Why do we not ALWAYS need to worry about cable impedances?

[Infraviolet]

I've been reading through Horowitz and Hill and have come to the section about cable impedances, capacitances/inductances/speed-of-light-lag. But it is fairly clear this only seems to count for cables of a few metres and more, and mostly only for coax. Certainly one doesn't think this sort of thing gets considered in designs when connecting up lower speed equipment with ribbon cables and other multi-wire connectors. And yet it would appear to be the speed of the rising/falling edges in a signal which give it the ability to generate reflected pulses and other nasty effects on the waveform, particularly things like the spikes of doubled or below ground voltage which can arise. And pretty much all chips thesedays are going to give very fast edges on their logic rises and falls, even if the datarate is slow and the time between a rise and the subsequent fall is relatively long.

To my understanding in practice there is a point in designing systems above which it is generally considered that things like terminations are needed, and below which such considerations are ignored. Where does this occur?

Clearly power wires, DC ones too, not just building-scale AC wiring, aren't usually planned with these impedance configurations, is tha just an assumption that there is always enough capacitance in the load, however distant it may be, to give a slow rise time and not trigger those huge positive and negative spikes?

And ribbon cables, as well as wire-to-board connected cable types with multiple separate wire conductors in them, don't come with any mentions of typical capacitances or impedances in their datasheets, which makes one assume that when they are used in a design of something the designers have gone and said "transmission line effects aren't going to matter here".

There are clearly quite a lot of scenarios where one can have several metres of wiring (for I2C or SPI or UART or CAN or USB within a larger robotic arm or around a car)and although the data rate might be slow, the edges are however fast they naturally are from the chips being used, but one doesn;t see much suggestion that transmission line effects play a major role in the design of these sort of things, while they do matter for superfast signals between test-bench kit. Are the transmission line effects magnified a lot by coax cabling of the type neded for keeping sensitive signals clear of environmental noise, and that is why they mainly crop up for coax not other cable types?

Can anyone explain where the dividing line is for considering these effects, and why they don't always get considered? If they did always matter one wouldn't even be able to wire a 10cm PCB trace without fear of voltage steps going well outside the receiving end device's voltage limits, because even for slow data rates the rise and fall from almost any logic is pretty fast indeed. What stops transmssion line effects from mattering except for high speed signals in coax, and extraordinarily high speed microwave type signals elsewhere?

[TimFox]

As a quick check, which should be followed by a more careful computation when caution is indicated, assume a propagation speed down the line (ribbon, wire, coax, trace, etc.) between 0.5 and 0.9 times c, then compare the electrical length (in seconds) with the rise time of the signal involved.  At 0.6 c, 18 cm = 7 in has an electrical length of 1 ns.  If the rise time is comparable to or shorter than the length, you should do the more detailed calculations found in textbooks such as AoE.
Good ribbon cables have published specifications for characteristic impedance, propagation speed, capacitance, and inductance;  they can carry differential (balanced) signals or be connected with alternate wires grounded to form single-ended (unbalanced) lines.
A typical 3M color-coded unshielded ribbon cable (0.050 in pitch) is well-specified:  https://multimedia.3m.com/mws/media/22071O/3mtm-color-coded-flat-cbl-05026-awg-stranded-pvc-3811-ts0122.pdf
In the ground-signal-ground-signal... configuration, the impedance in the datasheet is 95 ohms, speed 72% or 4.63 ns/m.
With 60 Hz building wiring, where the half-period is 8.3 msec, you might worry about reflections when the wire length approaches 2.5 km = 1.5 miles.

Ed:  2500 km = 1500 miles (see below).

[tggzzz]

...
Can anyone explain where the dividing line is for considering these effects, and why they don't always get considered? If they did always matter one wouldn't even be able to wire a 10cm PCB trace without fear of voltage steps going well outside the receiving end device's voltage limits, because even for slow data rates the rise and fall from almost any logic is pretty fast indeed. What stops transmssion line effects from mattering except for high speed signals in coax, and extraordinarily high speed microwave type signals elsewhere?

Once you start to ask questions like that, which you should ask, then you are one step away from finding these useful: https://www.edn.com/bogatins-rules-of-thumb/

[bdunham7]

With 60 Hz building wiring, where the half-period is 8.3 msec, you might worry about reflections when the wire length approaches 2.5 km = 1.5 miles.

How are you calculating that?  It seems off to me by about a factor of 100.

[TimFox]

Oops!  The correct answer is (8.3 msec) x (3 x 10⁸ m/sec) = 2.5 x 10⁶ m = 2500 km = 1500 miles.

[Terry Bites]

In many applications the line impedance and terminations don’t matter as much as the magnitude of the current or more usually voltage appearing at the receiving end. This is the case in most low frequncy small signal circuits. Signal intergity requires that the signal lines do not couple signals to or from adjacent lines. That does not always demand matched lines and terminations. Though it often helps at higher frequencies.

Impedance matching dericves from in the maximum power transfer theorem. It applies, as the name suggests, to the transmission of power. Power not dissipated in the load must be dissipated in the source.
This is undesirable in RF power systems. "Reflected" power may just melt your transmitter, fry the antenna and its transmission lines if not controlled. You dont want reflections in your logic lines to be seen as extra edges.The line capcitance is a given  i=CdV/dt and with a given voltage level , there's going more power being delivered into the line on a fast edge.

Your utility company doesn't want to be burning gigawatts just to satisfy a theorem. The maximum voltage (or minimum drop) criterion is what matters most.
High power AC and HVDV distribution systems do have manged impedances- The main purpose of this to control the effects of transients, a bit more than blips on a 750kV line!

[mag_therm]

I recall a nasty problem with a long track on a rather large ( physical dimension) LS TTL video memory board back in 1980's.
The problem, of course, was intermittent, and after boards were in production, and went away when the fastest scope or the logic analyser was even nearby let alone being connected.
Searching now, immediately this OnSemi legacy doc came up. The "0.15 nanosec per inch" problem is mentioned on page 18 and 19
https://www.slideshare.net/windelcito/74ls

But there were more detailed application notes at the time including some actual waveforms showing the staircase sending and receiving ends as the transitions bounced back and forth.

[David Hess]

When the transmission line is not terminated, it appears as a lumped capacitance at low frequencies.  At high frequencies, reflections in the transmission line reflect back and forth and produce the *same* exponential charging curve as the lumped capacitance would produce.  So if the signal transition time is slower than the equivalent RC time constant, termination is not required.

At higher frequencies, the exponential RC charging waveform is produced from the reflections if no termination is present, and if the termination is present, then the transition time of the transmission line is the same as the signal source.

In both cases longer unterminated transmission lines have lower bandwidth with low frequency signals because of capacitance, and lower bandwidth with high frequency signals because of reflections, with the results being the same.