Jump to content

Talk:Differential signalling/Archive 1

Page contents not supported in other languages.
From Wikipedia, the free encyclopedia
Archive 1

Merge interface articles

I would have included all sorts of other technologies under the topic of differential signaling, including RS-422, RS-485, ethernet and even MIDI. Is there some technical reason these have not been included in these collective topics yet? Is there a better term to use for the technology involved in these other major signalling technologies?

Charlie Richmond 23:05, 5 December 2005 (UTC)

So what do we do now? Charlie Richmond 11:35, 19 December 2005 (UTC)

Differential signaling just means you have two conductors and that you send your signal as a difference of voltages (or currents) rather than as the voltage (current) on a single wire (single ended singnaling). There are a host of circuit families that fall under the heading of differential signaling, including: current mode logic (CML), emitter coupled logic (ECL), positive supply emitter coupled logic (PECL), low-voltage ecl/pecl (LVECL/LVPECL), LVDS and the others mentioned (though I'm not an expert on RS-422, RS-485, or MIDI). These differences between these families have to do with where the termination is referenced, what the common-mode levels are set to, how large a differential signal amplitude is required, etc. The term "Differential signaling" is a generic topic which I think deserves its own entry. Citing examples like LVDS, ECL, CML, RS-422, etc might be useful. The latter, RS-422, is a signalling standard which describes an electrical interface and communications protocols. I don't see how it would be correct to merge the generic topic of "differential signaling" with any of the specific I/O standards cited.

It would be wrong to merge this article with LVDS. LVDS is a specific technology, whilst this article encompasses all forms of differential signalling. --Heron 17:51, 17 March 2006 (UTC)
As above, I say no merge. A link in the other page to here is fine. -- RevRagnarok 01:32, 3 May 2006 (UTC)
Agreed with Heron... it would be wrong to merge unless every type of differential signal was pulled into the article (also not a wise thing to do). There are lots of different types of differential signals, LVDS is just one type. Mrand 23:20, 7 June 2006 (UTC)

Two l's?

Single-ended signalling has two l's while Differential signaling only has one l. We need to make this consistent. 129.162.1.32 13:58, 16 October 2006 (UTC)

UK vs US spelling... This situation is rather contentious on Wikipedia, and it's unlikely there will be a standard any time soon (current policy is just, stick with whatever the article was originally written in unless the article is clearly written about a US or UK subject). -- mattb @ 2006-10-16T15:18Z

Accuracy

Balanced lines reject noise; not differential signals. As long as the lines are at the same impedance, they'll pick up the same noise, which can be rejected at the other end. The desired signal can be differential, only on one wire, or whatever.

"... only the common-mode impedance balance of the driver, line, and receiver play a role in noise or interference rejection. This noise or interference rejection property is independent of the presence of a desired differential signal. Therefore, it can make no difference whether the desired signal exists entirely on one line, as a greater voltage on one line than the other, or as equal voltage on both of them." - (IEC 602689-3:2001, page 111)Omegatron 02:03, 6 April 2007 (UTC)

I think I agree. I haven't visited this article for a while, so I haven't noticed its decline. It looks as if it is the result of a merge or two.
If you are right, which seems likely, then most of this article is BS. We are then left with these reasons for using DS:
  • No need for solid ground connection between source and load, since receiver is floating. The only requirement is to keep the signal within the common-mode range of the receiver.
  • The differential voltage is bipolar and the threshold voltage is zero. This makes the system work with high line attenuation, as long as attenuation is similar in both wires.
  • Differential swing is twice the swing in each wire, so you get more noise immunity from a given power supply rail voltage.
Did I miss any? There are lots of people out there claiming that DS cuts down on EMI (incoming and outgoing), e.g. this article by a 'recognized industry expert' (PDF) and this one from Lattice (PDF), but all these claims contradict your IEC reference and, if you read them carefully, are assuming that the line is balanced. --Heron 22:43, 7 April 2007 (UTC)

Yeah, you get 6 dB more signal (and therefore SNR) by sending differential signals, but it isn't the property that allows cancellation of noise. Parallel discussion on Talk:Balanced_line#Shielding. — Omegatron 19:49, 9 April 2007 (UTC)

Agreed. See you at Balanced line. --Heron 20:21, 9 April 2007 (UTC)

See Talk:Balanced_audio#Incorrect and Talk:Balanced_line#Accuracy for parallel discussion. — Omegatron 18:14, 22 May 2007 (UTC)

