space over using a single supply from an linear regulator or zener diode, but it would make amplifier design simpler by eliminating the virtual ground. They are most efficient when used under high loads and when converting to a multiple of the input frequency, in this case the multiple would be -1.

Buck converters, without some additional engineering, risk being noisy, plus they're limited by the size of the inductor. On the scale of a guitar, the size probably isn't a huge problem either way. For a guitar design, where the switching converter would be in close proximity to the pickups, even if used in a shielded enclosure or cavity, I suspect it will be ill-advised. This will be the last model I test.

These converters have low current ripple, but only provide negative voltages. Could be an interesting option for a low voltage rail.

After some simulations, I don't think they'll be an appropriate choice. They appear to need large inductors to work well. When I learn more about how they work, will revisit them.

https://sound-au.com/appnotes/an008.htm
https://sound-au.com/appnotes/an007.htm
TI - Phantom Power with Operational Amplifiers
Burr-Brown The Instrumentation Amplifier Handbook.pdf
Basic Concepts of Linear Regulator and Switching.pdf
Techniques that Reduce System Noise in ADC Circuits.pdf
https://eepower.com/technical-articles/intro-to-cuk-converters-part-1/!