HHD

Imagine every picture you took over the past year, all saved within the area of a single dot of ink
from a ballpoint pen. Well, this is approximately how compact data is stored inside a hard disk drive,
and in this video, we’re going to open one up and see how an entire library worth of books is able to
fit within the surface of this metal disk. We’ll start by opening up this hard drive and detailing the
components inside. After that, we’ll dive into exactly how the drive stores data using the read and
write head, and in the process, we’ll look at the tracks, sectors, and magnetic domains of the metal
disk. Finally, we’ll explore some of the latest advances that enable over a terabit of data to fit within
every square inch of the disk. This video is sponsored by PCBWay; more on them later, but for now
let’s jump right in. On the inside of this drive, we find a variety of components. Here’s the disk or
platter that stores all the data, and, depending on the storage capacity of the drive, might be
multiple platters tall. The disk is composed of an aluminum magnesium alloy with multiple coatings
of other alloys, but the magnetic, functional layer is this 120-nanometer thin layer of a cobalt
chromium tantalum alloy which has small magnetic domains or regions whose direction can be
manipulated via external magnetic fields. The platter is mounted on a spindle which spins at a
speed of 7200 rpm using a brushless DC motor at its center. Next, there’s a head stack assembly,
with one arm above and one arm below each disc, and with a slider and a read/write head at the
end of each arm. The slider is uniquely designed such that it catches the airflow generated by the
ludicrously fast-spinning disk and uses the air flow to float or fly the read/write head so that it’s only
15 nanometers or about 100 atoms away from the surface of the disk. For reference, here’s the
thickness of a sheet of aluminum foil. Because the arm assembly flies on top of the spinning disk,
it’s only brought over the surface when it’s at full speed, and when the disk is not spinning, the arm
assembly is parked to the side on a small piece of plastic.

Back here a voice coil motor composed of a coil of wire and two strong neodymium magnets above
and below is used to move the entire arm stack assembly. When electric current is run through the
coil it creates an electromagnet which is influenced by the neodymium magnets, thus generating a
force that causes the arm to move across the disk. When a reverse current is sent through the voice
coil, the arm is forced in the opposite direction, thereby enabling control of the exact position of the
read/write head within 30 or so nanometers. Additionally, the magnets and voice coil make a rather
strong motor that enable the lightweight arm stack assembly and read/write head to move back and
forth to different tracks across the platter up to 20 times a second and to make small adjustments
incredibly fast. In order to connect to the read/write head, a flexible ribbon of wires is routed along
the side of the arm and down to this connector which feeds signals to the outside of the hard drive
enclosure and to the printed circuit board or PCB. On the PCB we have the main processor as well
as a DRAM chip, which is used as a scratchpad for the processor and a buffer for the incoming
and outgoing data. Additionally mounted on the PCB is a chip for controlling the voice coil and
brushless DC spindle motor, and then on the edge of the PCB is a SATA connector which connects
to the motherboard for communications and a separate connector which goes to the power supply.
Additional important components are the gasket that seals the disk from the exterior environment
and two filters that catch any stray dust particles. These filters are necessary since the read/write
heads are just 15 nanometers away from the platter, and a single dust particle can be up to 10,000
nanometers large and could cause major damage if it were to collide with the 7,200-rpm disk. Now
that we’ve looked through many of these components, let’s see how they work. To begin, the disk is
divided into concentric circles of tracks. The latest hard drives can have more than 500,000 tracks
on just one side. These tracks are then divided into sectors, and in each sector is a preamble or
synchronization zone which tells the read/write head the exact speed of the spinning disk and the
length of each bit of data. The next part of the sector is the address which helps the read/write head
know which track and sector it’s currently positioned over. After that we have the actual data that’s
stored, typically 4 kilobytes of data per sector. Next is an area for an error correcting code, or ECC,
which is used to verify that the data stored in the block is accurately written and properly read, and
finally there’s a gap between this sector and the next which allows the read/write head some
tolerance when writing the contents of a block. Now let’s zoom in on the read/write head and the disk
to see exactly how data is written and read. Writing data to the disk is done by manipulating the
direction of magnetization of a localized region or domain of the cobalt-chromium-tantalum layer in
the disk and forcing the region to be magnetized in the up direction or the down direction. This tiny
magnetic domain or region is around 90 by 100 by 125 nanometers, and when magnetized, all the
atoms will have their even tinier magnetic north/south poles pointing in the same direction. In order
to magnetize a single domain, which is equivalent to writing a single bit of data, a current is applied
to a coil of wire at the back of the write head, thus creating a strong magnetic field back here. The
magnetic field is channeled through the write head and focused into a small point at the tip and
then jumps across the 15-nanometer air gap and into the disk. When the focused magnetic field
passes into a single domain of cobalt-chromium-tantalum, all these atoms are forced to align their
tiny atomic magnetic fields with the applied magnetic field from the write head, thus turning the small
domain or region into a permanent magnet. The key is that even when the write head is moved
away, the direction of the magnetic domains in this layer of the disk is maintained for years, and
they emit a permanent magnetic field which can be repeatedly sensed by the read head every time
you read out the data stored in this sector. That is, of course, until the computer and write head
rewrite a new bit of data to the domain by either flipping the direction or keeping it the same. Let’s
explore how we read data from the disk.

