OK, settle in for a long post. Taking this from the top.....
For a typical CNC machine there may be many different "grounds".
The first is the safety ground. It starts as the ground wire that comes
to the machine through the mains power cord and should be connected to
the frame and cabinetry of the machine. When other considerations tempt
you to not ground the frame to the mains safety ground, think again.
Safety is the most important thing. Your computer, unless it is a laptop
or tablet, will also connect to this safety ground.
The next ground is the signal ground used by signals coming out of the
USB and parallel ports of the host computer. This ground is necessary
for the interface to "look" at the signals.
Desktop computers internally connect the safety ground to the signal ground
of USB ports and parallel ports. This is where problems can arise from ground
loops when the safety ground of the machine and the safety ground of the host
compute are not at the same voltage potential. This can happen when other
devices on the mains circuit are causing current to flow in the safety ground
wires of their mains connection. In industrial settings this can be tens of
volts sometimes. When the USB or parallel port cable is connected to the host
computer, and the signal ground at the machine is also connected to the
safety ground, then the ground wire in the USB or parallel port cable will
try to short these two grounds together and current will flow through
the ground of the cable. This is called a ground loop and it is not a good
thing.
The first step in avoiding ground loops is to power the host computer
from the same outlet or power strip as the machine. This will minimize the
magnitude of any voltage offset between the grounds. The next step is
to use an interface design that electrically isolates the signal ground of
the host computer from the safety ground of the machine. Low cost
interfaces such as the SmartBOB USB do NOT provide this isolation so
the quality of the safety ground wiring becomes more important.
Ground loops can also be avoided by using interfaces with isolation. This
can take the form of isolation on the host port side, or isolation on the
machine signal side. Sometimes both are utilized.
Isolation on the host port side can be implemented via Ethernet connections
which are, by design, isolated by a signal transformer. Isolation can also be
provided by special versions of the USB port on the controller or by an
isolation device in the cable between the controller and the host computer.
Isolation on the machine side can be provided by devices such as opto-isolators
and/or by simply avoiding connections between the machine interface signal
ground and the safety ground of the machine. It is common for axis motor
drivers to internally provide opto-isolation of their Step and Direction
signals. Geckodrive, Leadshine, and most packaged stepper and servo drives
do this. For low cost interfaces that do not have isolation circuits for the
machine control and status signals typically extend this concept by not making
connections between the control signals and the machine frame. This can be as
simple as not grounding limit switches and sensors to the frame of the machine,
and by using mechanical or solid state relays on output signals. When doing
this it becomes important that power sources for the sensors or relays are
also not grounded to the machine frame.
You mentioned that you are using a PMDX-132, and it provides opto-isolated
inputs for limit switches and sensors. This will allow you to connect the
ground side of these sensor inputs to the machine frame if necessary, but
without any special reason to do so, I would not.
When using axis motor drivers with isolated Step and Direction signals the
negative side of the power to the drivers can usually be connected to the
frame of the machine as long as that power supply is not powering anything
else such as sensors that connect to non-isolated inputs on the controller.
It is common to connect the negative side of this power supply to the frame
ground as it will often reduce the level of electrical noise radiated from
the motor wiring. Do not do this if your axis motor drivers do not have
isolated Step and Direction inputs.
In your case, you mention that have a PMDX-132 for mounting your axis motor
drivers. The DC power connections to it go only to the Gecko motor drivers which
have opto-isolators for the Step and Direction signals. The power supply going
to the PMDX-132s motor power connector can safely have its negative side tied
to frame ground of the machine as long as nothing else powered by it has a
grounding conflict.
Again, in your case, you mentioned a scavenged PC power supply being used for
accessory power. This power supply will, by design, have its safety ground
and the negative side of its outputs internally connected together.
Your proposed ElCheapo laser vendor does not document how the PWM interface
is implemented, but it is likely that the PWM signal uses the same ground
reference as the power input. This in combination with the scavenged power
supply having its negative side tied to safety ground means that whatever
drives the PWM input of the laser will also get its signal ground tied to
safety ground through the lasers power supply. In the specific case of the
PMDX-132, the logic level control outputs are opto-isolated from the host
computer port, but share a ground with the limit and home signal inputs.
This means that you should be able to safely connect the negative side of your
scavenged power supply to the ground of the J9 output signal connector of the
PMDX-132. This analysis applies ONLY to the PMDX-132 and is not valid for
other boards or to the PMDX-410's direct outputs. You must use the J9
connector on the PMDX-132.
I know this was a long description, but I hate giving a simple answer without
explaining the reasons for it.