It says "Differential signaling is commonly transmitted on balanced lines, which offer advantages beyond those of differential signalling itself", but this still kind of implies that the differential signaling is responsible for the rejection. We should state this pretty clearly since it's a common misconception. — Omegatron 18:17, 22 May 2007 (UTC)

I've fixed the statement in this article, but I'm not going to do the same for the audio article because I have little practical experience in that area, so I would just be hypothesising. --Heron 20:11, 22 May 2007 (UTC)
Yeah. In light of this, though, I'm not really sure why audio circuits use differential signalling at all, besides the extra dynamic range allowed. — Omegatron 20:58, 22 May 2007 (UTC)
I suppose it allows rejection of common-mode signals in the output amps? Like if the output amp didn't have good power supply rejection?
And older circuits used (center-tapped?) output transformers, which the differential out tries to emulate for some compatibility reason? — Omegatron 21:00, 22 May 2007 (UTC)
Audio Balancing Issues by Graham Blyth of Soundcraft [1] is a good bit of debunking. He says that the point of differential balancing is that it keeps the source impedances of the two wires equal over the whole band, since they are both driven by similar amps, while a single-ended balanced line would only be accurately balanced at DC. --Heron 09:54, 23 May 2007 (UTC)
That's an excellent reference. Thanks. It seems like filter caps at the output (like RF caps from each output pin to chassis) would keep the impedance balanced at higher frequencies? But I can see it being marginally better in some circumstances.
As he points out, the fully-differential "cross-coupled" outputs will send the same signal level down an unbalanced cable that shorts one pin to ground. Although the "impedance balanced" output can also do this for a standard TS cable, it doesn't work if the hot is shorted, so interfacing with incorrectly-wired cables is also a very marginal benefit. — Omegatron 15:44, 23 May 2007 (UTC)

Can you look at Talk:Balanced_audio#Incorrect? (And also Talk:Faraday_cage#Mesh_size.3B_explanation.3B_magnetic_fields.) I'm not really fluent with things like "wave impedance". — Omegatron 15:16, 24 May 2007 (UTC)

OK. See you there. --Heron 15:20, 26 May 2007 (UTC)

I'd like to point out that for a single signal over long wires, it often doesn't matter if the transmitter starts out with single-ended (signal and ground) or differential signal (S+ and S- swinging in opposite directions). After the wires pass through a braid-breaker or choke (to reduce electromagnetic interference), the signal seen at the far end looks like a differential signal, one voltage swinging in the opposite direction as the other. (For example, many MIDI hardware *drivers* often have one "+5 V" line and another line that is driven single-ended 0 to +5 V. Other MIDI drivers have one "GND" line and a "signal" line that is driven 0 to +5 V. But on the other side of the the resistors and the choke, the voltages at the reciever -- for both kinds of drivers -- look like differential signals swinging in opposite directions).

I wonder if this patent has any relevance to the "differential signal" dispute: "Ground return for high speed digital signals that are capacitively coupled across a DC-isolated interface" United States Patent 6023202

it can make no difference

"... only the common-mode impedance balance of the driver, line, and receiver play a role in noise or interference rejection. This noise or interference rejection property is independent of the presence of a desired differential signal. Therefore, it can make no difference whether the desired signal exists entirely on one line, as a greater voltage on one line than the other, or as equal voltage on both of them." - (IEC 602689-3:2001, page 111)

Perhaps I'm taking this statement out of context, or perhaps I've been suckered into believing a myth, but the very last bit of this statement seems very wrong to me: "it can make no difference whether the desired signal exists ... as a greater voltage on one line than the other, or as equal voltage on both of them."

Often we have 2 balanced wires feeding into a receiver. And often the "desired signal" is some analog audio music. And often we get some unwanted outside noise that (because we designed the wires to be balanced) couples equally into the 2 wires.

If we were to hypothetically put "the desired signal ... as equal voltage on both of them", I don't see how it is possible to reject the noise. But if we do the usual thing and make "the desired signal ... as a greater voltage on one line than the other", then we can reject the (common-mode) noise with a standard differential amplifier. So it does make a difference. Right? --76.209.28.72 07:59, 16 June 2007 (UTC)

I assume they mean 'equal in amplitude but opposite in phase', otherwise it doesn't make sense. Does anyone have a full copy of the standard? I wonder if it's a misquote, or a mistranslation from the French, since the standard seems to be bilingual. --Heron 19:42, 16 June 2007 (UTC)

