Where to put ground wire on a connector among many signal wires and power?

[John Celo]

Lets say I have 5V, signal1, signal2, signal3 and GND (configuration 1)

Ideally, there would be return path for every signal. But lets say that for (space restrictions) reasons we can only have one gnd wire.
Where would you put it? Next to 5V line or in the middle somewhere, or on the opposite end of 5v?
Or does this even matter?

You could have arrangement like 5V, GND, s1, s2, s3.
Or 5V, s1, GND, s2, s3, etc.

Which one would be better, or does it even matter? Or if you had a second GND wire, where would you put it?

Or even in a simpler case of having just three wire connector - signal1, signal2 and GND.
Would the best arrangement be putting signal1, GND, signal2, such that gnd is in the middle?

[tggzzz]

Assuming they are digital signals inside a single box, the Gnd and Vcc are equivalent from a transmission line perspective. Hence I'd use s,Gnd,s,Vcc,s.

If they are analogue signals or if they go outside box/cabinet, the answer is more complex, nuanced, and dependent on many subtle factors.

[John Celo]

I rarely (if ever) see connectors being arranged this way.
Vcc, Ground - is almost always on one end. (or opposite ends)
Why is that?
Is that the case of it simply "not mattering" in most hobbyist applications?
ie. Arduino Uno pinout with whole line of signal lines and one sad single ground at the very end.

(I definitely haven't seen anything like s,vcc,s,gnd,s so far)

[tggzzz]

I rarely (if ever) see connectors being arranged this way.
Vcc, Ground - is almost always on one end. (or opposite ends)
Why is that?

Unless there is a specific reason, ignorance plus laziness is sufficient.

It is rare to find a beginner, amateur (or software engineer!) who thinks of EM fields. Typically they also think that "ground" is a valid concept.

Is that the case of it simply "not mattering" in most hobbyist applications?

(I definitely haven't seen anything like s,vcc,s,gnd,s so far)

(ie. Arduino Uno pinout with whole line of signal lines and one sad single ground at the very end)

The issue is made less important by slower transition times and lower loads. Doesn't mean the issue isn't there, though.

[u666sa]

Consider geneous apple design, both macbooks and imacs and macs pros. GND, VCC, data
This way, if there is any liquid VCC goes onto data pins and kills the damn device. As a manufacturer you sell more and your income will be more. Makes perfect sense.

Correct way is VCC, GND, data
You put a fuse on VCC. That way in most cases fuse goes, and other cases vcc shorts to gnd, not touching data pins.

[Siwastaja]

Yes, decoupling capacitors at both ends between Vcc and GND and then you can pretty much treat Vcc as "return path" as well. Therefore, signal - gnd - signal - vcc - signal offers best shielding between the signals, and closest possible return path on any signal.

Why it's not done that way? Well, good enough is good enough. If the edge rates are not that high, then some sort of Vcc-GND-signal-signal-signal "feels" easier for the designer, or maybe offers real, actual layout simplifications. And if it works and passes EMC with flying colors, then why do anything else?

[Nominal Animal]

signal - gnd - signal - vcc - signal offers best shielding between the signals, and closest possible return path on any signal.  Why it's not done that way?

It is when the limits of the other orders are felt, but not before  .

The best example I can think of is 80-wire parallel ATA cables (UDMA/ATA-4); every second one is a ground wire, and in the round cables the pairs are even twisted, because the default 40-wire configuration had too much crosstalk.

Isn't the exact same thing a problem on PCBs?  With crosstalk between unrelated parallel signal wires, fixed by adding a "guard" ground trace in between?

[Siwastaja]

The best example I can think of is 80-wire parallel ATA cables (UDMA/ATA-4); every second one is a ground wire, and in the round cables the pairs are even twisted, because the default 40-wire configuration had too much crosstalk.

Isn't the exact same thing a problem on PCBs?  With crosstalk between unrelated parallel signal wires, fixed by adding a "guard" ground trace in between?

The original 40-pin IDE cable has crosstalk because of lack of close-by return path; the grounding of that pinout is seriously fucked up.

Single-layer PCB indeed has the same issue. But if you have a ground plane close (in a 4-layer stackup where prepreg thickness is something around 200um), guard traces in-between are usually not needed. I mean, long parallel tracks with minimum clearance between them may have crosstalk. But if you separate the traces as much to allow decent guard traces (they must have the width of the smallest annular ring of a via, because they need to be grounded and not flapping in the breeze, so this is some 0.6mm wide guard track + 2*0.2mm separation, so 1mm between the signal traces), this separation alone combined with much shorter distance to the ground plane (one fifth or so) removes the crosstalk even without actually laying out the guard traces.