Thus far we’ve been showing domains as pointing up as a binary 1 and pointing down as a 0. While
this is conceptually simple, it isn’t actually the case. Rather the read head is designed to detect the
changes in orientation from magnetic domains pointing in one direction and then the adjacent
domain pointing in the opposite direction. This is because emitted magnetic fields from adjacent
regions that switch their orientations are much stronger than the emitted field from just a single
domain pointing one direction or the other. Therefore, each change in magnetic domain pointing in
one direction to the opposite direction is assigned a 1, and an absence of a transition from one
domain to the next is assigned a 0. Therefore, the write head would record a binary sequence of
0011 0010 like this. Or a sequence of 1101 1110 like this, where the 1’s are changes, and the 0’s are
lack of changes. So then, what’s inside the read head that detects these magnetic fields? Well,
inside is a multilayer conductive material composed of alternating layers of ferromagnetic and non-
magnetic materials. This multilayer material has a property called giant magnetoresistance or GMR,
and, put simply, it’s a material that changes its resistivity depending on the strength of magnetic
fields that pass through it. Therefore, using GMR it’s a simple matter of just measuring the resistivity,
and when there’s a low resistivity that means there are strong magnetic fields below the read head
resulting from a change in domain orientation and it’s a 1, and when there’s high resistivity and no
change it’s a 0. However, this poses the problem of how a string of dozens of non-changing domains
can result in an ambiguous number of zeroes. To fix this, in each sector the preamble is simply a
set of alternating domains and is used to set the size of each domain, and then the error correcting
code at the back of the sector is used to ensure no data is lost. Next, we’re going to explore some
advancements in hard disk drive technology that improve the areal density, which is the number of
bits that can fit within a given area. In this graph, you can see how areal density has increased by
over 50 million times throughout the past 60 years. However, perhaps more important is that the
cost to store trillions of bits of data has dropped by over 100 million times. Just imagine, if we were
to time travel this hard drive back to the 1960s, it would be worth over 4 billion dollars, and now in
2022 it costs less than 40 dollars and is far faster and more reliable than the disk drives from the
60s! Similar to the trend in hard drives, you can buy inexpensive, and yet incredibly reliable printed
circuit boards from our sponsor PCBWay. Whether you’re prototyping your next project, or ready to
mass produce thousands of your finalized devices, PCBWay can quickly manufacture your PCBs
with competitive prices and impeccable standards. Additionally, if you don’t want to spend weeks
soldering all the components to every board,PCBWay provides PCB assembly services where they
populate and solder the components to the PCB for you. They’ll send pictures throughout the
assembly process, and you can work directly with PCBWay’s engineers to provide programming and
testing protocols. The next time you have a project and want to save both time and money, consider
using PCBWay to manufacture and populate all your Printed Circuit Boards.

SSD
However as seeing the outside of this memory storage microchip tells us little about how
these smartphones and solid-state drives can store tens of thousands of photos and files,
let’s explore deeper and zoom in until we get to a nanoscopic view, and it's here that we can
see the structures called VNAND that hold all the data in your smartphone and computer.
Here is where the real magic happens. Every picture, message, and bit of information gets
saved as quantities of electrons inside these memory cells which are called charge trap flash
and, in this episode, we'll learn how smartphone memory and solid-state drives work. Now,
hold on- these insanely small and intricate structures seem very complex, and yeah- they
are- I’m not going to say this marvel of engineering is simple. But you have to trust me- stick
around, watch closely, maybe watch this video twice, and by the end of it, this technology will
amaze you, it will blow your mind at least twice over, and yeah, you'll have a thorough
understanding as to how such a small device, can store weeks of high quality video, tens of
thousands of pictures, or hundreds of thousands of songs in such an itty bitty little space. So,
let’s get started. We’re going to use a real-life example and explore how it works when you
save a picture to your smartphone or computer. First, this picture is made up of pixels and
each pixel has a color so let’s zoom in so that we can see the individual pixels. The color of
every pixel is defined by a combination of 3 numbers, ranging from 0 to 255, each
representing red, green, or blue.