They mean the voltage is equal in amplitude (but opposite in polarity). A "greater voltage on one line than the other" (if the signal were on one wire and the other were connected to ground, for instance) is not typical in balanced systems; but still works as long as the impedances are balanced. — Omegatron 22:57, 17 June 2007 (UTC)

Diff Sig Diagram

So I had made this diagram for diff signaling because I thought it would help with understanding. Feel free to take out or replace with a better one if needed. Guerberj 20:05, 21 June 2007 (UTC)


Excellent. Thanks. — Omegatron 01:48, 22 June 2007 (UTC)

Clarification on high frequency signals: The article states an advantage of Diff Sig. over Single Ended is greater ability at high frequency. Gives example of coaxial cable as single ended but the link to coax describes it's ability at high frequency. My understanding is coax is superior at high frequency signaling due to such things as smaller variation in impedance. All seems a bit contradictory. Suggest some clarification by an expert (which I'm not). Stuzzo 20:35, 18 July 2007 (UTC)

Twice the noise immunity

I don't know enough about the area to change anything myself here but it would be useful to say what sort of noise differential signaling provides twice the protection from. Presumably if you include noise in the ground at different points in the device differential signaling provides more than twice the protection against this. It would also be useful to say exactly what this means. These may be obvious to people who know the field but not to me.

For instance maybe we could change the sentence to read something like, "At a given voltage differential signaling can tolerate twice as much noise induced in the lines." However, now that I think about it I would guess that some sort of statement about the shape of the noise (independent Gausians) might need to be made to avoid the situation where both wires experience perfectly negatively correlated noise. Anyway I don't really have a clue hence the reason to ask someone who does. 128.103.187.69 (talk) 19:15, 26 November 2007 (UTC)

Much more than twice - Common mode rejection ratio --195.137.93.171 (talk) 20:55, 30 December 2010 (UTC)

differential signaling in computers

The article currently claims "Differential signaling has to be used in computers".

Certainly some computers do use differential signaling in some places. But my understanding is that many computers don't use differential signaling anywhere.

What should we say about computers in this article? --68.0.124.33 (talk) 21:07, 26 April 2008 (UTC)

PCI Express uses it. Robijn (talk) 11:19, 8 January 2012 (UTC)

Semantics : Balanced lines reject noise; not differential signals ????

No - the noise is present equally on both lines. It's only when the signals on the two lines are subtracted (or 'differenced') that the noise cancels. eg at the 'inverting' and 'non-inverting' inputs of an operational amplifier. It's called Common-mode rejection ratio. So it is the differential between the signals that rejects the noise ! --195.137.93.171 (talk) 20:53, 30 December 2010 (UTC)

Wire 1 in = signal (+S)
Wire 2 in = signal (-S)
Wire 1 out = noise + signal (N+S)
Wire 2 out = noise - signal (N-S)
Detect (Wire 1 - Wire 2) = ((N+S) - (N-S)) = 2xS
It gets a bit philosophical or metaphysical - we can describe the same thing in different words, and argue about which words are correct.
I suppose a single-ended signal is like the sound of one hand clapping ?
--195.137.93.171 (talk) 21:40, 30 December 2010 (UTC)
And Twisted pair helps, too !--195.137.93.171 (talk) 21:44, 30 December 2010 (UTC)

Differential pair merge

I have proposed a merge from Differential pair into this article. The source article has significant overlap with this one and does not have enough material to stand on its own. --Kvng (talk) 13:02, 3 May 2012 (UTC)

Support - same topic, two titles, better to have one article than two stubs. --Wtshymanski (talk) 13:11, 3 May 2012 (UTC)
109.146.20.31 has recreated Differential pair - https://wiki.riteme.site/w/index.php?title=Differential_pair&oldid=609539942. I have reverted it back to a redirect and directed further discussion here. ~KvnG 23:00, 27 May 2014 (UTC)

No disadvantages?

The article only has a section about advantages. I think it is fair to also list some disadvantages.

For higher frequencies (wire length >> wavelength) differential signals are a big source of RFI. In general this is circumvented by moving to a three wire system +, - and GND. The latter begin a shield in many cases. So where is the benifit over a single wire + shield, ie coaxial cable.

Furthermore a three wire system can support two modes, the odd and even modes, (or in low-frequency terms: the differential and common modes) each mode will have different propagation delays, causing mode conversion to happen. Leading to all kinds of problems.

Next in a multilayer PCB it is almost impossible to create a real differential line (ie odd mode impedance much less than the even mode impedance). Most lines that are intended to be differential are in reality two time a single mode line. By having two single mode lines most benefits of a differential line are lost.

