csma/cd do packets or frames or signals collide?

with csma/cd, is it more correct to say that signals collide, packets
collide, or frames collide? are any of those wrong?

The detection must occur at the physical layer, so it's signals there.
But once there's a collission, i'd have thought it's frames. and all
frames contain packets.

i know that cisco refers to signals colliding. But most other sources
refer to packets colliding, and many - but fewer - refer to frames
colliding. IF it is packets colliding then that's more specific than
signals. But is it as correct?

 

All of them are technically wrong, but "signals" is the least so.
Unlike "signals" and "frames", "packets" is very wrong, albeit
common usage.

Not all frames contain packets (consider LLC2, Spanning tree BPDUs,
etc). Frame is an informal name for a "MAC protocol data unit," but is
also an informal name for a "LLC protocol data unit." In the latter
sense, it could be incorrect because a collision could occur before
the LLC portion of the frame ever gets put on the Ether.

My personal preference is to just state that a collision occurs, because
the MAC PDU which is being sent could be colliding with ANYTHING,
including noise on the line.

Whether or not signal is correct depends upon your definition of
"signal." In common dictionary usage, a "signal" is something that
serves as a means of communication, so noise and line faults would
be erroneously excluded if we required that "signals collide." Even
in the absence of failure, a jam or flow control signalling from
a switch would not be considered a "frame," so there does not
have to be another frame present on the line to cause a "frame"
collision. And, as already stated, the presence of a network PDU
(packet) is optional, so a "packet" collision is clearly misleading.

As always, YMMV. If prepping for a test, the correct answer is whatever
the book/lecturer/test writer says it is.