Consider that a stand-alone machine with a PLC: you are in control of what and what is not common grounded on most PLCs by the way you wire it. Often times there is a jumper you can removed so that you can force the common ground point to be that desired central point in your main panel. Now throw a PC into it; if your IT group mandates what PC is used (instead of being able to choose an industrial PC for this and other merits) you suddenly are not in control of the ground point! Most commercial PCs internally ground the DC power to the chassis via the motherboard! In some cases this creates a nightmare of intermittently working peripherals where the solution is NOT to ground the 24VDC in the panel but have the central ground point be something as close to where the PC plugs into as possible. If you can provide a dedicated power connection for the PC (think a hard wired line cord as short as possible) that is grounded in the panel at the 'single point' then common the 24 to this point. You may be 'floating' the 24V supply because some device is forcing the common ground point to be somewhere else. I often find myself checking peripheral devices and sensors to verify if they make the big mistake of internally connecting chassis and/or power to signals. One must visualize the circuit as a whole when everything is connected together as there are some devices that are making choices for you and now you must figure out how to undo their nefarious acts so that a true single point ground can be achieved and if additional isolation needs to be thrown in.
Grounding the negative of a 24VDC power supply is good as a common point. Quite a few years ago I was called to a plant where we'd just installed a replacement Aluminum melting furnace (the original had blown up due to ice in the scrap added to the metal pool). The furnace had shutdown. When I checked, the multi-loop programmable controller had lost part of it's memory. I called the manufacturer of the controller and one of their Engineers came to the plant. He suspected grounding as the problem and hooked up his Oscilloscope up to the ground rod that the control panel was connected to. The ground point had lots of noise and when he switched the scope to Ascii instead of normal voltage he pointed out that the dump program ascii sequence was apparent.
The controller was programmed such that it's Serial port was programmed as a printer port for printing process temperatures. The port received handshaking signals from the printer so, was getting noise induced commands.
I asked him what could be done to test the ground rod system and he said drive five ground rods down to the water table, connect them to one common point and use this as the reference. I said the plant might as well replace the plant ground rod(s) as the cost would be about the same he nodded and said exactly.
We found that the source of the noise were Tow-motor magnetos when the tow-motors passed the control panel. The plant would not spend the money to test or replace the ground rod(s) so, we stopped using the printer port and installed a multipoint temperature recorder instead. Grounding to the water table is essential keeping in mind that water tables rise and fall.