The main advantage of differential lines is that it allows people without a proper electromagnetic background to generate layouts with a somewhat decent performance. As soon as the line lengths become a substantial part of a wavelength, the real problems start. For normal PCB sizes up to roughly 1 or 2 GHz it can be used. Above that single ended designs will give much better and more reliable performance. No sane microwave engineer would think of using differential lines. — Preceding unsigned comment added by 86.81.172.19 (talk) 01:22, 29 January 2013 (UTC)

One more very simple disadvantage of differential signaling is increased cost and complexity of the transfer lines, and the input, output and processing electronics156.17.91.253 (talk) 15:00, 8 October 2013 (UTC)

Differential pair currently is redirected to this page (Differential signaling). However, this is a problem since there are two different meanings of "Differential Pair". The earlier meaning is a specific two-transistor circuit. The second meaning is a pair of conductors using differential signaling. I propose to change the Differential pair link to a disambiguation page to separate these two different meanings.

Some references for the first definition: Chapter 4 of Designing Analog Chips, analog.com, All About Circuits, lecture notes

KenShirriff (talk) 01:47, 16 September 2019 (UTC)

ensemble signalling

Currently the 64b/66b encoding article briefly mentions "Ensemble NRZ coding", linking to a section of this "differential signaling" article that does not currently exist (did it ever exist?).

Recently (in a series of edits around 2020 May 2 to 2020 May 4) a section on "ensemble signalling" / "ensemble signaling" was added to this article, alas without any references.

Is that "ensemble signaling" the same kind of signaling as the "Ensemble NRZ coding" mentioned in the "64b/66b encoding" article?

Is ensemble signaling (of either kind) actually used in any communication system? (A quick google search turned up a bunch of references to a musical term and a neuro-medical term that seem to be unrelated). --DavidCary (talk) 23:42, 5 May 2020 (UTC)

Confused advantages

Despite best efforts, I believe that some of the advantages attributed to differential signaling in this article are not in fact the result of differential signaling, but should rather be attributed to balanced lines. A quick example would be the caption beneath the first illustration: “Elimination of noise by using differential signaling.”. Yes, this image displays an application of differential signaling, but the noise cancellation shown is the result of the balanced line and differential amplifier, not the signaling scheme.

Another example is the very first paragraph under the Advantages heading: “Provided that the impedances of two conductors in a circuit are equal (it is balanced), external electromagnetic interference tends to affect both conductors identically. Since the receiving circuit only detects the difference between the wires, the technique resists electromagnetic noise compared to one conductor with an un-balanced reference (low-Ω connection to ground).”. This is none other than a description of the advantages of balanced lines and again has nothing to do with differential signaling. At the very least, there is a follow up statement immediately after: “Contrary to popular belief, differential signaling does not affect noise cancellation. Balanced lines with differential receivers will reject noise regardless of whether the signal is differential or single-ended, but since balanced line noise rejection requires a differential receiver anyway, differential signaling is often used on balanced lines. This improves SNR, reduces EMI, and makes the signal more immune to ground currents or differences.”, but without any proper citations or explanations, I think this is more confusing than enlightening.

With these issues in mind, I will try to improve this article to the best of my ability and differentiate the topics to promote understanding. Balanced lines and differential signaling have been confused for too long, and I believe we have some responsibility clear up some of this confusion if possible. Gutten på Hemsen (talk) 11:19, 1 September 2021 (UTC)

I’ve located a great point by point list of some advantages of driving lines with differential signaling and this list happened to be part of the Balanced audio article which really isn’t about differential signaling but rather about the application of balanced lines in audio systems. It just so happens to have a decent section on applications of differential signaling. Gutten på Hemsen (talk) 00:20, 2 September 2021 (UTC)

