Custom PC - UK (2020-08)

far you can go. The PATA standard for hard
drives used 40 data conductors per cable and
needed another 40 ground wires in between
them to help deal with interference and
crosstalk between the data lines. That’s a hell
of a lot of cable to ensure that every conductor
is within tolerance and isn’t damaged.
Finally, there’s the inherent lack of
versatility in a pure parallel system. If we
look at the old PCI standard used to connect
graphics cards and other components to
a motherboard, it used its 32 data lanes
to connect all components on the bus.
Components would then send interrupt
requests to signal that they required the bus
to send or receive some data, and duly the
connection was opened up for them to use.
That’s fine if every component is only
occasionally using the bus, or is far slower
than the bus, which was the case with old
hard drives, floppy drives and parallel printer
cables. However, if data is constantly flowing
back and forth – and especially where real-
time feedback is important, as with graphics
cards – you can quickly find a huge backlog
of interrupt requests and components being
held up waiting for access.

The surge of serial
We’ve established that parallel
communications really do have some
issues, but up until relatively recently we
were still stick with slow serial connections,
so what’s changed? Well, silicon got better.
Good old Moore’s law played its part, with
huge advancements in processor speed
and complexity allowing for ever faster clock
speeds to be employed, along with ever
greater error correction and sophistication in
the way data is handled.
It’s now relatively cheap and easy to make
a very low-power, but blisteringly fast, USB
controller chip that runs at 5GHz, for instance.

You can have a large error correcting overhead

and still end up with a faster connection than if

you were to opt for a parallel system.

Another key change is the adoption of low

voltage differential signalling (LVDS). This is

where, instead of having a single wire whose

voltage fluctuates between two relatively

high or far apart voltages (-5V to +5V for

instance), you have two wires, each of which

carries the same signal but in mirror image to

each other and at a lower voltage. Such cables

are generally twisted together, like you’ll often

see in an Ethernet cable, and they’re known as

a twisted pair.

There are several advantages to this

approach. The lower voltage means the

system uses less power than high-voltage

systems with a data signal cable and a ground

cable. Also, the two cables help to create

a stronger signal that’s less susceptible

to interference, as the equal and opposite

current flow in the two wires results in equal

and opposite electromagnetic fields that

cancel each other out. What’s more, any

electromagnetic noise interference that does

affect the line will equally affect each wire, so

it can be detected and ignored by the receiver.

One disadvantage of LVDS is that, just like

parallel communication, you’re relying on two

signals arriving in sync when propagating

down two different wires, so cables and

PCB traces need to take this into account.

However, because you’re only dealing with

at most two conductors – not the 40 of PATA

or 32 of PCI – it’s much easier to achieve

this synchronisation.

So, the fundamentals of data transfer boil

down to parallel and serial communications,

but there's still a stack of different

technologies, some parallel, some serial. Why

are there so many? From here on, we’ll break

it all down and explain some of the reasoning

behind different interconnect technologies.

USB

While it may not seem central to the overall

performance of a PC, USB was the original

poster child for modern applications of serial

communications, and it’s utterly ubiquitous in

our lives now. It preceded all the other moves

from legacy parallel comms to modern serial

comms by many years, so it's the natural place

for us to start showing the way forward for

this modern new communication approach.

Prior to USB’s arrival, if we wanted to

connect just about any external device to

our PCs, we were stuck with either relatively

speedy (but wide and cumbersome) parallel

cables or deathly slow serial cables. With USB,

we suddenly had a compact, convenient and

highly versatile connection standard.

Fundamentally, though, USB still uses

the same serial communication principle

as those old serial connections. The original

USB standard uses two pairs of LVDS wires

for transmitting and receiving data at the

same time, and data is packaged in much the

same frame-based way as RS232. However,

improvements in semiconductor production

have allowed for faster, more sophisticated

chips to be embedded in even the most

rudimentary devices. This allows for more

control and complexity in the way the system

manages its connected devices, as well as

faster data rates.

Specifically, while older serial comms

used essentially direct lines between the

system and the connected device, USB has

an intermediary layer that requires software

control and distribution of the signal. That’s why

our modern computers can have half a dozen

USB ports on the back of the motherboard,

and a single controller can support up to 127

devices (power concerns aside), all with easy

plug-and-play compatibility. Conversely, you

were lucky to get more than two serial ports on

older computers.

Four twisted pair LVDS cables make up a UTP
network cable

USB has gone through umpteen iterations but has maintained backwards compatibility throughout

USB

type A

USB

type B

USB

type C

USB

mini A

USB

mini B

USB

micro A

USB

micro B

USB micro B

super speeed