This is only a preview of the January 2024 issue of Practical Electronics. You can view 0 of the 72 pages in the full issue. Articles in this series:
Items relevant to "Active Mains Soft Starter":
Items relevant to "ADVANCED SMD TEST TWEEZERS":
Items relevant to "Active Subwoofer For Hi-Fi at Home":
Articles in this series:
|
Oscillating onions,
Batman!
Techno Talk
Max the Magnificent
The thought that we are now capable of creating multi-billion-transistor semiconductor devices
with structures whose sizes are measured in billionths of a meter makes my eyes water. I’m too
young for all this excitement!
I
n my previous Techno Talk
column (PE, December 2023), I cogitated on the concept of Precision Time
Protocol (PTP), a.k.a. IEEE 1588, used to
synchronise the nodes forming a packetbased network with an accuracy in the
sub-microsecond range.
The way this works is that somewhere
in the network is a grandmaster clock –
which typically obtains its time from
some GNSS (global navigation satellite
system) source – that propagates its concept of time throughout the network. One
thing we didn’t discuss was the fact that
each node in the network maintains its
own local time-of-day (ToD) value, as
part of which it employs an oscillator, but
what sort of oscillator might it employ?
Oscillating onions, Batman!
To be honest, there are more layers to
this onion than you might imagine. We
start with a resonator, which is a passive
device, such as a quartz crystal, that vibrates at a fixed frequency (its resonant
frequency). The next step up is an oscillator, which is an active device that
combines a resonator with an oscillation
circuit to generate a clock signal. The
first quartz-based crystal oscillator (XO)
was built by Walter Cady in 1921, more
than 100 years ago as I pen these words.
Now, this is where things get interesting. The typical frequency stability
variation over temperature of quartzbased XOs is between ±10 and ±100
parts-per-million (ppm). This isn’t too
shabby and will satisfy a wide variety of
use cases, but it’s insufficient for many
of today’s more demanding applications.
The next step up are TCXOs (temperature-compensated crystal oscillators),
which typically have frequency stability of ±0.05 ppm to ±5 ppm over their
operating temperature range.
For those who demand even more, we
have OCXOs (oven-controlled crystal oscillators) that achieve high stability by
encasing the crystal along with temperature-sensing and compensation circuits
inside a heated metal enclosure to create
a miniature ‘oven’ with a relatively constant temperature. In this case, we can
achieve frequency stability in the range
of ±0.5 to ±20 parts per billion (ppb).
8
Ovens don’t cool things down
When you think about it, an oven can
only heat things up (it can’t cool things
down). This means the inside of the
OCXO’s oven must be maintained at a
higher temperature than the outside ambient temperature (‘duh’). What does this
mean in these days of climate change in
which a temperature of 40.3°C was recorded at Coningsby, Lincolnshire, on 19
July 2022 (a temperature of 53.9°C was
recorded in Death Valley, California, on
16 July 2023)?
Well, fear not, because we are talking
about oven temperatures around 75°C.
If the outside temperature ever exceeds
this value, keeping accurate time will be
the least of our problems.
A rose by any other name
The first quartz-based OCXO was created
in 1929 and this legacy technology is still
ticking along (pun intended) to this day.
Having said this, although quartz resonators remain the mainstay of the oscillation
industry, devices using other materials
– such as ceramic resonators or MEMS
(micro-electromechanical systems) – are
becoming increasingly common.
Theoretically, oven-controlled MEMSbased oscillators should be called OCMOs,
but that’s one battle no one in the industry appears prepared to fight. Instead,
they refer to these bodacious beauties
as MEMS OCXOs, and I cannot find it
in my heart to fault them.
The reason I’m waffling on about all
this is that I was recently chatting with
the folks at SiTime. These little scamps
have just introduced their Epoch MEMS
OCXOs, which are truly OCXOs for the
21st Century. These silicon-based devices
– which have a frequency stability of 1
ppb and an internal oven temperature of
95°C – are claimed to be eight-times more
consistent, two-times more resilient, use
three-times lower power, 30-times more
reliable, and 25-times smaller than their
legacy quartz-based OCXO counterparts.
How low can we go?
The term ‘technology node’ (a.k.a. ‘process technology,’ ‘process node,’ or just
‘node’) refers to a specific semiconductor
manufacturing process. The first ASIC I
designed deep in the mists of time we
used to call 1980 was a device at the
5-micron (5µm) technology node.
In those days, depending on who you
were talking to, the numerical qualifier
referred to the width of a track or the
length of the channel between the source
and drain diffusion regions of a field-effect transistor (FET). I typically think of
this number as reflecting the size of the
smallest structure that can be created
in or on the surface of the silicon chip.
Every time we move to a new technology node, we either reduce the area
used or increase the number of transistors that can be squeezed into the same
area. We also increase the speed of the
transistors while reducing the amount
of power they consume.
We started creating devices at the 1µm
technology node circa 1985, where 1µm
is 100th the diameter of a human hair
(assuming a human hair has a diameter
of 0.1mm). At that time, the naysayers
started to proclaim that we were reaching the limits of what was possible. But
we kept on overcoming problems and
coming up with new solutions, and we
started to describe nodes in terms of
nanometres (nm).
I remember the progression well:
800nm in 1987, 600nm in 1990, 350nm
in 1993, 250nm in 1996, 180nm in 1999,
and 130nm in 2001. Surely this was as low
as we could go… but no! We saw 90nm
in 2003, 65nm in 2005, 45nm in 2007,
32nm in 2009, 22nm in 2012, 14nm in
2014, 10nm in 2016, 7nm in 2018, and
5nm in 2020. Apple’s latest processor,
the M3 is built with 3nm technology –
it’s most advanced version, the M3 Max,
boasts 92 billion transistors. TSMC, the
Taiwanese world leader in chip fabrication plans on introducing its 2nm node
in 2025/2026, and pundits are predicting
the 1nm node in 2028. (For comparison’s sake, the atomic radius of silicon is
0.132nm, so we are talking about structures just a few times bigger than the
atoms used to build them.)
All I can say is the thought that we are
now capable of creating multi-billiontransistor devices with structures whose
sizes are measured in billionths of a metre makes my eyes water.
Practical Electronics | January | 2024
|