What ultimately determines the speed of electrical media

I have heard several myths about this. Some have said "thicker wires mean faster speed" which I don't believe. Others have said different voltages and others have said higher frequencies.

Take Coax for example. Several years ago connections of 5mbit/sec were common, now today I am starting to see 100mbit/sec connections over the same physical medium of 5 years. Same goes with twisted pair (cat5e vs cat6), one can now get 10Gbit/sec over copper. DSL vs 56k. Once again these are the same physical medium yet ADSL is sufficiently faster.

With that being said why are cable (coax) speeds always faster than DSL speeds (twisted pair)?

My question is essentially what determines the limits of speed of any particular physical medium

• Impedance, reflections, and latency. Aug 31, 2013 at 13:43
• You need to realize that we rarely (if ever) achieve the speed that is theoretically achievable for a given physical medium. The data rate is further limited by the electronics that transmit and receive the data, as well as the modulation scheme. Over time we develop better electronics and modulation, so we get faster data rates for a given medium (56k modem vs. DSL). Aug 31, 2013 at 14:22

There's two things you might call speed: bandwidth and latency.

Latency is the duration of time needed for a signal at one node of the network to reach another node of the network. Processing time for the electronics to packetize the signal and place it on the wire often dominates the latency, but the physical medium does also affect it. As far as the physical medium is concerned latency is largely determined by the distance between the nodes (measured along the actual connecting cables between them), and the propagation velocity of the transmission lines.

The propagation velocity of the transmission lines is determined mainly by the dielectric material between the conductors. It is typically between 1/2 and 3/4 c.

Bandwidth (as used in the field of networking) is the channel capacity, or amount of data that can be delivered by the system in a given time. For example, "100 megabits per second". The physical bandwidth capability of a transmission network is determined by the Shannon-Hartley equation

$C = B \log_2\left(1+\frac{S}{N}\right)$

C is the channel capacity, B is the analog bandwidth (in Hz) of the system, and $S/N$ is the signal to noise ratio (SNR) of the receiver.

The SNR term is generally determined by the receiver electronics, though losses in the transmission medium will reduce the S portion of SNR.

B is the term where the physical medium has the most effect. Most media can be characterized by a bandwidth-distance product. Meaning a certain medium can carry more bandwidth if the distance is shorter.

Many of the technological advances that increased the network bandwidth (C) were achieved by using coding to take advantage of SNR to deliver more bits through cables with the same analog bandwidth (B). This was a major driver of advances in telephone modem speeds from 1200 to 56k baud, for example. These codes provide both error correction (to deal with the occasional incorrectly received bit) and equalization (to maximize use of the analog bandwidth of the medium).

10 Gb/s ethernet transmission over twisted pair is achieved by using 4 physical transmission lines per link, and by keeping the distance short (15 m), as well as by appropriate coding techniques.

• May I summarize that: Coax has a higher SNR ratio and thus a higher data bandwidth because of its resistance to crosstalk and it's ability to more accurately propagate rapidly changing, high frequency signals used to expand its data capacity through high frequency coding. Sep 1, 2013 at 2:44
• @AndyzSmith I think coax was used in the early days of Ethernet because the only twisted-pair widely available at the time was telephone grade, and modulation techniques to achieve high bandwidths in this cable (as in DSL) weren't known yet. When twisted-pair manufacturing was improved to create Cat5 cable, it became a new ball-game and Ethernet moved to point-to-point instead of multidrop topologies. Sep 1, 2013 at 2:53
• So, although a bit stretching it, better cables, some might say thicker cables, were a part of the improvement. Sep 1, 2013 at 2:56
• @AndyzSmith, improving cables has been part of the development of Ethernet from 2 to 10 Mb/s and from 100 Mb/s to 10 Gb/s. Whether those cables were thicker or not has little to do with it. The improvement from 1200 to 56k baud to ADSL over the installed plant of the telephony network was due largely to improvements in coding. Sep 1, 2013 at 3:01
• Agreed, just trying to clarify regarding OP's question. Could you characterize the improvements in physical cabling beyond 'manufacturing was improved to create Cat5 cable'? Sep 1, 2013 at 3:07