8

I am a bit confused about this Wikipedia article about Category 5 cables.

It states:

The cable standard provides performance of up to 100 MHz and is suitable for most varieties of Ethernet over twisted pair up to 2.5GBASE-T.

How is it possible to transmit data with a speed of 2.5 gigabit/s when using a frequency of 100 MHz? Wouldn't you need at least 2.5 GHz?

Peter Mortensen
  • 1,676
  • 3
  • 17
  • 23
Zciurus
  • 331
  • 2
  • 10
  • 8
    "it uses a frequency of 100 MHz", no a *bandwidth* of 100 MHz, and not even that is correct. – Marcus Müller Aug 03 '21 at 08:18
  • See https://en.wikipedia.org/wiki/Line_code – NStorm Aug 03 '21 at 08:33
  • @MarcusMüller Ohh, I see. So there's not *the* signal beeing transmitted at 100 MHz, there are multiple signals beeing transmitted simultaniously over the same pair at different frequencies, correct? – Zciurus Aug 03 '21 at 10:35
  • @Zciurus-Alt-Del no. Any transmission of data has a *bandwidth*, not just a single frequency. – Marcus Müller Aug 03 '21 at 10:38
  • @Marcus Muller: The cabling standard TIA-EIA-568 specifies performance of a cable type at specific sinusoidal frequencies, up to 16 MHz for Cat 3 and up to 100 MHz for Cat 5, for example. There is nothing in that document that specifies performance over a bandwidth. – SteveSh Aug 03 '21 at 14:21
  • 1
    @SteveSh the question asked was how to transmit data *using a frequency*, so that's what that I commented on, not the cabling standard (which indeed only guarantees any particular behaviour on frequencies, but it does limit the delay spectrum, so that's inherently a guarantee over a bandwidth). – Marcus Müller Aug 03 '21 at 15:14
  • 5
    basically the same as this question yesterday: [How am I getting 1 Gbps through a CAT3 cable?](https://electronics.stackexchange.com/q/579895/27052) – phuclv Aug 04 '21 at 02:18

3 Answers3

33

No, you don't need 2.5 GHz bandwidth to transmit 2.5 Gbit/s. You can do it of course, but that would be very inefficient.

The data is encoded into symbols.

100Base-T Ethernet uses one pair in one direction and uses MLT-3 encoding to convert 4B5B encoded 125 Mbit/s bit stream into a signal that has 31.25 MHz bandwidth, so it uses 3.2 bits per hertz per channel.

2.5Base-T Ethernet uses all four pairs and complex encoding methods to achieve 6.25 bits per hertz per channel to end up being 100 MHz bandwidth for which the Cat 5e cable is rated for.

The encoding method includes the use of PAM-16 modulation, which means that each symbol on each wire pair is one of 16 voltage levels and thus carries four bits per symbol. It also uses Tomlinson-Harashima precoding (THP) precoding and DSQ128 encoding to achieve the goal of getting 2.5 Gbit/s sent over four pairs with 100 MHz bandwidth per pair, or 400 MHz total bandwidth for the whole link.

Peter Mortensen
  • 1,676
  • 3
  • 17
  • 23
Justme
  • 127,425
  • 3
  • 97
  • 261
  • How did you arrive at the number of 6.25 bits per channel per Hz? Shouldn't it be 12.5? Or is 2.5Base-T half duplex? – jaskij Aug 03 '21 at 23:39
  • 5
    Somehow I never realized this before, but it looks like it's full duplex using all four pairs in both directions: http://www.qdpma.com/GigabitEthernet.html . That sounds wacky, but it's no different from an ordinary telephone line. You just have to filter your own transmissions from reaching your receiver, which is fairly straightforward. – Glenn Willen Aug 04 '21 at 01:31
  • 4
    AIUI, the bandwidth limits of CAT5e are part of why 2.5GBase-T even exists, it provides an acceptable improvement over 1GBazse-T that (usually) can use existing 1GBase-T infrastructure, thus allowing some degree of upgrade with limited up-front costs. – Austin Hemmelgarn Aug 04 '21 at 02:03
  • @AustinHemmelgarn it’s also for longer runs of cat6 and it is at a sweetspot where internet uplink or wlan might be over 1GE capacity but below 10. (but the uptake was not reflecting that) – eckes Aug 04 '21 at 07:34
  • @JanDorniak I looked some docs and calculations add up. All four pairs are used simultaneously in both directions in the link, so yes, it's full duplex. – Justme Aug 04 '21 at 07:38
  • @Justme this is what I was missing: https://networkengineering.stackexchange.com/questions/56042/how-1gbe-works-in-full-duplex - seems like there are some shenanigans so that each pair by itself is full duplex. No idea how it works, but yes, your numbers do check out in this case. – jaskij Aug 04 '21 at 08:22
  • how are there fractional bits per Hz? – Dean MacGregor Aug 04 '21 at 11:46
  • 3
    @DeanMacGregor Are you specifically asking how is it possible to have fractional bits per Hz, or why there is? It's a calculated value from total link speed (2.5Gbps) divided by number of wire pairs (4) which gives 625 Mbps for single pair. That is sent over 100 MHz bandwidth in wire. 6.25 bits per channel per Hz. – Justme Aug 04 '21 at 12:09
  • Does the fraction come from dividing by the number of wires? That would mean it's 25 bits per hertz for the combination of the wires? – Dean MacGregor Aug 04 '21 at 12:21
  • @DeanMacGregor Total bandwidth for all four pairs of 100 MHz each will total 25 bits per Hz. Or 6.25 bits per Hz over a 400 MHz bandwidth of all pairs total. It's just a number. 1000Base-T uses 4 bits/sec/Hz. – Justme Aug 04 '21 at 13:02
11

The rate at which you can communicate information through any channel (like a CAT5 link) is determined by its channel capacity

For normal channels like CAT5, this is well approximated by the Shannon-Hartley Theorem. This determines the channel capacity for a given bandwidth in the presence of noise as:

C = Wlog2(1 + S/N)

... where C is the channel capacity in bits/s, W is the bandwidth, and S/N is signal power divided by noise power.

So you can see that the amount of information that you can transmit is proportional to bandwidth, but also increases according to the clarity of the channel and the power of the transmitter.

100 MHz corresponds to 100 Mb/s only when the noise is as powerful as the signal. Cat 5 cable is designed to minimize high-frequency noise, so you can transmit a lot of data at the higher frequencies. It also has 4 pairs that you can use independently.

Matt Timmermans
  • 1,358
  • 6
  • 7
5

Thanks to @Justme and @MarcusMüller for pointing me in the right direction. I found this video by Ben Eater which answers my question quite well.

Basically, the frequency of 100 MHz is not directly related to the bits pre second that get transmitted. I assumed there was one signal (a square wave or something) per pair, which is not the case. Instead, sine wave modulation is used to transmit "symbols", which can be just ones and zeros (that's the case for slow transmissions like 10BaseT), but for gigabit ethernet, a symbol can have more than two states. This is part of the Line Code that @Nstorm mentioned. See this table for which ethernet variant uses which line code. The symbols that get transmitted per second are called the "baud rate". The 100 MHz bandwith limits how "sophisticated" a symbol can be.

Feel free to edit this answer if I understood something wrong.

Zciurus
  • 331
  • 2
  • 10
  • 3
    bps was divorced from Hz forty years ago, using Information Theory back when 300 baud modems were Really Expensive. – RonJohn Aug 04 '21 at 19:08