For example, the numbers would be 55-53-55 for this pixel’s color right here, and then 124-
121-119 for this pixel. Each of these 3 numbers from 0 to 255 is represented by 8 bits in
binary, or eight ones and zeros ya know, because computers work in binary. So, 3 colors,
red, green and blue, and 8 bits each, means each pixel takes 24 bits to define its color. This
picture is a grid of colored pixels, so let’s turn it into a grid of values, kind of like a
spreadsheet in excel, but called an array instead of a spreadsheet. This array of bits is what
your computer cares about and noncoincidentally, it’s also the information that the camera
on my smartphone recorded when I took the picture. One quick note: if you want to see the
pixels in any picture, just open it in an image editing program like paint or 3D paint in this
case, and zoom in. And then if you want to see the red, green and blue or RGB values, just
use the eye dropper, click on a pixel, and then click on the edit color option. Right here you
can see the 3 values for red, green, and blue, and the resulting color. Ok, with that covered,
let’s get back to this episode, first, we’re gonna zoom out to see the full picture, which is
3024 pixels wide and 4032 pixels tall, which is a total of around 12 million pixels, or, 12
megapixels- which relates to the resolution of the 12 megapixel camera on my smartphone.
Next, by doing some multiplication we calculate that an array of this size, where each pixel is
defined by 24 bits, or 24 0s or 1s only requires 293 million bits or a unique set of 293 million
0s or 1s. That’s a ton of bits, so let’s figure out how your smartphone or this solid-state drive
seamlessly stores every single one of them. Ok: so let’s open up that solid state drive again
and zoom into a simplified nanoscopic view kind of like the one we had earlier. It's here that
we can see the memory cells that are used in every single one of your smartphones or
tablets, as well as inside the solid-state drive in your computer. This is the basic unit of a
computer’s long term memory storage and it’s called Charge Trap Flash Memory- so how
does it work? Well, in each cell we can store information by placing different levels of
electrons onto a charge trap, which is the key component inside the memory cell. Older
technology could only store two different levels of electrons, a lot of electrons or very few
electrons, which were used to store a single bit as a 1 or a 0. However, engineers have been
developing more finely tuned capabilities for trapping and measuring different amounts of
electrons or charges onto the charge trap. Most memory cells in 2020 can hold 8 different
levels, but newer technology can have 16 different levels of electrons. This means that a
single cell, instead of holding only one bit as a lot of electrons or no electrons, can now hold
3 or more bits but, for this example, let’s stick with 3 bits. So- in this cell, if we were to have
very few electrons on it, it would be 1-1-1, while some electrons get designated as 1-0-0 and
a lot of electrons are 0-0-0 There are 8 different levels for all the various amounts of electron
charges that our charge trap can be set or written to. The key to the charge trap is that it is
specially designed so that after it gets charged with electrons, it can hold onto those
electrons for decades, which is how information is saved or written to the solid-state drive. I
mean- it’s called a charge trap for a reason. It traps electrons, or charges for years on end,
and in order to read the information, the electron charge level is measured, and the amount
of charge on the charge trap is unchanged. However, in order to erase the contents of a
memory cell, all the electron charges are forcibly removed from the charge trap returning it to
its lowest level, which is 1-1-1, and leaving no excess electron charges behind. Let’s move
on and explore how these memory cells are organized so that we can store more than just 3
bits of information. After we zoom out a little, you can see that the memory cells are stacked
vertically. This is where the vertical part in Vertical NAND or VNAND comes from.
This stack of memory cells, what is technically called a string is composed of 10 charge trap
flash cells layered one top of another. when information is written to or read from a string,
only one cell can be activated at any given time, and to do that we use separate control
gates attached to every layer in the string. It works like this: the bottom control gate first says
“Hey you, charge trap 1 what’s your electron charge level at?” Then the bottom cell sends
that information through the center of the string up to the information highway at the top,
which is technically called a bitline. Then the next control gate for the 2nd layer asks for the
charge level in the 2nd cell, and so on, up the string, each cell sending their information up
to the highway or bitline. The same kinda sequence happens when charges are being added
to a charge trap which is how information is written to a memory cell. The main thing is that
only one layer in the string is either written to or read from at any given time. Let’s move on
in complexity, next we duplicate this string 32 times, and this gets us a page of strings. Let’s
review some terminology: this a memory cell and this is a string. And now here we have a
page, and we are going to call this entire page of strings a row. When we duplicate the
string, we also duplicate the bitline 32 times, however rather than duplicate the control gates,
we are going to have every cell in the same page share a common control gate. This makes
it such that when information is written to or read from a row, an entire page composed of 32
adjacent cells, all in the same layer, are activated at the same time. Let’s step up in
complexity again: Next, we duplicate these rows 6 times until we get a block, but we are
going to do it 12 times so we can see 2 blocks. Okay, so again, here we have a column, here
is a row, this is a layer. And now here is a cell and here is a string. Next we have a page, and
finally we have a block. We are going to connect the tops of each string in a column
together, so they all share the same bitline and our bitline is looking more like a highway
now. In addition, we have to add a control gate that selects between rows, so that only one
row is using the bitline at a time. These are called bitline selectors. As discussed these
bitlines are like highways, and the selectors at the top act as traffic lights that mediate the
flow of information so that only a single row can use the highway, or is active at a time.
