Talk:Differential signalling

The crosstalk claim

@SpinningSpark: These last few days I’ve been thinking more and more about the claim that differential signaling reduces crosstalk. The current version of the article mentions this in two places (neither cite any source). First one is in the introduction: “Any signals radiated by the conductors tend to cancel out resulting in reduced emission that could affect nearby circuits.”. The second one is in the Advantages section: “The electromagnetic field around a differential line is ideally zero, which reduces crosstalk into adjacent cables, useful for telephone pairs.”. What I cannot understand is how differential signaling has any effect here. Wouldn’t this be true of any two wire circuit where the conductors are close together? Even in a single-ended system, the signal currents in the wires are always moving in opposite directions so the magnetic fields should cancel each other out, right? Hopefully you could reduce my ignorance regarding this topic :-) Gutten på Hemsen (talk) 00:49, 28 December 2021 (UTC) If my assumption is correct, perhaps reduced crosstalk an advantage of twisted-pair or twin-lead rather than differential signaling. Gutten på Hemsen (talk) 12:35, 28 December 2021 (UTC)

Are you pinging me because I wrote some of that? My apologies if I have got something wrong. Reduction in crosstalk is largely a result of balanced lines rather than differential signalling as I'm sure you already know. This may be yet another case of confusion between the two because they are so often used together. It is the common return of single-ended signalling that causes the crosstalk problem because the return current is creating the (what would otherwise be the cancelling) field in a different place. Differential signalling implies a pair of conductors. Separate return conductors is bound to improve crosstalk over a common return conductor system, even if the pair of conductors is not balanced. That really has nothing directly to do with differential signalling since non-differential would have the same advantages in those circumstances. SpinningSpark 11:50, 29 December 2021 (UTC)

The ping was simply because I value your input on the matter. I believe you are one of the most active editors in this field. Now, who wrote what doesn’t really interest me, I’m more interested in getting the facts right and you’ve been a great help these past months. So thank you for that! As for the matter at hand: it seems my gut feeling was correct, which is always nice. You’ve perfectly summed up the idea: separate return conductors are what reduces crosstalk, not differential signaling. I’ll go ahead and remove reduced crosstalk from the list of advantages and walk away from this with more knowledge than before :-) Gutten på Hemsen (talk) 14:12, 29 December 2021 (UTC)

Ensemble signalling

@Kvng: Do you actually know anything about this? I can't find anything in sources, at least not under this name. It looks like a mathematical plaything to me rather than anything that ever had a practical use, if indeed it is not pure WP:OR. It's kind of reminiscent of stacked phantom circuits which can indeed provide n-1 sigals from n wires (if n is an even power of two). But this was never a practical proposition beyond the first layer of phantoms. SpinningSpark 22:38, 29 January 2022 (UTC)

No I don't and I should have noticed it was unsourced before creating redirects for it. ~Kvng (talk) 15:26, 30 January 2022 (UTC)

In that case I propose removing it. SpinningSpark 18:03, 30 January 2022 (UTC)

Probably a good idea. The material was added by an anonymous editor and no edit description or references were included. This is not the easiest topic to research and I haven't spent a lot of time on it but I haven't been able to find any sources to support the content. ~Kvng (talk) 18:20, 2 February 2022 (UTC)

Uses of differential pairs

This section is missing sources and the concept isn’t introduced with a formal definition. I’m sure it’s relevant to the article but as it currently stands, it’s unclear to readers what is actually discussed. Gutten på Hemsen (talk) 10:58, 5 February 2022 (UTC)

Revert

@Kvng: Hi could you elaborate on why you reverted my edits? Thanks! Gutten på Hemsen (talk) 21:58, 10 February 2022 (UTC)

His point is that once a signal has been put through a transformer, it is no longer referenced to anything, ground or otherwise (unless one conductor is then connected to ground). His objection is that by describing the signal in terms of a reference voltage (as your edit did) does not cover this case. However, I would point out that with a transformer coupled output there is no longer any difference between a differential and a single-ended signal. They are both floating signals. SpinningSpark 17:40, 11 February 2022 (UTC)

Now I will ask for elaboration. Doesn't a single-ended signal always have a ground reference? ~Kvng (talk) 14:15, 14 February 2022 (UTC)

Well it always has a reference, not necessarily ground. If you are about to say that a single-ended signal put through a transformer should not be called single-ended on the output side, then I would agree. SpinningSpark 14:45, 14 February 2022 (UTC)

This conversation feels ungrounded. I hope @Gutten på Hemsen got a satisfactory answer. If not, let us know and we will try again. ~Kvng (talk) 18:02, 14 February 2022 (UTC)

Sorry, I’m confused. Could you describe which parts of my edits were incorrect? Thanks. Gutten på Hemsen (talk) 21:55, 14 February 2022 (UTC)

You added (both with respect to a reference point such as ground). The signals only need to reference each other. Some differential transmitters use a ground reference for each signal but a transformer-coupled transmitter does not. ~Kvng (talk) 23:43, 14 February 2022 (UTC)

As SpinningSpark pointed out, once a signal passes through a transformer it's no longer differential or single-ended – it's floating. Without a reference point (such as ground) how are you to determine the signaling scheme? Gutten på Hemsen (talk) 11:24, 15 February 2022 (UTC)

We will need more than a parenthetical in the lead to explain this situation. I don't see any support in the body for this. Readers will be confused and even reading this talk page won't help them much. I clearly could use some schooling on this topic so go ahead and write something I can understand. ~Kvng (talk) 15:56, 15 February 2022 (UTC)

Good idea; I could write a section on transformer coupled transmitters. Gutten på Hemsen (talk) 20:15, 15 February 2022 (UTC)

What is the difference between noise rejection and noise reduction?

I'd like to correct article Adam Stankiewicz (talk) 20:18, 20 September 2022 (UTC)

My male answer syndrome is telling me that rejection is keeping the noise out in the first place and reduction is removing it once it has already been introduced. We need to find a citation. ~Kvng (talk) 14:32, 23 September 2022 (UTC)

Magnitude

According to Magnitude (mathematics)#Numbers, The magnitude of any number $x$ is usually called its absolute value or modulus, denoted by $|x|$. By that definition, magnitude does not capture polarity and Gutten på Hemsen has a point with this edit. SpinningSpark 09:38, 19 December 2022 (UTC)

polarity errors vs swapped-wire errors

Perhaps I'm overthinking it, but I feel that connecting 2 pins of a differential pair transmitter on one device to 2 pins of a differential pair receiver on another device, but getting the wires swapped, is not a "polarity error" -- the signals on those wires have the correct polarity, it's just that they got physically swapped. I imagine that some people would argue that there's no measurable difference at the receiver between swapping the 2 wires, vs. both wires having inverted polarity, so they are actually the same thing, and I feel that's a reasonable enough argument that I wouldn't disagree. But I am baffled at one edit that seems to be saying that getting 2 wires swapped *is* *not* a "swapped-wire error", it's actually a "polarity error". --DavidCary (talk) 18:22, 14 June 2023 (UTC)

Swapped wires causes a polarity error. The cause and effect are commonly conflated. The edit you link to is good because not caring about polarity fixes any polarity error regardless of origin. ~Kvng (talk) 12:53, 20 June 2023 (UTC)