@Gutten på Hemsen: I have quibbles with your recent changes. You've removed points about immunity but have reinforced points about emissions. I believe the same caveats apply to emissions and immunity. I think it is important to explain the relationship between balanced line and differential signaling, but you have removed this discussion. What kind of balanced line does not use differential signaling? When is differential signaling applied to a non-balanced line? ~Kvng (talk) 17:43, 6 September 2021 (UTC)
Thanks for your reply! These are very valid concerns you bring up. Let me start with the relation to balanced lines:
I absolutely agree that the article should explain how differential signaling and balanced lines are related, but in my opinion, the previous version of the text included these references in very unstrategic places. For example, the section on advantages began with two whole paragraphs about balanced lines and how they protect the signals against noise and how they should not be confused with differential signaling. The way I see it, that made the text more confusing and less readable. Why shouldn’t the section on advantages start with actual advantages of differential signaling?
Differential signaling is a concept that is way too often confused with balanced lines (among amateurs and professionals alike), so in my opinion, having references to balanced lines in many places in this article is only going to strengthen the confusion. Instead, I suggest that we limit references to balanced lines to a single section where it’s rigorously defined and explained in such a way as to promote understanding of the topic and its relation to differential signaling. I’d be thrilled if you would like to cooperate on writing this section together with me.
Now onto another concern; not every balanced line uses differential signaling. The article on balanced circuits discusses how this is performed (have a look at fig. 7). As you probably know, the only requirement for a line to be considered balanced is that the two conductors have equal (balanced) impedances to the circuit’s reference point. The type of signal that the line is driven with, is besides the point.
Differential signals applied to unbalanced lines is a bit different; as you’ve probably guessed, it is indeed possible, but I cannot bring forwards any examples of this. If you were to connect an unbalanced cable (such as a coaxial cable) to a differential output, it would considered differential signaling on an unbalanced line, but I cannot think of a reason why anyone would want to do so.
In the end I think it is important to reinforce the fact that one doesn’t require the other, and that they are two entirely separate concepts, despite often appearing together. I think the last thing we want is another generation of people who never understood that balanced lines and differential signaling are two different things. My preferred distinction is that one concerns the electrical properties of the circuit, while the other concerns the method of signal transmission.
Finally, could you rephrase your concern about immunity and emission? I don’t understand what exactly it is that you’re saying there.
I’d be happy to receive suggestions on how to further improve the article. Thanks again! Gutten på Hemsen (talk) 19:00, 6 September 2021 (UTC)
Gutten på Hemsen, I believe readers are looking for practical information in these articles. I'm fine with explaining that these are separable concepts but, in practice, they are used together and that is even more deserving of explanation.
There is symmetry between emission and immunity. A balanced line with very good immunity also typically has low emissions. The same physics that cancels out emissions also can be used to cancel out common-mode interference. ~Kvng (talk) 17:04, 9 September 2021 (UTC)
I’ve started the section on balanced interfaces. Feedback is very welcome! Gutten på Hemsen (talk) 20:34, 6 September 2021 (UTC)
I added a short paragraph about balanced circuits to the introduction. Gutten på Hemsen (talk) 21:16, 6 September 2021 (UTC)
I haven't looked at all your edits yet, but I generally support your assertion that balanced line and differential signalling are completely separate concepts. In answer to Kvng, when I worked in telecomms (back in the stone age), it was passingly common to send an unbalanced telegraph signal down a twisted pair balanced line in situations where no unbalanced line was available. SpinningSpark 10:30, 8 September 2021 (UTC)
I’m basing my assertion on the writings of experts in the field: Neil Muncy and Bill Whitlock. The two of them worked together in the Audio Engineering Society’s standard working group on grounding and EMC practices and have been instrumental to the literature on grounding and interfacing ever since. The book I have in front of me is the Handbook for Sound Engineers where Whitlock has written a full chapter on grounding and interfacing. Gutten på Hemsen (talk) 14:17, 8 September 2021 (UTC)
Gutten på Hemsen, I am familiar with Muncy and Whitlock's work from attending their presentations at AES conventions and my work on AES standards. I don't have a copy of the book though.
Unless there are significant technical errors, I don't consider removing poorly organized material to be an improvement and I am inclined to revert such changes. We should leave that stuff in until we come up with a better presentation for the information ~Kvng (talk) 16:51, 9 September 2021 (UTC)
Kvng Thanks for the feedback. Have you had a look at the article since your previous comments? I’ve written a section on the relationship between differential signaling and balanced interfaces. I’ll also have to ask: would you say there is something important missing from the article at this point? Gutten på Hemsen (talk) 16:19, 11 September 2021 (UTC)
Spinningspark, that would be a single-ended signal over a balanced line in a pinch. Doesn't sound like something that would give any perfrmance advantage. ~Kvng (talk) 16:46, 9 September 2021 (UTC)
I never claimed it did have performance advantages (unless you particular want to increase earth loop currents). I was merely responding to your request for examples. It does, however, have significant economic advantages. Laying new cables is expensive, especially if you have to dig up roads. Twisted pair would never have become the de facto standard for LAN cabling if it were not for the ability in the early days to repurpose building telephone cable infrastructure. Telephone cables don't have performance advantages over coax, which was the previous most common way of doing it. SpinningSpark 17:19, 9 September 2021 (UTC)