Similarly, the control gates attached to each layer act as traffic lights for the layers. With
bitline selectors along the tops of each row, and control gate selectors along each layer, the
solid state drive can read from or write to a single page at any given time. Additionally, in
order to connect to the bitline selectors and control gate selectors there are wires that drop
down from above and run perpendicular to the bitlines. So, let’s quickly recap: 8 different
levels of electrons are placed on charge traps in order to store 3 bits of information. These
charge trap flash memory cells are stacked into strings 10 cells tall, which are duplicated into
pages of 32 strings in a row. Next, those pages of strings are duplicated until we have a
block 6 rows deep, and here we are showing 2 blocks. Doing some quick multiplication we
find that there are 3,840 memory cells here capable of storing a total of 11,520 bits. With
each pixel in our picture requiring 24 bits, that means that we can store 480 pixels, or this
much of our overall picture. That means you need about 25 thousand times the size of this
layout to store the contents of this single picture. Aaand, here’s where we learn about the
actual size of a memory chip. All the principles we have discussed remain the same, so keep
those in mind, it’s just that the size is much more extensive than we discussed in our
example. It’s hard to pin down exact numbers because manufacturers are continually
improving their designs and they are very secretive regarding what their designs look like.
But I’ll tell you what I know: the latest designs utilize not 10 layers as in the example, but
rather somewhere around 96 to 136 layers tall. Here's a single sheet of paper so you can get
a sense of the of the approximiate height of these stacks of memory cells. Now that we
understand the height, lets think about the width.
A page is around 30,000 to 60,000 adjacent memory cells wide. That means there are
30,000 to 60,000 bitlines in our information superhighway. Blocks are every 4 to 8 rows and
there are around 4000 to 6000 blocks. Along the edges are the control gate selectors and
the bitline selectors on the other side. Together, they comprise what is called a row decoder,
and by using both sets of selectors as traffic lights, we're able to accesss a single page. To
repeat this, only one page, 45 thousand or so cells wide, ever uses the bitline to read or
write information at any given time All tens of thousands of bitlines feed down here to the
page buffer where the information from a single page is written to or read from. Let’s
transition to see what an overall chip might look like. Here we have the arrays of 3D memory
cells, the row decoder and the page buffer at the bottom. Additional peripheral circuitry can
be found here for supporting the chip. In order to fit more capacity, engineers copied this
layout onto the other side. This chip can read or write at a rate of around 500megabytes per
second. That means that it can read from or write to around 63 blocks every single second.
That’s incredibly fast! Ok, let’s add the last level of complexity. Engineers like to fit even
more stuff in as small of a space as possible, so on top of having a massive array of memory
cells in this insanely complex layout, they decided to copy this chip 8 times. and stack it into
a single microchip. At the bottom, an additional interface chip is used to coordinate between
the 8 different chips. And that’s it, that’s all there is in this one microchip that can found at the
center of every one of your smartphones, tablets, or solid-state drives. This video covered a
lot, and I hope you kept up. You can always watch this video a second time, and if you do
watch it a second time, we added our notes and commentary into the English Canada
subtitles. Turn them on by clicking the settings gear over here. On the contrary the notes that
are placed up here are caveats or footnotes, but the notes we placed in the English Canada
subtitles include commentary, additional information, and much more. Let us know what you
think of them in the comments Also, I will be making a follow up set of episodes that will
branch off and explain how each part works in detail. In separate episodes we'll cover
specifics as to how the charge trap flash works, how the bitline and control gate selectors
work, and how these microchips are manufactured. Also, take a look at our channel page
where we cover other topics such as how touchscreens work, how PCBs work, or how
cameras in your smartphone work. If you have any questions or want me to add more
branches relating to solid state drives.