Electronics > Projects, Designs, and Technical Stuff

Twisted pair, effect on non balanced/differential systems?

<< < (4/5) > >>

bsdphk:

--- Quote from: duak on June 13, 2020, 08:40:14 pm ---The ultimate in twisted pairs was worked out by the Telcos.  I remember driving along an interstate in New Mexico that was parallel to an old telephone line that used open wires strung on doorknob insulators on crossbar poles.  I'd say there were maybe 16 pairs in some sections.  If you followed each pair you would see there usually only two doorknob insulators but sometimes there were four where the wires crossed over.  Not all pairs crossed over at same pole but there was a pattern to it.  If memory serves it's a Walsh function.

--- End quote ---

For more info, see:  https://archive.org/details/bstj15-4-1/page/n7/mode/2up

MFX:

--- Quote from: grouchobyte on June 14, 2020, 02:56:41 am ---
--- Quote from: bson on June 14, 2020, 01:51:36 am ---
--- Quote from: grouchobyte on June 13, 2020, 05:40:30 pm ---Bottom line: YES, twisted pair to connect to something like a switch does provide marginally better noise immunity. Quantifying that is a little more involved.

--- End quote ---
Well, he didn't ask for it quantified, it was only a simple question and he only wanted a simple answer.  No reason to start thinking.

--- End quote ---

Actually he did ask us to quantify it.... read his question.......

Does using twisted pair have any noise pickup reduction benefit in this case and if so how much (roughly) benefit compared to a proper differential system?

--- End quote ---

"(roughly)" generally implies rough finger in the air estimate, sorry if that wasn't clearer.

SuperFungus:
The "rough" general rule of thumb is: the closer you bring a current to it's return (twisted pair) the less it radiates and the less susceptible it is to interference.

Further quantifying the benefit is no longer "simple" because it can be incredibly sensitive to every implementation detail.  I know it's not satisfying, but to give you an idea of what I'm talking about I've tightened a nut from "snug" to "tight" on a bulkhead connector and gotten ~10dB better radiated EMI performance (which itself is only true because the whole thing was very quiet to begin with).  In comparison the variability inherent in a twisted pair vs non twisted pair is almost infinite.  How tight is the twist?  How thick is the insulation?  How do the wires run in the "non-twisted" comparison?  What frequency are you making the comparison at?  You could compare a specific twisted pair to a specific different wire configuration (generally by building both and measuring), but anyone who you find who thinks they can give you a number for this that applies "in general" is fooling themselves.

T3sl4co1l:
It helps even on single ended signals, when the aim is towards fair signal quality, rather than RF emissions for example.

Consider a short length of twisted pair over ground plane.  The transmission line has two modes, CM and DM.  Because it's used unbalanced, the modes are coupled to some extent.  With one wire grounded at both ends, and the other wire the signal, that coupling is about 50%.  That is, the common mode is simply (V(signal) + V(GND)) / 2.

Since the TL is short, it won't be exposed to too much interference, nor radiate too much itself; if it were long, CM would in turn couple with ambient fields, i.e. a lot of signal evaporates into space, bad news.  Evidently, about half, which isn't good news even for digital signals.

Note that TL length is relative to the signal, its analog bandwidth or digital edge transition rate (not the clock rate!).

So, since CM loss depends on length and balance, if we have a short, unbalanced stretch, it's not a big deal, and we won't lose a huge amount of signal.

For example, 1dB of signal loss is merely 11% amplitude, likely no problem for a digital signal.  If the loss is entirely radiated, that 11% lost is -20dB to the signal, and if the signal is some 120dBuV (i.e., about a volt), that's 100dBuV of emissions -- not a great start when your limit is, whatever, 60dBuV or so!

In simpler terms: if you lose say 1% of a signal, that's 99% still making it through, which isn't much bother to the intended signal path.  And it's only 1% leaking away, who cares, right?  Trouble is when the signal is large enough, relative to some emissions limit, that even that 1% causes further problems.

For a mechanical analogy perhaps -- consider the power output of a car engine.  100kW say comes out the drive shaft, and, Idunno, 200kW comes out of the exhaust pipe thermally and another 50kW is soaked up by the coolant, and so on.  Well, heat isn't a bother, it's fine as long as it has somewhere to go (hence the tailpipe and the radiator).  But acoustic energy is a problem.  It's literally illegal (in many places) to operate a vehicle without a muffler, it's irritating for bystanders and it's damaging to the driver!  It might only be a few hundred or thousand watts of actual acoustic power -- a tiny fraction of the total power, but one that, if not dealt with, becomes a big problem, for everyone!

The engine is probably perfectly fine either way, with or without a muffler; it won't make much difference to overall efficiency, or maximum power output (analogous to signal quality, as it were).  The pipes themselves, might need to be tuned for best operation -- particularly true of two-cycle engines; a more direct signal quality analogy.  But after that, what happens to the noise leaving the vehicle doesn't matter much to what's inside, so you can get away with various things; heh, well, I'm not sure there's an analogy for coax vs. twisted pair in terms of acoustics.  But in any case, in electronics, we get concerned when the signal quality must be especially good, and when the emissions must be low.

Tim

David Hess:
Analog Devices discussed the subject in their application note 347 - Shielding and Guarding:


https://www.analog.com/media/en/technical-documentation/application-notes/41727248AN_347.pdf

Navigation

[0] Message Index

[#] Next page

[*] Previous page

There was an error while thanking
Thanking...
Go to full version
Powered by SMFPacks Advanced Attachments Uploader Mod